diff --git "a/checkpoint-34000/trainer_state.json" "b/checkpoint-34000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-34000/trainer_state.json"
@@ -0,0 +1,40969 @@
+{
+  "best_metric": 0.05467037484049797,
+  "best_model_checkpoint": "wav2vec2-base-pemlsb-la/checkpoint-34000",
+  "epoch": 0.9434224035073115,
+  "global_step": 34000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 0.2249,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 0.2536,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 0.1986,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 0.2285,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 0.1936,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 0.0825,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 0.1878,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 0.1953,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 0.3766,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 0.5279,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.5e-07,
+      "loss": 0.2282,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 0.1752,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5e-07,
+      "loss": 0.2045,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 0.2753,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.5e-07,
+      "loss": 0.0828,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 0.1613,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 0.2658,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.900000000000001e-07,
+      "loss": 0.1503,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.400000000000001e-07,
+      "loss": 0.2618,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.9e-07,
+      "loss": 0.2933,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.04e-06,
+      "loss": 0.1629,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0900000000000002e-06,
+      "loss": 0.1351,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.14e-06,
+      "loss": 0.1777,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.19e-06,
+      "loss": 0.1647,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2400000000000002e-06,
+      "loss": 0.107,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2900000000000001e-06,
+      "loss": 0.2022,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.34e-06,
+      "loss": 0.562,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3900000000000002e-06,
+      "loss": 0.2811,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.44e-06,
+      "loss": 0.2064,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.48e-06,
+      "loss": 0.4073,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5300000000000002e-06,
+      "loss": 0.1724,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 0.2262,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6300000000000003e-06,
+      "loss": 0.3025,
+      "step": 165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6800000000000002e-06,
+      "loss": 0.3449,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.73e-06,
+      "loss": 0.089,
+      "step": 175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7800000000000001e-06,
+      "loss": 0.2062,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.83e-06,
+      "loss": 0.1239,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8800000000000002e-06,
+      "loss": 0.215,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.93e-06,
+      "loss": 0.2719,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.98e-06,
+      "loss": 0.5067,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0300000000000005e-06,
+      "loss": 0.2739,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.08e-06,
+      "loss": 0.1794,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.13e-06,
+      "loss": 0.2751,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1800000000000003e-06,
+      "loss": 0.185,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2300000000000002e-06,
+      "loss": 0.0968,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.28e-06,
+      "loss": 0.2331,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.33e-06,
+      "loss": 0.1052,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.38e-06,
+      "loss": 0.3041,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.43e-06,
+      "loss": 0.3321,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4800000000000004e-06,
+      "loss": 0.6221,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5300000000000003e-06,
+      "loss": 0.1997,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 0.1825,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6300000000000002e-06,
+      "loss": 0.2278,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.68e-06,
+      "loss": 0.2432,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7300000000000005e-06,
+      "loss": 0.1503,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7800000000000005e-06,
+      "loss": 0.1733,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.83e-06,
+      "loss": 0.0788,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.88e-06,
+      "loss": 0.2403,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.93e-06,
+      "loss": 0.1724,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9800000000000003e-06,
+      "loss": 0.5096,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0300000000000002e-06,
+      "loss": 0.2124,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.08e-06,
+      "loss": 0.2383,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.13e-06,
+      "loss": 0.1581,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1800000000000005e-06,
+      "loss": 0.2122,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2300000000000004e-06,
+      "loss": 0.2645,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2800000000000004e-06,
+      "loss": 0.1838,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3300000000000003e-06,
+      "loss": 0.1749,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3800000000000007e-06,
+      "loss": 0.1447,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4300000000000006e-06,
+      "loss": 0.2553,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.48e-06,
+      "loss": 0.5503,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.53e-06,
+      "loss": 0.2164,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.58e-06,
+      "loss": 0.1653,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6300000000000004e-06,
+      "loss": 0.1861,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6800000000000003e-06,
+      "loss": 0.2385,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7300000000000003e-06,
+      "loss": 0.2468,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7800000000000002e-06,
+      "loss": 0.1379,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.830000000000001e-06,
+      "loss": 0.0878,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.88e-06,
+      "loss": 0.2242,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9300000000000005e-06,
+      "loss": 0.3307,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.980000000000001e-06,
+      "loss": 0.4413,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.03e-06,
+      "loss": 0.1462,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.08e-06,
+      "loss": 0.205,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.13e-06,
+      "loss": 0.2098,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.18e-06,
+      "loss": 0.191,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.23e-06,
+      "loss": 0.0629,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2800000000000005e-06,
+      "loss": 0.1265,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.33e-06,
+      "loss": 0.2014,
+      "step": 435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.38e-06,
+      "loss": 0.3309,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.430000000000001e-06,
+      "loss": 0.3537,
+      "step": 445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.48e-06,
+      "loss": 0.4845,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.530000000000001e-06,
+      "loss": 0.2915,
+      "step": 455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.58e-06,
+      "loss": 0.233,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6300000000000006e-06,
+      "loss": 0.1498,
+      "step": 465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.680000000000001e-06,
+      "loss": 0.1597,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7300000000000005e-06,
+      "loss": 0.182,
+      "step": 475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.78e-06,
+      "loss": 0.2874,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.83e-06,
+      "loss": 0.1666,
+      "step": 485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.880000000000001e-06,
+      "loss": 0.3407,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.93e-06,
+      "loss": 0.2495,
+      "step": 495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.980000000000001e-06,
+      "loss": 0.443,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.03e-06,
+      "loss": 0.1806,
+      "step": 505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0800000000000005e-06,
+      "loss": 0.217,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.130000000000001e-06,
+      "loss": 0.1761,
+      "step": 515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.18e-06,
+      "loss": 0.305,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.230000000000001e-06,
+      "loss": 0.1023,
+      "step": 525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.28e-06,
+      "loss": 0.0879,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.330000000000001e-06,
+      "loss": 0.1694,
+      "step": 535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.380000000000001e-06,
+      "loss": 0.1529,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4300000000000005e-06,
+      "loss": 0.3211,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.470000000000001e-06,
+      "loss": 0.6838,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5200000000000005e-06,
+      "loss": 0.2006,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.570000000000001e-06,
+      "loss": 0.1613,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.620000000000001e-06,
+      "loss": 0.2125,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.67e-06,
+      "loss": 0.2194,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.72e-06,
+      "loss": 0.2017,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.77e-06,
+      "loss": 0.0922,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.82e-06,
+      "loss": 0.1526,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8700000000000005e-06,
+      "loss": 0.2183,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.92e-06,
+      "loss": 0.197,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9700000000000004e-06,
+      "loss": 0.5106,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.02e-06,
+      "loss": 0.2169,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.07e-06,
+      "loss": 0.1802,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.120000000000001e-06,
+      "loss": 0.2202,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.17e-06,
+      "loss": 0.2362,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.220000000000001e-06,
+      "loss": 0.1626,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.27e-06,
+      "loss": 0.1174,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3200000000000005e-06,
+      "loss": 0.1051,
+      "step": 635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.370000000000001e-06,
+      "loss": 0.1218,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.42e-06,
+      "loss": 0.2457,
+      "step": 645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.470000000000001e-06,
+      "loss": 0.3494,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.520000000000001e-06,
+      "loss": 0.2136,
+      "step": 655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.570000000000001e-06,
+      "loss": 0.2451,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.620000000000001e-06,
+      "loss": 0.2849,
+      "step": 665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.6700000000000005e-06,
+      "loss": 0.2196,
+      "step": 670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.720000000000001e-06,
+      "loss": 0.2496,
+      "step": 675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.770000000000001e-06,
+      "loss": 0.0549,
+      "step": 680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.820000000000001e-06,
+      "loss": 0.0748,
+      "step": 685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.870000000000001e-06,
+      "loss": 0.2986,
+      "step": 690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.92e-06,
+      "loss": 0.2527,
+      "step": 695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.97e-06,
+      "loss": 0.5124,
+      "step": 700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0200000000000006e-06,
+      "loss": 0.2632,
+      "step": 705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.07e-06,
+      "loss": 0.204,
+      "step": 710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.1200000000000004e-06,
+      "loss": 0.1899,
+      "step": 715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.17e-06,
+      "loss": 0.2731,
+      "step": 720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.22e-06,
+      "loss": 0.2023,
+      "step": 725
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.270000000000001e-06,
+      "loss": 0.1433,
+      "step": 730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.32e-06,
+      "loss": 0.1118,
+      "step": 735
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.370000000000001e-06,
+      "loss": 0.2098,
+      "step": 740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.420000000000001e-06,
+      "loss": 0.1431,
+      "step": 745
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.4700000000000005e-06,
+      "loss": 0.3622,
+      "step": 750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.520000000000001e-06,
+      "loss": 0.1973,
+      "step": 755
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.57e-06,
+      "loss": 0.1582,
+      "step": 760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.620000000000001e-06,
+      "loss": 0.2339,
+      "step": 765
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.670000000000001e-06,
+      "loss": 0.2483,
+      "step": 770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.72e-06,
+      "loss": 0.0592,
+      "step": 775
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.77e-06,
+      "loss": 0.1738,
+      "step": 780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.820000000000001e-06,
+      "loss": 0.1896,
+      "step": 785
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.870000000000001e-06,
+      "loss": 0.2247,
+      "step": 790
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.92e-06,
+      "loss": 0.4066,
+      "step": 795
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.970000000000002e-06,
+      "loss": 0.6385,
+      "step": 800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.020000000000001e-06,
+      "loss": 0.1883,
+      "step": 805
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.07e-06,
+      "loss": 0.2734,
+      "step": 810
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.120000000000002e-06,
+      "loss": 0.2096,
+      "step": 815
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.17e-06,
+      "loss": 0.1888,
+      "step": 820
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.220000000000001e-06,
+      "loss": 0.1583,
+      "step": 825
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.27e-06,
+      "loss": 0.097,
+      "step": 830
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.32e-06,
+      "loss": 0.1855,
+      "step": 835
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.370000000000001e-06,
+      "loss": 0.1659,
+      "step": 840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.42e-06,
+      "loss": 0.3064,
+      "step": 845
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.47e-06,
+      "loss": 0.5692,
+      "step": 850
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.52e-06,
+      "loss": 0.1638,
+      "step": 855
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.570000000000001e-06,
+      "loss": 0.2457,
+      "step": 860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.62e-06,
+      "loss": 0.2183,
+      "step": 865
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.67e-06,
+      "loss": 0.2386,
+      "step": 870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.720000000000001e-06,
+      "loss": 0.1225,
+      "step": 875
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.77e-06,
+      "loss": 0.2076,
+      "step": 880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.82e-06,
+      "loss": 0.2056,
+      "step": 885
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.870000000000001e-06,
+      "loss": 0.3954,
+      "step": 890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.920000000000001e-06,
+      "loss": 0.1682,
+      "step": 895
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.97e-06,
+      "loss": 0.6697,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.020000000000002e-06,
+      "loss": 0.1666,
+      "step": 905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.070000000000001e-06,
+      "loss": 0.2045,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.12e-06,
+      "loss": 0.2174,
+      "step": 915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.17e-06,
+      "loss": 0.2384,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.220000000000002e-06,
+      "loss": 0.2176,
+      "step": 925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.270000000000001e-06,
+      "loss": 0.136,
+      "step": 930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.32e-06,
+      "loss": 0.2383,
+      "step": 935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.370000000000002e-06,
+      "loss": 0.1341,
+      "step": 940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.42e-06,
+      "loss": 0.2451,
+      "step": 945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.47e-06,
+      "loss": 0.9591,
+      "step": 950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.52e-06,
+      "loss": 0.1867,
+      "step": 955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.57e-06,
+      "loss": 0.1873,
+      "step": 960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.620000000000001e-06,
+      "loss": 0.1948,
+      "step": 965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.67e-06,
+      "loss": 0.2792,
+      "step": 970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.72e-06,
+      "loss": 0.3987,
+      "step": 975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.770000000000001e-06,
+      "loss": 0.039,
+      "step": 980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.820000000000001e-06,
+      "loss": 0.2448,
+      "step": 985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.87e-06,
+      "loss": 0.1216,
+      "step": 990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920000000000002e-06,
+      "loss": 0.249,
+      "step": 995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970000000000001e-06,
+      "loss": 0.433,
+      "step": 1000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999429207454551e-06,
+      "loss": 0.1561,
+      "step": 1005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.998002226090927e-06,
+      "loss": 0.1989,
+      "step": 1010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.996575244727305e-06,
+      "loss": 0.1724,
+      "step": 1015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.995148263363682e-06,
+      "loss": 0.1805,
+      "step": 1020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.993721282000058e-06,
+      "loss": 0.1212,
+      "step": 1025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.992294300636434e-06,
+      "loss": 0.1385,
+      "step": 1030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99086731927281e-06,
+      "loss": 0.1,
+      "step": 1035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.989440337909187e-06,
+      "loss": 0.1665,
+      "step": 1040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.988013356545563e-06,
+      "loss": 0.3426,
+      "step": 1045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.986586375181941e-06,
+      "loss": 0.5495,
+      "step": 1050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985159393818318e-06,
+      "loss": 0.1594,
+      "step": 1055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983732412454694e-06,
+      "loss": 0.1865,
+      "step": 1060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98230543109107e-06,
+      "loss": 0.2308,
+      "step": 1065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980878449727447e-06,
+      "loss": 0.1686,
+      "step": 1070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979451468363825e-06,
+      "loss": 0.2432,
+      "step": 1075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978024487000201e-06,
+      "loss": 0.1348,
+      "step": 1080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976597505636578e-06,
+      "loss": 0.1699,
+      "step": 1085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975170524272954e-06,
+      "loss": 0.1712,
+      "step": 1090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97374354290933e-06,
+      "loss": 0.298,
+      "step": 1095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972316561545707e-06,
+      "loss": 0.5211,
+      "step": 1100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970889580182085e-06,
+      "loss": 0.2375,
+      "step": 1105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.969462598818461e-06,
+      "loss": 0.2803,
+      "step": 1110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.968035617454837e-06,
+      "loss": 0.2163,
+      "step": 1115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.966608636091214e-06,
+      "loss": 0.2415,
+      "step": 1120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.96518165472759e-06,
+      "loss": 0.2971,
+      "step": 1125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.963754673363966e-06,
+      "loss": 0.0985,
+      "step": 1130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.962327692000343e-06,
+      "loss": 0.1321,
+      "step": 1135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.960900710636719e-06,
+      "loss": 0.1606,
+      "step": 1140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.959473729273097e-06,
+      "loss": 0.2521,
+      "step": 1145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.958046747909473e-06,
+      "loss": 0.6916,
+      "step": 1150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.95661976654585e-06,
+      "loss": 0.1813,
+      "step": 1155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.955192785182226e-06,
+      "loss": 0.2017,
+      "step": 1160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.953765803818602e-06,
+      "loss": 0.2034,
+      "step": 1165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.95233882245498e-06,
+      "loss": 0.1885,
+      "step": 1170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.950911841091357e-06,
+      "loss": 0.2131,
+      "step": 1175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.949484859727733e-06,
+      "loss": 0.0847,
+      "step": 1180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.94805787836411e-06,
+      "loss": 0.1071,
+      "step": 1185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.946630897000486e-06,
+      "loss": 0.2301,
+      "step": 1190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.945203915636864e-06,
+      "loss": 0.1886,
+      "step": 1195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.94377693427324e-06,
+      "loss": 0.4755,
+      "step": 1200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.942349952909616e-06,
+      "loss": 0.1931,
+      "step": 1205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.941208367818717e-06,
+      "loss": 3.617,
+      "step": 1210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.939781386455095e-06,
+      "loss": 0.2011,
+      "step": 1215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.938354405091471e-06,
+      "loss": 0.1695,
+      "step": 1220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.936927423727848e-06,
+      "loss": 0.2074,
+      "step": 1225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.935500442364224e-06,
+      "loss": 0.1075,
+      "step": 1230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9340734610006e-06,
+      "loss": 0.1331,
+      "step": 1235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.932646479636977e-06,
+      "loss": 0.2502,
+      "step": 1240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.931219498273353e-06,
+      "loss": 0.2703,
+      "step": 1245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92979251690973e-06,
+      "loss": 0.386,
+      "step": 1250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928365535546106e-06,
+      "loss": 0.2368,
+      "step": 1255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926938554182484e-06,
+      "loss": 0.2148,
+      "step": 1260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.92551157281886e-06,
+      "loss": 0.2385,
+      "step": 1265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.924084591455236e-06,
+      "loss": 0.1964,
+      "step": 1270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.922657610091613e-06,
+      "loss": 0.1165,
+      "step": 1275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.921230628727989e-06,
+      "loss": 0.1667,
+      "step": 1280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.919803647364367e-06,
+      "loss": 0.2257,
+      "step": 1285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.918376666000743e-06,
+      "loss": 0.2837,
+      "step": 1290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.91694968463712e-06,
+      "loss": 0.5581,
+      "step": 1295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.915522703273496e-06,
+      "loss": 0.731,
+      "step": 1300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.914095721909872e-06,
+      "loss": 0.2324,
+      "step": 1305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.91266874054625e-06,
+      "loss": 0.2118,
+      "step": 1310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.911241759182627e-06,
+      "loss": 0.2389,
+      "step": 1315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.909814777819003e-06,
+      "loss": 0.2752,
+      "step": 1320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.90838779645538e-06,
+      "loss": 0.2481,
+      "step": 1325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.906960815091756e-06,
+      "loss": 0.0591,
+      "step": 1330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.905533833728132e-06,
+      "loss": 0.1931,
+      "step": 1335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.904106852364508e-06,
+      "loss": 0.2633,
+      "step": 1340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.902679871000885e-06,
+      "loss": 0.4182,
+      "step": 1345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.901252889637261e-06,
+      "loss": 0.4839,
+      "step": 1350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.899825908273639e-06,
+      "loss": 0.2509,
+      "step": 1355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.898398926910015e-06,
+      "loss": 0.251,
+      "step": 1360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.896971945546392e-06,
+      "loss": 0.1829,
+      "step": 1365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.895544964182768e-06,
+      "loss": 0.2305,
+      "step": 1370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894117982819144e-06,
+      "loss": 0.2734,
+      "step": 1375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892691001455522e-06,
+      "loss": 0.1588,
+      "step": 1380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891264020091899e-06,
+      "loss": 0.1621,
+      "step": 1385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.889837038728275e-06,
+      "loss": 0.3336,
+      "step": 1390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888410057364652e-06,
+      "loss": 0.26,
+      "step": 1395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886983076001028e-06,
+      "loss": 0.3407,
+      "step": 1400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885556094637406e-06,
+      "loss": 0.2269,
+      "step": 1405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884129113273782e-06,
+      "loss": 0.2338,
+      "step": 1410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882702131910159e-06,
+      "loss": 0.175,
+      "step": 1415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881275150546535e-06,
+      "loss": 0.2675,
+      "step": 1420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879848169182911e-06,
+      "loss": 0.2215,
+      "step": 1425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878421187819288e-06,
+      "loss": 0.1028,
+      "step": 1430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876994206455664e-06,
+      "loss": 0.1014,
+      "step": 1435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87556722509204e-06,
+      "loss": 0.195,
+      "step": 1440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874140243728417e-06,
+      "loss": 0.3388,
+      "step": 1445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872713262364795e-06,
+      "loss": 0.3025,
+      "step": 1450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871286281001171e-06,
+      "loss": 0.1817,
+      "step": 1455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869859299637547e-06,
+      "loss": 0.2523,
+      "step": 1460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868432318273924e-06,
+      "loss": 0.1917,
+      "step": 1465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8670053369103e-06,
+      "loss": 0.2508,
+      "step": 1470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865578355546678e-06,
+      "loss": 0.0938,
+      "step": 1475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864151374183054e-06,
+      "loss": 0.1335,
+      "step": 1480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86272439281943e-06,
+      "loss": 0.2164,
+      "step": 1485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.861297411455807e-06,
+      "loss": 0.2721,
+      "step": 1490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.859870430092183e-06,
+      "loss": 0.3324,
+      "step": 1495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.858443448728561e-06,
+      "loss": 0.6563,
+      "step": 1500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.857016467364938e-06,
+      "loss": 0.205,
+      "step": 1505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.855589486001314e-06,
+      "loss": 0.1993,
+      "step": 1510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.85416250463769e-06,
+      "loss": 0.1924,
+      "step": 1515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.852735523274067e-06,
+      "loss": 0.2587,
+      "step": 1520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.851308541910443e-06,
+      "loss": 0.1098,
+      "step": 1525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.84988156054682e-06,
+      "loss": 0.0455,
+      "step": 1530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.848454579183196e-06,
+      "loss": 0.1789,
+      "step": 1535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.847027597819572e-06,
+      "loss": 0.1117,
+      "step": 1540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.84560061645595e-06,
+      "loss": 0.2805,
+      "step": 1545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.844173635092326e-06,
+      "loss": 0.4923,
+      "step": 1550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.842746653728703e-06,
+      "loss": 0.2292,
+      "step": 1555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.841319672365079e-06,
+      "loss": 0.2311,
+      "step": 1560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.839892691001455e-06,
+      "loss": 0.2332,
+      "step": 1565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.838465709637834e-06,
+      "loss": 0.1955,
+      "step": 1570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.83703872827421e-06,
+      "loss": 0.1574,
+      "step": 1575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.835611746910586e-06,
+      "loss": 0.1634,
+      "step": 1580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.834184765546963e-06,
+      "loss": 0.3306,
+      "step": 1585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.832757784183339e-06,
+      "loss": 0.235,
+      "step": 1590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.831330802819717e-06,
+      "loss": 0.3012,
+      "step": 1595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.829903821456093e-06,
+      "loss": 0.4106,
+      "step": 1600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.82847684009247e-06,
+      "loss": 0.2449,
+      "step": 1605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.827049858728846e-06,
+      "loss": 0.2571,
+      "step": 1610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.825622877365222e-06,
+      "loss": 0.2535,
+      "step": 1615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.824195896001599e-06,
+      "loss": 0.2197,
+      "step": 1620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.822768914637975e-06,
+      "loss": 0.0862,
+      "step": 1625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.821341933274351e-06,
+      "loss": 0.1091,
+      "step": 1630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.819914951910728e-06,
+      "loss": 0.1436,
+      "step": 1635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.818487970547106e-06,
+      "loss": 0.2528,
+      "step": 1640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.817060989183482e-06,
+      "loss": 0.3583,
+      "step": 1645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.815634007819858e-06,
+      "loss": 0.4629,
+      "step": 1650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.814207026456235e-06,
+      "loss": 0.2162,
+      "step": 1655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.812780045092611e-06,
+      "loss": 0.2192,
+      "step": 1660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.811353063728989e-06,
+      "loss": 0.157,
+      "step": 1665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.809926082365365e-06,
+      "loss": 0.208,
+      "step": 1670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.808499101001742e-06,
+      "loss": 0.2118,
+      "step": 1675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.807072119638118e-06,
+      "loss": 0.1081,
+      "step": 1680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.805645138274496e-06,
+      "loss": 0.198,
+      "step": 1685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.804218156910872e-06,
+      "loss": 0.2535,
+      "step": 1690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.802791175547249e-06,
+      "loss": 0.3795,
+      "step": 1695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.801364194183625e-06,
+      "loss": 0.4864,
+      "step": 1700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.799937212820001e-06,
+      "loss": 0.1615,
+      "step": 1705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.798510231456378e-06,
+      "loss": 0.2312,
+      "step": 1710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.797083250092754e-06,
+      "loss": 0.2066,
+      "step": 1715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.79565626872913e-06,
+      "loss": 0.2505,
+      "step": 1720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.794229287365507e-06,
+      "loss": 0.0999,
+      "step": 1725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.792802306001883e-06,
+      "loss": 0.1156,
+      "step": 1730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.791375324638261e-06,
+      "loss": 0.0962,
+      "step": 1735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.789948343274638e-06,
+      "loss": 0.2936,
+      "step": 1740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.788521361911014e-06,
+      "loss": 0.3334,
+      "step": 1745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.78709438054739e-06,
+      "loss": 0.5398,
+      "step": 1750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.785667399183767e-06,
+      "loss": 0.2475,
+      "step": 1755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.784240417820145e-06,
+      "loss": 0.1535,
+      "step": 1760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.782813436456521e-06,
+      "loss": 0.1862,
+      "step": 1765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.781386455092897e-06,
+      "loss": 0.2029,
+      "step": 1770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.779959473729274e-06,
+      "loss": 0.1054,
+      "step": 1775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.778532492365652e-06,
+      "loss": 0.166,
+      "step": 1780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.777105511002028e-06,
+      "loss": 0.2208,
+      "step": 1785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.775678529638404e-06,
+      "loss": 0.1572,
+      "step": 1790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.77425154827478e-06,
+      "loss": 0.2488,
+      "step": 1795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.772824566911157e-06,
+      "loss": 0.2578,
+      "step": 1800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.771397585547533e-06,
+      "loss": 0.2531,
+      "step": 1805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.76997060418391e-06,
+      "loss": 0.2395,
+      "step": 1810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.768543622820286e-06,
+      "loss": 0.2433,
+      "step": 1815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.767116641456664e-06,
+      "loss": 0.2235,
+      "step": 1820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.76568966009304e-06,
+      "loss": 0.2172,
+      "step": 1825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.764262678729417e-06,
+      "loss": 0.1271,
+      "step": 1830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.762835697365793e-06,
+      "loss": 0.439,
+      "step": 1835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.76140871600217e-06,
+      "loss": 0.1604,
+      "step": 1840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.759981734638546e-06,
+      "loss": 0.3628,
+      "step": 1845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.758554753274922e-06,
+      "loss": 0.4847,
+      "step": 1850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.7571277719113e-06,
+      "loss": 0.1674,
+      "step": 1855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.755700790547676e-06,
+      "loss": 0.2104,
+      "step": 1860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.754273809184053e-06,
+      "loss": 0.2227,
+      "step": 1865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.752846827820429e-06,
+      "loss": 0.163,
+      "step": 1870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.751419846456807e-06,
+      "loss": 0.2878,
+      "step": 1875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.749992865093183e-06,
+      "loss": 0.1738,
+      "step": 1880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.74856588372956e-06,
+      "loss": 0.109,
+      "step": 1885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.747138902365936e-06,
+      "loss": 0.2777,
+      "step": 1890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.745711921002312e-06,
+      "loss": 0.08,
+      "step": 1895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.744284939638689e-06,
+      "loss": 0.537,
+      "step": 1900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.742857958275065e-06,
+      "loss": 0.2149,
+      "step": 1905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.741430976911443e-06,
+      "loss": 0.2135,
+      "step": 1910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.74000399554782e-06,
+      "loss": 0.2176,
+      "step": 1915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.738577014184196e-06,
+      "loss": 0.2804,
+      "step": 1920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.737150032820572e-06,
+      "loss": 0.1747,
+      "step": 1925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.735723051456949e-06,
+      "loss": 0.1037,
+      "step": 1930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.734296070093325e-06,
+      "loss": 0.1026,
+      "step": 1935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.732869088729701e-06,
+      "loss": 0.1829,
+      "step": 1940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.731442107366078e-06,
+      "loss": 0.4844,
+      "step": 1945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.730015126002456e-06,
+      "loss": 0.3903,
+      "step": 1950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.728588144638832e-06,
+      "loss": 0.5092,
+      "step": 1955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.727161163275208e-06,
+      "loss": 0.1894,
+      "step": 1960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.725734181911585e-06,
+      "loss": 0.2359,
+      "step": 1965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.724307200547963e-06,
+      "loss": 0.2379,
+      "step": 1970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.722880219184339e-06,
+      "loss": 0.2554,
+      "step": 1975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.721453237820715e-06,
+      "loss": 0.1861,
+      "step": 1980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.720026256457092e-06,
+      "loss": 0.192,
+      "step": 1985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.718599275093468e-06,
+      "loss": 0.3486,
+      "step": 1990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.717172293729844e-06,
+      "loss": 0.2169,
+      "step": 1995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.715745312366222e-06,
+      "loss": 0.4418,
+      "step": 2000
+    },
+    {
+      "epoch": 0.06,
+      "eval_loss": 0.08974172174930573,
+      "eval_runtime": 2327.969,
+      "eval_samples_per_second": 6.881,
+      "eval_steps_per_second": 1.72,
+      "eval_wer": 0.26016949152542374,
+      "step": 2000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.714318331002599e-06,
+      "loss": 0.202,
+      "step": 2005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.712891349638975e-06,
+      "loss": 0.4201,
+      "step": 2010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.711464368275351e-06,
+      "loss": 0.2236,
+      "step": 2015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.710037386911728e-06,
+      "loss": 0.1821,
+      "step": 2020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.708610405548104e-06,
+      "loss": 0.2104,
+      "step": 2025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.70718342418448e-06,
+      "loss": 0.102,
+      "step": 2030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.705756442820857e-06,
+      "loss": 0.1826,
+      "step": 2035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.704329461457233e-06,
+      "loss": 0.288,
+      "step": 2040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.702902480093611e-06,
+      "loss": 0.5527,
+      "step": 2045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.701475498729987e-06,
+      "loss": 0.4803,
+      "step": 2050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.700048517366364e-06,
+      "loss": 0.2576,
+      "step": 2055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.69862153600274e-06,
+      "loss": 0.2127,
+      "step": 2060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.697194554639118e-06,
+      "loss": 0.2714,
+      "step": 2065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.695767573275494e-06,
+      "loss": 0.1932,
+      "step": 2070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.69434059191187e-06,
+      "loss": 0.1743,
+      "step": 2075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.692913610548247e-06,
+      "loss": 0.2139,
+      "step": 2080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.691486629184624e-06,
+      "loss": 0.2656,
+      "step": 2085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.690059647821002e-06,
+      "loss": 0.4477,
+      "step": 2090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.688632666457378e-06,
+      "loss": 0.215,
+      "step": 2095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.687205685093754e-06,
+      "loss": 0.3432,
+      "step": 2100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.68577870373013e-06,
+      "loss": 0.229,
+      "step": 2105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.684351722366507e-06,
+      "loss": 0.2255,
+      "step": 2110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.682924741002883e-06,
+      "loss": 0.1953,
+      "step": 2115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.68149775963926e-06,
+      "loss": 0.1701,
+      "step": 2120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.680070778275636e-06,
+      "loss": 0.2611,
+      "step": 2125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.678643796912012e-06,
+      "loss": 0.1288,
+      "step": 2130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.677216815548389e-06,
+      "loss": 0.0965,
+      "step": 2135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.675789834184767e-06,
+      "loss": 0.2093,
+      "step": 2140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.674362852821143e-06,
+      "loss": 0.2346,
+      "step": 2145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.67293587145752e-06,
+      "loss": 0.6203,
+      "step": 2150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.671508890093896e-06,
+      "loss": 0.2262,
+      "step": 2155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.670081908730274e-06,
+      "loss": 0.4529,
+      "step": 2160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.66865492736665e-06,
+      "loss": 0.235,
+      "step": 2165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.667227946003026e-06,
+      "loss": 0.1584,
+      "step": 2170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.665800964639403e-06,
+      "loss": 0.2514,
+      "step": 2175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.664373983275779e-06,
+      "loss": 0.0816,
+      "step": 2180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.662947001912157e-06,
+      "loss": 0.0859,
+      "step": 2185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.661520020548533e-06,
+      "loss": 0.1528,
+      "step": 2190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.66009303918491e-06,
+      "loss": 0.3267,
+      "step": 2195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.658666057821286e-06,
+      "loss": 0.3738,
+      "step": 2200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.657239076457662e-06,
+      "loss": 0.214,
+      "step": 2205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.655812095094039e-06,
+      "loss": 0.2082,
+      "step": 2210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.654385113730415e-06,
+      "loss": 0.179,
+      "step": 2215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.652958132366791e-06,
+      "loss": 0.1861,
+      "step": 2220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.651531151003168e-06,
+      "loss": 0.1895,
+      "step": 2225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.650104169639544e-06,
+      "loss": 0.0996,
+      "step": 2230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.648677188275922e-06,
+      "loss": 0.0948,
+      "step": 2235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.647250206912298e-06,
+      "loss": 0.1832,
+      "step": 2240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.645823225548675e-06,
+      "loss": 0.2559,
+      "step": 2245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.644396244185051e-06,
+      "loss": 0.6217,
+      "step": 2250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.64296926282143e-06,
+      "loss": 0.2921,
+      "step": 2255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.641542281457806e-06,
+      "loss": 0.146,
+      "step": 2260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.640115300094182e-06,
+      "loss": 0.2137,
+      "step": 2265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.638688318730558e-06,
+      "loss": 0.41,
+      "step": 2270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.637261337366935e-06,
+      "loss": 0.2422,
+      "step": 2275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.635834356003313e-06,
+      "loss": 0.1457,
+      "step": 2280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.634407374639689e-06,
+      "loss": 0.1725,
+      "step": 2285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.632980393276065e-06,
+      "loss": 0.1379,
+      "step": 2290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.631553411912442e-06,
+      "loss": 0.2867,
+      "step": 2295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.630126430548818e-06,
+      "loss": 0.4942,
+      "step": 2300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.628699449185194e-06,
+      "loss": 0.2345,
+      "step": 2305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.62727246782157e-06,
+      "loss": 0.1619,
+      "step": 2310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.625845486457947e-06,
+      "loss": 0.1986,
+      "step": 2315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.624418505094323e-06,
+      "loss": 0.2719,
+      "step": 2320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.6229915237307e-06,
+      "loss": 0.2429,
+      "step": 2325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.621564542367078e-06,
+      "loss": 0.2239,
+      "step": 2330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.620137561003454e-06,
+      "loss": 0.1842,
+      "step": 2335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.61871057963983e-06,
+      "loss": 0.2596,
+      "step": 2340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.617283598276207e-06,
+      "loss": 0.3677,
+      "step": 2345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.615856616912585e-06,
+      "loss": 0.5166,
+      "step": 2350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.614429635548961e-06,
+      "loss": 0.1937,
+      "step": 2355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.613002654185337e-06,
+      "loss": 0.1555,
+      "step": 2360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.611575672821714e-06,
+      "loss": 0.1997,
+      "step": 2365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.61014869145809e-06,
+      "loss": 0.2785,
+      "step": 2370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.608721710094468e-06,
+      "loss": 0.1963,
+      "step": 2375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.607294728730844e-06,
+      "loss": 0.2269,
+      "step": 2380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.60586774736722e-06,
+      "loss": 0.1675,
+      "step": 2385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.604440766003597e-06,
+      "loss": 0.1251,
+      "step": 2390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.603013784639973e-06,
+      "loss": 0.4726,
+      "step": 2395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.60158680327635e-06,
+      "loss": 0.6281,
+      "step": 2400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.600159821912726e-06,
+      "loss": 0.2328,
+      "step": 2405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.598732840549102e-06,
+      "loss": 0.2014,
+      "step": 2410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.597305859185479e-06,
+      "loss": 0.2418,
+      "step": 2415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.595878877821855e-06,
+      "loss": 0.2288,
+      "step": 2420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.594451896458233e-06,
+      "loss": 0.2354,
+      "step": 2425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.59302491509461e-06,
+      "loss": 0.1047,
+      "step": 2430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.591597933730986e-06,
+      "loss": 0.1122,
+      "step": 2435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.590170952367362e-06,
+      "loss": 0.4132,
+      "step": 2440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.58874397100374e-06,
+      "loss": 0.2479,
+      "step": 2445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.587316989640117e-06,
+      "loss": 0.3581,
+      "step": 2450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.585890008276493e-06,
+      "loss": 0.214,
+      "step": 2455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.58446302691287e-06,
+      "loss": 0.2046,
+      "step": 2460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.583036045549246e-06,
+      "loss": 0.1714,
+      "step": 2465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.581609064185624e-06,
+      "loss": 0.2386,
+      "step": 2470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.580182082822e-06,
+      "loss": 0.0804,
+      "step": 2475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.578755101458376e-06,
+      "loss": 0.2579,
+      "step": 2480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.577328120094753e-06,
+      "loss": 0.2384,
+      "step": 2485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.575901138731129e-06,
+      "loss": 0.2612,
+      "step": 2490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.574474157367505e-06,
+      "loss": 0.245,
+      "step": 2495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.573047176003882e-06,
+      "loss": 0.5089,
+      "step": 2500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.571620194640258e-06,
+      "loss": 0.2194,
+      "step": 2505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.570193213276634e-06,
+      "loss": 0.1558,
+      "step": 2510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.56876623191301e-06,
+      "loss": 0.1905,
+      "step": 2515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.567339250549389e-06,
+      "loss": 0.2005,
+      "step": 2520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.565912269185765e-06,
+      "loss": 0.0812,
+      "step": 2525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.564485287822141e-06,
+      "loss": 0.1355,
+      "step": 2530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.563058306458518e-06,
+      "loss": 0.2427,
+      "step": 2535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.561631325094896e-06,
+      "loss": 0.2012,
+      "step": 2540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.560204343731272e-06,
+      "loss": 0.2796,
+      "step": 2545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.558777362367648e-06,
+      "loss": 0.5535,
+      "step": 2550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.557350381004025e-06,
+      "loss": 0.1669,
+      "step": 2555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.555923399640401e-06,
+      "loss": 0.2318,
+      "step": 2560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.554496418276779e-06,
+      "loss": 0.2368,
+      "step": 2565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.553069436913155e-06,
+      "loss": 0.2134,
+      "step": 2570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.551642455549532e-06,
+      "loss": 0.1166,
+      "step": 2575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.550215474185908e-06,
+      "loss": 0.0954,
+      "step": 2580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.548788492822284e-06,
+      "loss": 0.1802,
+      "step": 2585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.54736151145866e-06,
+      "loss": 0.2293,
+      "step": 2590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.545934530095037e-06,
+      "loss": 0.275,
+      "step": 2595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.544507548731413e-06,
+      "loss": 0.3557,
+      "step": 2600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.54308056736779e-06,
+      "loss": 0.1596,
+      "step": 2605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.541653586004168e-06,
+      "loss": 0.1963,
+      "step": 2610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.540226604640544e-06,
+      "loss": 0.2286,
+      "step": 2615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.53879962327692e-06,
+      "loss": 0.2489,
+      "step": 2620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.537372641913297e-06,
+      "loss": 0.2016,
+      "step": 2625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.535945660549673e-06,
+      "loss": 0.1457,
+      "step": 2630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.534518679186051e-06,
+      "loss": 0.1496,
+      "step": 2635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.533091697822428e-06,
+      "loss": 0.219,
+      "step": 2640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.531664716458804e-06,
+      "loss": 0.2844,
+      "step": 2645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.53023773509518e-06,
+      "loss": 0.5178,
+      "step": 2650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.528810753731557e-06,
+      "loss": 0.2695,
+      "step": 2655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.527383772367935e-06,
+      "loss": 0.1573,
+      "step": 2660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.525956791004311e-06,
+      "loss": 0.1983,
+      "step": 2665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.524529809640687e-06,
+      "loss": 0.2395,
+      "step": 2670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.523102828277064e-06,
+      "loss": 0.2201,
+      "step": 2675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.52167584691344e-06,
+      "loss": 0.1012,
+      "step": 2680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.520248865549816e-06,
+      "loss": 0.247,
+      "step": 2685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.518821884186193e-06,
+      "loss": 0.2196,
+      "step": 2690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.517394902822569e-06,
+      "loss": 0.2505,
+      "step": 2695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.515967921458947e-06,
+      "loss": 0.4629,
+      "step": 2700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.514540940095323e-06,
+      "loss": 0.2163,
+      "step": 2705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.5131139587317e-06,
+      "loss": 0.212,
+      "step": 2710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.511686977368076e-06,
+      "loss": 0.2412,
+      "step": 2715
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.510259996004452e-06,
+      "loss": 0.1869,
+      "step": 2720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.508833014640829e-06,
+      "loss": 0.2049,
+      "step": 2725
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.507406033277207e-06,
+      "loss": 0.1012,
+      "step": 2730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.505979051913583e-06,
+      "loss": 0.1866,
+      "step": 2735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.50455207054996e-06,
+      "loss": 0.1811,
+      "step": 2740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.503125089186336e-06,
+      "loss": 0.4811,
+      "step": 2745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.501698107822712e-06,
+      "loss": 0.6059,
+      "step": 2750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.50027112645909e-06,
+      "loss": 0.2447,
+      "step": 2755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.498844145095466e-06,
+      "loss": 0.1582,
+      "step": 2760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.497417163731843e-06,
+      "loss": 0.1435,
+      "step": 2765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.49599018236822e-06,
+      "loss": 0.258,
+      "step": 2770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.494563201004596e-06,
+      "loss": 0.1405,
+      "step": 2775
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.493136219640972e-06,
+      "loss": 0.122,
+      "step": 2780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.491709238277348e-06,
+      "loss": 0.1433,
+      "step": 2785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.490282256913726e-06,
+      "loss": 0.141,
+      "step": 2790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.488855275550103e-06,
+      "loss": 0.2074,
+      "step": 2795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.487428294186479e-06,
+      "loss": 0.3473,
+      "step": 2800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.486001312822855e-06,
+      "loss": 0.2921,
+      "step": 2805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.484574331459232e-06,
+      "loss": 0.2629,
+      "step": 2810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.483147350095608e-06,
+      "loss": 0.1963,
+      "step": 2815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.481720368731984e-06,
+      "loss": 0.184,
+      "step": 2820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.480293387368362e-06,
+      "loss": 0.1284,
+      "step": 2825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.478866406004739e-06,
+      "loss": 0.153,
+      "step": 2830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.477439424641115e-06,
+      "loss": 0.1785,
+      "step": 2835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.476012443277491e-06,
+      "loss": 0.1201,
+      "step": 2840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.474585461913868e-06,
+      "loss": 0.6185,
+      "step": 2845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.473158480550246e-06,
+      "loss": 0.3666,
+      "step": 2850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.471731499186622e-06,
+      "loss": 0.1744,
+      "step": 2855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.470304517822998e-06,
+      "loss": 0.205,
+      "step": 2860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.468877536459375e-06,
+      "loss": 0.1366,
+      "step": 2865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.467450555095751e-06,
+      "loss": 0.2189,
+      "step": 2870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.466023573732129e-06,
+      "loss": 0.1519,
+      "step": 2875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.464596592368505e-06,
+      "loss": 0.1423,
+      "step": 2880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.463169611004882e-06,
+      "loss": 0.2421,
+      "step": 2885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.461742629641258e-06,
+      "loss": 0.2549,
+      "step": 2890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.460315648277634e-06,
+      "loss": 0.1883,
+      "step": 2895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.45888866691401e-06,
+      "loss": 0.2442,
+      "step": 2900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.457461685550387e-06,
+      "loss": 0.261,
+      "step": 2905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.456034704186763e-06,
+      "loss": 0.201,
+      "step": 2910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.45460772282314e-06,
+      "loss": 0.2201,
+      "step": 2915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.453180741459518e-06,
+      "loss": 0.17,
+      "step": 2920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.451753760095894e-06,
+      "loss": 0.1684,
+      "step": 2925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.45032677873227e-06,
+      "loss": 0.1543,
+      "step": 2930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.448899797368647e-06,
+      "loss": 0.1616,
+      "step": 2935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.447472816005023e-06,
+      "loss": 0.2266,
+      "step": 2940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.446045834641401e-06,
+      "loss": 0.2079,
+      "step": 2945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.444618853277778e-06,
+      "loss": 0.5384,
+      "step": 2950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.443191871914154e-06,
+      "loss": 0.2111,
+      "step": 2955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.44176489055053e-06,
+      "loss": 0.1835,
+      "step": 2960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.440337909186907e-06,
+      "loss": 0.2175,
+      "step": 2965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.438910927823285e-06,
+      "loss": 0.2433,
+      "step": 2970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.437483946459661e-06,
+      "loss": 0.1842,
+      "step": 2975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.436056965096037e-06,
+      "loss": 0.1738,
+      "step": 2980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.434629983732414e-06,
+      "loss": 0.1602,
+      "step": 2985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.43320300236879e-06,
+      "loss": 0.3211,
+      "step": 2990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.431776021005166e-06,
+      "loss": 0.2437,
+      "step": 2995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.430349039641543e-06,
+      "loss": 0.5782,
+      "step": 3000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.428922058277919e-06,
+      "loss": 0.2063,
+      "step": 3005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.427495076914295e-06,
+      "loss": 0.2657,
+      "step": 3010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.426068095550673e-06,
+      "loss": 0.2458,
+      "step": 3015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.42464111418705e-06,
+      "loss": 0.1792,
+      "step": 3020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.423214132823426e-06,
+      "loss": 0.215,
+      "step": 3025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.421787151459802e-06,
+      "loss": 0.1876,
+      "step": 3030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.420360170096179e-06,
+      "loss": 0.2164,
+      "step": 3035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.418933188732557e-06,
+      "loss": 0.3034,
+      "step": 3040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.417506207368933e-06,
+      "loss": 0.3196,
+      "step": 3045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.41607922600531e-06,
+      "loss": 0.3141,
+      "step": 3050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.414652244641686e-06,
+      "loss": 0.1715,
+      "step": 3055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.413225263278062e-06,
+      "loss": 0.2603,
+      "step": 3060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.41179828191444e-06,
+      "loss": 0.207,
+      "step": 3065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.410371300550816e-06,
+      "loss": 0.2695,
+      "step": 3070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.408944319187193e-06,
+      "loss": 0.2754,
+      "step": 3075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.407517337823569e-06,
+      "loss": 0.1798,
+      "step": 3080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.406090356459945e-06,
+      "loss": 0.1376,
+      "step": 3085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.404663375096322e-06,
+      "loss": 0.2561,
+      "step": 3090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.403236393732698e-06,
+      "loss": 0.4096,
+      "step": 3095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.401809412369074e-06,
+      "loss": 0.6031,
+      "step": 3100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.40038243100545e-06,
+      "loss": 0.1809,
+      "step": 3105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.398955449641829e-06,
+      "loss": 0.1677,
+      "step": 3110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.397528468278205e-06,
+      "loss": 0.2089,
+      "step": 3115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.396101486914582e-06,
+      "loss": 0.1739,
+      "step": 3120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.394674505550958e-06,
+      "loss": 0.1027,
+      "step": 3125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.393247524187334e-06,
+      "loss": 0.277,
+      "step": 3130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.391820542823712e-06,
+      "loss": 0.2017,
+      "step": 3135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.390393561460089e-06,
+      "loss": 0.1102,
+      "step": 3140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.388966580096465e-06,
+      "loss": 0.2556,
+      "step": 3145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.387539598732841e-06,
+      "loss": 0.3088,
+      "step": 3150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.386112617369218e-06,
+      "loss": 0.2412,
+      "step": 3155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.384685636005596e-06,
+      "loss": 0.1554,
+      "step": 3160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.383258654641972e-06,
+      "loss": 0.2364,
+      "step": 3165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.381831673278348e-06,
+      "loss": 0.1928,
+      "step": 3170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.380404691914725e-06,
+      "loss": 0.2412,
+      "step": 3175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.378977710551101e-06,
+      "loss": 0.0616,
+      "step": 3180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.377550729187477e-06,
+      "loss": 0.1695,
+      "step": 3185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.376123747823854e-06,
+      "loss": 0.2261,
+      "step": 3190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.37469676646023e-06,
+      "loss": 0.3939,
+      "step": 3195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.373269785096606e-06,
+      "loss": 0.3573,
+      "step": 3200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.371842803732984e-06,
+      "loss": 0.1546,
+      "step": 3205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.37041582236936e-06,
+      "loss": 0.1995,
+      "step": 3210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.368988841005737e-06,
+      "loss": 0.1817,
+      "step": 3215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.367561859642113e-06,
+      "loss": 0.2167,
+      "step": 3220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.36613487827849e-06,
+      "loss": 0.0847,
+      "step": 3225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.364707896914868e-06,
+      "loss": 0.3077,
+      "step": 3230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.363280915551244e-06,
+      "loss": 0.1731,
+      "step": 3235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.36185393418762e-06,
+      "loss": 0.2065,
+      "step": 3240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.360426952823997e-06,
+      "loss": 0.2309,
+      "step": 3245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.358999971460373e-06,
+      "loss": 0.4587,
+      "step": 3250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.357572990096751e-06,
+      "loss": 0.1917,
+      "step": 3255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.356146008733127e-06,
+      "loss": 0.2364,
+      "step": 3260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.354719027369504e-06,
+      "loss": 0.2863,
+      "step": 3265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.35329204600588e-06,
+      "loss": 0.2016,
+      "step": 3270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.351865064642256e-06,
+      "loss": 0.0938,
+      "step": 3275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.350438083278633e-06,
+      "loss": 0.1143,
+      "step": 3280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.34901110191501e-06,
+      "loss": 0.1648,
+      "step": 3285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.347584120551385e-06,
+      "loss": 0.1564,
+      "step": 3290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.346157139187762e-06,
+      "loss": 0.2529,
+      "step": 3295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.34473015782414e-06,
+      "loss": 0.2956,
+      "step": 3300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.343303176460516e-06,
+      "loss": 0.2099,
+      "step": 3305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.341876195096893e-06,
+      "loss": 0.2067,
+      "step": 3310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.340449213733269e-06,
+      "loss": 0.1962,
+      "step": 3315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.339022232369645e-06,
+      "loss": 0.2011,
+      "step": 3320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.337595251006023e-06,
+      "loss": 0.2443,
+      "step": 3325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.3361682696424e-06,
+      "loss": 0.1325,
+      "step": 3330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.334741288278776e-06,
+      "loss": 0.0892,
+      "step": 3335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.333314306915152e-06,
+      "loss": 0.2842,
+      "step": 3340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.331887325551529e-06,
+      "loss": 0.3977,
+      "step": 3345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.330460344187907e-06,
+      "loss": 0.5102,
+      "step": 3350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.329033362824283e-06,
+      "loss": 0.2043,
+      "step": 3355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.32760638146066e-06,
+      "loss": 0.2347,
+      "step": 3360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.326179400097036e-06,
+      "loss": 0.1802,
+      "step": 3365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.324752418733412e-06,
+      "loss": 0.2627,
+      "step": 3370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.323325437369788e-06,
+      "loss": 0.2043,
+      "step": 3375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.321898456006165e-06,
+      "loss": 0.175,
+      "step": 3380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.320471474642541e-06,
+      "loss": 0.1667,
+      "step": 3385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.319044493278917e-06,
+      "loss": 0.1856,
+      "step": 3390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.317617511915295e-06,
+      "loss": 0.1524,
+      "step": 3395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.316190530551672e-06,
+      "loss": 0.5217,
+      "step": 3400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.314763549188048e-06,
+      "loss": 0.1621,
+      "step": 3405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.313336567824424e-06,
+      "loss": 0.2443,
+      "step": 3410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.3119095864608e-06,
+      "loss": 0.1285,
+      "step": 3415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.310482605097179e-06,
+      "loss": 0.2026,
+      "step": 3420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.309055623733555e-06,
+      "loss": 0.1533,
+      "step": 3425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.307628642369931e-06,
+      "loss": 0.1703,
+      "step": 3430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.306201661006308e-06,
+      "loss": 0.173,
+      "step": 3435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.304774679642684e-06,
+      "loss": 0.2034,
+      "step": 3440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.303347698279062e-06,
+      "loss": 0.3843,
+      "step": 3445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.301920716915438e-06,
+      "loss": 0.6026,
+      "step": 3450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.300493735551815e-06,
+      "loss": 0.2296,
+      "step": 3455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.299066754188191e-06,
+      "loss": 0.2542,
+      "step": 3460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.297639772824568e-06,
+      "loss": 0.2546,
+      "step": 3465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.296212791460944e-06,
+      "loss": 0.1721,
+      "step": 3470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.29478581009732e-06,
+      "loss": 0.1002,
+      "step": 3475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.293358828733697e-06,
+      "loss": 0.1049,
+      "step": 3480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.291931847370073e-06,
+      "loss": 0.1701,
+      "step": 3485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.290504866006451e-06,
+      "loss": 0.1776,
+      "step": 3490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.289077884642827e-06,
+      "loss": 0.3081,
+      "step": 3495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.287650903279204e-06,
+      "loss": 0.4832,
+      "step": 3500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.28622392191558e-06,
+      "loss": 0.1629,
+      "step": 3505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.284796940551956e-06,
+      "loss": 0.1529,
+      "step": 3510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.283369959188334e-06,
+      "loss": 0.2222,
+      "step": 3515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.28194297782471e-06,
+      "loss": 0.247,
+      "step": 3520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.280515996461087e-06,
+      "loss": 0.1803,
+      "step": 3525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.279089015097463e-06,
+      "loss": 0.1257,
+      "step": 3530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.27766203373384e-06,
+      "loss": 0.249,
+      "step": 3535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.276235052370218e-06,
+      "loss": 0.1671,
+      "step": 3540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.274808071006594e-06,
+      "loss": 0.3034,
+      "step": 3545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.27338108964297e-06,
+      "loss": 0.7076,
+      "step": 3550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.271954108279347e-06,
+      "loss": 0.198,
+      "step": 3555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.270527126915723e-06,
+      "loss": 0.2329,
+      "step": 3560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.2691001455521e-06,
+      "loss": 0.2115,
+      "step": 3565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.267673164188476e-06,
+      "loss": 0.241,
+      "step": 3570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.266246182824852e-06,
+      "loss": 0.1814,
+      "step": 3575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.26481920146123e-06,
+      "loss": 0.1618,
+      "step": 3580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.263392220097606e-06,
+      "loss": 0.1482,
+      "step": 3585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.261965238733983e-06,
+      "loss": 0.3956,
+      "step": 3590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.260538257370359e-06,
+      "loss": 0.3081,
+      "step": 3595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.259111276006735e-06,
+      "loss": 0.7055,
+      "step": 3600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.257684294643112e-06,
+      "loss": 0.1578,
+      "step": 3605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.25625731327949e-06,
+      "loss": 0.2141,
+      "step": 3610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.254830331915866e-06,
+      "loss": 0.2331,
+      "step": 3615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.253403350552242e-06,
+      "loss": 0.1813,
+      "step": 3620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.251976369188619e-06,
+      "loss": 0.1458,
+      "step": 3625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.250549387824995e-06,
+      "loss": 0.1256,
+      "step": 3630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.249122406461373e-06,
+      "loss": 0.1983,
+      "step": 3635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.24769542509775e-06,
+      "loss": 0.1667,
+      "step": 3640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.246268443734126e-06,
+      "loss": 0.4398,
+      "step": 3645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.244841462370502e-06,
+      "loss": 0.5002,
+      "step": 3650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.243414481006879e-06,
+      "loss": 0.192,
+      "step": 3655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.241987499643255e-06,
+      "loss": 0.2471,
+      "step": 3660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.240560518279633e-06,
+      "loss": 0.2014,
+      "step": 3665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.23913353691601e-06,
+      "loss": 0.2114,
+      "step": 3670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.237706555552386e-06,
+      "loss": 0.2025,
+      "step": 3675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.236279574188762e-06,
+      "loss": 0.0411,
+      "step": 3680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.234852592825138e-06,
+      "loss": 0.206,
+      "step": 3685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.233425611461515e-06,
+      "loss": 0.1223,
+      "step": 3690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.231998630097891e-06,
+      "loss": 0.2817,
+      "step": 3695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.230571648734267e-06,
+      "loss": 0.4316,
+      "step": 3700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.229144667370645e-06,
+      "loss": 0.2052,
+      "step": 3705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.227717686007022e-06,
+      "loss": 0.2416,
+      "step": 3710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.226290704643398e-06,
+      "loss": 0.1811,
+      "step": 3715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.224863723279774e-06,
+      "loss": 0.1848,
+      "step": 3720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.22343674191615e-06,
+      "loss": 0.0785,
+      "step": 3725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.222009760552529e-06,
+      "loss": 0.1411,
+      "step": 3730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.220582779188905e-06,
+      "loss": 0.0565,
+      "step": 3735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.219155797825281e-06,
+      "loss": 0.3228,
+      "step": 3740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.217728816461658e-06,
+      "loss": 0.3325,
+      "step": 3745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.216301835098034e-06,
+      "loss": 0.2787,
+      "step": 3750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.214874853734412e-06,
+      "loss": 0.1688,
+      "step": 3755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.213447872370788e-06,
+      "loss": 0.2905,
+      "step": 3760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.212020891007165e-06,
+      "loss": 0.2357,
+      "step": 3765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.210593909643541e-06,
+      "loss": 0.2019,
+      "step": 3770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.209166928279917e-06,
+      "loss": 0.2344,
+      "step": 3775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.207739946916294e-06,
+      "loss": 0.0983,
+      "step": 3780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.20631296555267e-06,
+      "loss": 0.0607,
+      "step": 3785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.204885984189046e-06,
+      "loss": 0.1757,
+      "step": 3790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.203459002825423e-06,
+      "loss": 0.2125,
+      "step": 3795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.2020320214618e-06,
+      "loss": 0.6506,
+      "step": 3800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.200605040098177e-06,
+      "loss": 0.208,
+      "step": 3805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.199178058734554e-06,
+      "loss": 0.1704,
+      "step": 3810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.19775107737093e-06,
+      "loss": 0.1661,
+      "step": 3815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.196324096007306e-06,
+      "loss": 0.3135,
+      "step": 3820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.194897114643684e-06,
+      "loss": 0.2682,
+      "step": 3825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.19347013328006e-06,
+      "loss": 0.0732,
+      "step": 3830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.192043151916437e-06,
+      "loss": 0.1708,
+      "step": 3835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.190616170552813e-06,
+      "loss": 0.1189,
+      "step": 3840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.189189189189191e-06,
+      "loss": 0.2076,
+      "step": 3845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.187762207825568e-06,
+      "loss": 0.5824,
+      "step": 3850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.186335226461944e-06,
+      "loss": 0.2316,
+      "step": 3855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.18490824509832e-06,
+      "loss": 0.239,
+      "step": 3860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.183481263734697e-06,
+      "loss": 0.2332,
+      "step": 3865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.182054282371073e-06,
+      "loss": 0.2173,
+      "step": 3870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.18062730100745e-06,
+      "loss": 0.0937,
+      "step": 3875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.179200319643826e-06,
+      "loss": 0.2231,
+      "step": 3880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.177773338280202e-06,
+      "loss": 0.3681,
+      "step": 3885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.176346356916578e-06,
+      "loss": 0.3334,
+      "step": 3890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.174919375552956e-06,
+      "loss": 0.3968,
+      "step": 3895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.173492394189333e-06,
+      "loss": 0.5548,
+      "step": 3900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.172065412825709e-06,
+      "loss": 0.146,
+      "step": 3905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.170638431462085e-06,
+      "loss": 0.2168,
+      "step": 3910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.169211450098462e-06,
+      "loss": 0.2161,
+      "step": 3915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.16778446873484e-06,
+      "loss": 0.1638,
+      "step": 3920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.166357487371216e-06,
+      "loss": 0.168,
+      "step": 3925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.164930506007592e-06,
+      "loss": 0.0929,
+      "step": 3930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.163503524643969e-06,
+      "loss": 0.2191,
+      "step": 3935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.162076543280347e-06,
+      "loss": 0.1826,
+      "step": 3940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.160649561916723e-06,
+      "loss": 0.4173,
+      "step": 3945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.1592225805531e-06,
+      "loss": 0.5063,
+      "step": 3950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.157795599189476e-06,
+      "loss": 0.2068,
+      "step": 3955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.156368617825852e-06,
+      "loss": 0.1955,
+      "step": 3960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.154941636462228e-06,
+      "loss": 0.2014,
+      "step": 3965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.153514655098605e-06,
+      "loss": 0.2034,
+      "step": 3970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.152087673734981e-06,
+      "loss": 0.2596,
+      "step": 3975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.150660692371357e-06,
+      "loss": 0.0777,
+      "step": 3980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.149233711007734e-06,
+      "loss": 0.1958,
+      "step": 3985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.147806729644112e-06,
+      "loss": 0.1974,
+      "step": 3990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.146379748280488e-06,
+      "loss": 0.2286,
+      "step": 3995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.144952766916865e-06,
+      "loss": 0.4232,
+      "step": 4000
+    },
+    {
+      "epoch": 0.11,
+      "eval_loss": 0.08065038919448853,
+      "eval_runtime": 2611.7113,
+      "eval_samples_per_second": 6.133,
+      "eval_steps_per_second": 1.533,
+      "eval_wer": 0.2415795167688424,
+      "step": 4000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.143525785553241e-06,
+      "loss": 0.2607,
+      "step": 4005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.142098804189617e-06,
+      "loss": 0.2374,
+      "step": 4010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.140671822825995e-06,
+      "loss": 0.2258,
+      "step": 4015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.139244841462372e-06,
+      "loss": 0.2825,
+      "step": 4020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.137817860098748e-06,
+      "loss": 0.2166,
+      "step": 4025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.136390878735124e-06,
+      "loss": 0.1085,
+      "step": 4030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.134963897371502e-06,
+      "loss": 0.244,
+      "step": 4035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.133536916007879e-06,
+      "loss": 0.2392,
+      "step": 4040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.132109934644255e-06,
+      "loss": 0.3185,
+      "step": 4045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.130682953280631e-06,
+      "loss": 0.4469,
+      "step": 4050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.129255971917008e-06,
+      "loss": 0.2549,
+      "step": 4055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.127828990553384e-06,
+      "loss": 0.1936,
+      "step": 4060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.12640200918976e-06,
+      "loss": 0.2288,
+      "step": 4065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.124975027826137e-06,
+      "loss": 0.2042,
+      "step": 4070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.123548046462513e-06,
+      "loss": 0.207,
+      "step": 4075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.12212106509889e-06,
+      "loss": 0.0934,
+      "step": 4080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.120694083735267e-06,
+      "loss": 0.2389,
+      "step": 4085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.119267102371644e-06,
+      "loss": 0.1741,
+      "step": 4090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.11784012100802e-06,
+      "loss": 0.2289,
+      "step": 4095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.116413139644396e-06,
+      "loss": 0.7276,
+      "step": 4100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.114986158280773e-06,
+      "loss": 0.2186,
+      "step": 4105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.11355917691715e-06,
+      "loss": 0.1656,
+      "step": 4110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.112132195553527e-06,
+      "loss": 0.1967,
+      "step": 4115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.110705214189903e-06,
+      "loss": 0.3377,
+      "step": 4120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.10927823282628e-06,
+      "loss": 0.1491,
+      "step": 4125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.107851251462658e-06,
+      "loss": 0.0329,
+      "step": 4130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.106424270099034e-06,
+      "loss": 0.0635,
+      "step": 4135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.10499728873541e-06,
+      "loss": 0.2521,
+      "step": 4140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.103570307371787e-06,
+      "loss": 0.4768,
+      "step": 4145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.102143326008163e-06,
+      "loss": 0.4017,
+      "step": 4150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.10071634464454e-06,
+      "loss": 0.1938,
+      "step": 4155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.099289363280916e-06,
+      "loss": 0.1638,
+      "step": 4160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.097862381917292e-06,
+      "loss": 0.2338,
+      "step": 4165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.096435400553669e-06,
+      "loss": 0.199,
+      "step": 4170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.095008419190045e-06,
+      "loss": 0.0978,
+      "step": 4175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.093581437826423e-06,
+      "loss": 0.1591,
+      "step": 4180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.0921544564628e-06,
+      "loss": 0.1114,
+      "step": 4185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.090727475099176e-06,
+      "loss": 0.2239,
+      "step": 4190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.089300493735552e-06,
+      "loss": 0.285,
+      "step": 4195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.08787351237193e-06,
+      "loss": 0.5391,
+      "step": 4200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.086446531008306e-06,
+      "loss": 0.2093,
+      "step": 4205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.085019549644683e-06,
+      "loss": 0.1986,
+      "step": 4210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.083592568281059e-06,
+      "loss": 0.2044,
+      "step": 4215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.082165586917435e-06,
+      "loss": 0.2003,
+      "step": 4220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.080738605553813e-06,
+      "loss": 0.2496,
+      "step": 4225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.07931162419019e-06,
+      "loss": 0.1269,
+      "step": 4230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.077884642826566e-06,
+      "loss": 0.0929,
+      "step": 4235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.076457661462942e-06,
+      "loss": 0.124,
+      "step": 4240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.075030680099319e-06,
+      "loss": 0.3424,
+      "step": 4245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.073603698735695e-06,
+      "loss": 0.4852,
+      "step": 4250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.072176717372071e-06,
+      "loss": 0.1921,
+      "step": 4255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.070749736008448e-06,
+      "loss": 0.2314,
+      "step": 4260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.069322754644824e-06,
+      "loss": 0.1731,
+      "step": 4265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.0678957732812e-06,
+      "loss": 0.2128,
+      "step": 4270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.066468791917578e-06,
+      "loss": 0.2346,
+      "step": 4275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.065041810553955e-06,
+      "loss": 0.0798,
+      "step": 4280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.063614829190331e-06,
+      "loss": 0.034,
+      "step": 4285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.062187847826707e-06,
+      "loss": 0.341,
+      "step": 4290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.060760866463085e-06,
+      "loss": 0.3448,
+      "step": 4295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.059333885099462e-06,
+      "loss": 0.5544,
+      "step": 4300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.057906903735838e-06,
+      "loss": 0.2297,
+      "step": 4305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.056479922372214e-06,
+      "loss": 0.1696,
+      "step": 4310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.05505294100859e-06,
+      "loss": 0.2084,
+      "step": 4315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.053625959644969e-06,
+      "loss": 0.221,
+      "step": 4320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.052198978281345e-06,
+      "loss": 0.1946,
+      "step": 4325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.050771996917722e-06,
+      "loss": 0.1071,
+      "step": 4330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.049345015554098e-06,
+      "loss": 0.1961,
+      "step": 4335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.047918034190474e-06,
+      "loss": 0.3092,
+      "step": 4340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.04649105282685e-06,
+      "loss": 0.4163,
+      "step": 4345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.045064071463227e-06,
+      "loss": 0.6229,
+      "step": 4350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.043637090099603e-06,
+      "loss": 0.2903,
+      "step": 4355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.04221010873598e-06,
+      "loss": 0.1842,
+      "step": 4360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.040783127372358e-06,
+      "loss": 0.1753,
+      "step": 4365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.039356146008734e-06,
+      "loss": 0.1974,
+      "step": 4370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.03792916464511e-06,
+      "loss": 0.1202,
+      "step": 4375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.036502183281487e-06,
+      "loss": 0.0612,
+      "step": 4380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.035075201917863e-06,
+      "loss": 0.1177,
+      "step": 4385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.033648220554241e-06,
+      "loss": 0.2192,
+      "step": 4390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.032221239190617e-06,
+      "loss": 0.4009,
+      "step": 4395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.030794257826994e-06,
+      "loss": 0.7506,
+      "step": 4400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.02936727646337e-06,
+      "loss": 0.1961,
+      "step": 4405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.027940295099746e-06,
+      "loss": 0.1984,
+      "step": 4410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.026513313736124e-06,
+      "loss": 0.1924,
+      "step": 4415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.0250863323725e-06,
+      "loss": 0.1851,
+      "step": 4420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.023659351008877e-06,
+      "loss": 0.1969,
+      "step": 4425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.022232369645253e-06,
+      "loss": 0.1534,
+      "step": 4430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.02080538828163e-06,
+      "loss": 0.2427,
+      "step": 4435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.019378406918006e-06,
+      "loss": 0.2435,
+      "step": 4440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.017951425554382e-06,
+      "loss": 0.2561,
+      "step": 4445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.016524444190759e-06,
+      "loss": 0.284,
+      "step": 4450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.015097462827137e-06,
+      "loss": 0.1331,
+      "step": 4455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.013670481463513e-06,
+      "loss": 0.2267,
+      "step": 4460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.01224350009989e-06,
+      "loss": 0.1998,
+      "step": 4465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.010816518736266e-06,
+      "loss": 0.2609,
+      "step": 4470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.009389537372642e-06,
+      "loss": 0.1334,
+      "step": 4475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.007962556009018e-06,
+      "loss": 0.0818,
+      "step": 4480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.006535574645396e-06,
+      "loss": 0.1239,
+      "step": 4485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.005108593281773e-06,
+      "loss": 0.3317,
+      "step": 4490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.00368161191815e-06,
+      "loss": 0.216,
+      "step": 4495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.002254630554526e-06,
+      "loss": 0.5137,
+      "step": 4500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.000827649190902e-06,
+      "loss": 0.2356,
+      "step": 4505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.99940066782728e-06,
+      "loss": 0.1811,
+      "step": 4510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.997973686463656e-06,
+      "loss": 0.2463,
+      "step": 4515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.996546705100033e-06,
+      "loss": 0.2877,
+      "step": 4520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.995119723736409e-06,
+      "loss": 0.2499,
+      "step": 4525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.993692742372785e-06,
+      "loss": 0.2056,
+      "step": 4530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.992265761009162e-06,
+      "loss": 0.0899,
+      "step": 4535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.990838779645538e-06,
+      "loss": 0.2066,
+      "step": 4540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.989411798281916e-06,
+      "loss": 0.1908,
+      "step": 4545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.987984816918292e-06,
+      "loss": 0.324,
+      "step": 4550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.986557835554669e-06,
+      "loss": 0.1409,
+      "step": 4555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.985130854191045e-06,
+      "loss": 0.2014,
+      "step": 4560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.983703872827421e-06,
+      "loss": 0.1857,
+      "step": 4565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.982276891463798e-06,
+      "loss": 0.2537,
+      "step": 4570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.980849910100174e-06,
+      "loss": 0.2109,
+      "step": 4575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.979422928736552e-06,
+      "loss": 0.1228,
+      "step": 4580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.977995947372928e-06,
+      "loss": 0.1489,
+      "step": 4585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.976568966009305e-06,
+      "loss": 0.1028,
+      "step": 4590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.975141984645681e-06,
+      "loss": 0.3634,
+      "step": 4595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.973715003282057e-06,
+      "loss": 0.5033,
+      "step": 4600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.972288021918435e-06,
+      "loss": 0.1769,
+      "step": 4605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.970861040554812e-06,
+      "loss": 0.2308,
+      "step": 4610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.969434059191188e-06,
+      "loss": 0.1799,
+      "step": 4615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.968007077827564e-06,
+      "loss": 0.2012,
+      "step": 4620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.96658009646394e-06,
+      "loss": 0.2702,
+      "step": 4625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.965153115100317e-06,
+      "loss": 0.1508,
+      "step": 4630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.963726133736695e-06,
+      "loss": 0.1771,
+      "step": 4635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.962299152373071e-06,
+      "loss": 0.0945,
+      "step": 4640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.960872171009448e-06,
+      "loss": 0.2389,
+      "step": 4645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.959445189645824e-06,
+      "loss": 0.5115,
+      "step": 4650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.9580182082822e-06,
+      "loss": 0.2057,
+      "step": 4655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.956591226918577e-06,
+      "loss": 0.1715,
+      "step": 4660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.955164245554953e-06,
+      "loss": 0.1947,
+      "step": 4665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.95373726419133e-06,
+      "loss": 0.19,
+      "step": 4670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.952310282827708e-06,
+      "loss": 0.257,
+      "step": 4675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.950883301464084e-06,
+      "loss": 0.15,
+      "step": 4680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.94945632010046e-06,
+      "loss": 0.1584,
+      "step": 4685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.948029338736837e-06,
+      "loss": 0.1429,
+      "step": 4690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.946602357373213e-06,
+      "loss": 0.2701,
+      "step": 4695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.945175376009591e-06,
+      "loss": 0.3097,
+      "step": 4700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.943748394645967e-06,
+      "loss": 0.2693,
+      "step": 4705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.942321413282344e-06,
+      "loss": 0.1758,
+      "step": 4710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.94089443191872e-06,
+      "loss": 0.2927,
+      "step": 4715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.939467450555096e-06,
+      "loss": 0.2062,
+      "step": 4720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.938040469191474e-06,
+      "loss": 0.1183,
+      "step": 4725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.93661348782785e-06,
+      "loss": 0.1886,
+      "step": 4730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.935186506464227e-06,
+      "loss": 0.0822,
+      "step": 4735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.933759525100603e-06,
+      "loss": 0.2749,
+      "step": 4740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.93233254373698e-06,
+      "loss": 0.2977,
+      "step": 4745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.930905562373356e-06,
+      "loss": 0.4549,
+      "step": 4750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.929478581009732e-06,
+      "loss": 0.1402,
+      "step": 4755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.928051599646109e-06,
+      "loss": 0.1788,
+      "step": 4760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.926624618282485e-06,
+      "loss": 0.1905,
+      "step": 4765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.925197636918863e-06,
+      "loss": 0.2435,
+      "step": 4770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.92377065555524e-06,
+      "loss": 0.0795,
+      "step": 4775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.922343674191616e-06,
+      "loss": 0.1603,
+      "step": 4780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.920916692827992e-06,
+      "loss": 0.1377,
+      "step": 4785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.919489711464368e-06,
+      "loss": 0.2565,
+      "step": 4790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.918062730100746e-06,
+      "loss": 0.3557,
+      "step": 4795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.916635748737123e-06,
+      "loss": 0.4561,
+      "step": 4800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.915208767373499e-06,
+      "loss": 0.1781,
+      "step": 4805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.913781786009875e-06,
+      "loss": 0.1467,
+      "step": 4810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.912354804646252e-06,
+      "loss": 0.1543,
+      "step": 4815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.91092782328263e-06,
+      "loss": 0.2281,
+      "step": 4820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.909500841919006e-06,
+      "loss": 0.1932,
+      "step": 4825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.908073860555382e-06,
+      "loss": 0.1857,
+      "step": 4830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.906646879191759e-06,
+      "loss": 0.1831,
+      "step": 4835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.905219897828135e-06,
+      "loss": 0.207,
+      "step": 4840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.903792916464512e-06,
+      "loss": 0.2404,
+      "step": 4845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.902365935100888e-06,
+      "loss": 0.4327,
+      "step": 4850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.900938953737264e-06,
+      "loss": 0.16,
+      "step": 4855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.89951197237364e-06,
+      "loss": 0.1548,
+      "step": 4860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.898084991010019e-06,
+      "loss": 0.2546,
+      "step": 4865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.896658009646395e-06,
+      "loss": 0.2301,
+      "step": 4870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.895231028282771e-06,
+      "loss": 0.1885,
+      "step": 4875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.893804046919148e-06,
+      "loss": 0.1039,
+      "step": 4880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.892377065555524e-06,
+      "loss": 0.1215,
+      "step": 4885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.890950084191902e-06,
+      "loss": 0.2017,
+      "step": 4890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.889523102828278e-06,
+      "loss": 0.2429,
+      "step": 4895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.888096121464655e-06,
+      "loss": 0.4522,
+      "step": 4900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.886669140101031e-06,
+      "loss": 0.2326,
+      "step": 4905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.885242158737407e-06,
+      "loss": 0.2617,
+      "step": 4910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.883815177373785e-06,
+      "loss": 0.1861,
+      "step": 4915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.882388196010162e-06,
+      "loss": 0.2249,
+      "step": 4920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.880961214646538e-06,
+      "loss": 0.1508,
+      "step": 4925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.879534233282914e-06,
+      "loss": 0.1548,
+      "step": 4930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.87810725191929e-06,
+      "loss": 0.1668,
+      "step": 4935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.876680270555667e-06,
+      "loss": 0.1921,
+      "step": 4940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.875253289192043e-06,
+      "loss": 0.1872,
+      "step": 4945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.87382630782842e-06,
+      "loss": 0.5628,
+      "step": 4950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.872399326464796e-06,
+      "loss": 0.2048,
+      "step": 4955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.870972345101174e-06,
+      "loss": 0.1782,
+      "step": 4960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.86954536373755e-06,
+      "loss": 0.1839,
+      "step": 4965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.868118382373927e-06,
+      "loss": 0.2331,
+      "step": 4970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.866691401010303e-06,
+      "loss": 0.0948,
+      "step": 4975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.86526441964668e-06,
+      "loss": 0.071,
+      "step": 4980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.863837438283057e-06,
+      "loss": 0.2554,
+      "step": 4985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.862410456919434e-06,
+      "loss": 0.3874,
+      "step": 4990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.86098347555581e-06,
+      "loss": 0.3417,
+      "step": 4995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.859556494192186e-06,
+      "loss": 0.4399,
+      "step": 5000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.858129512828563e-06,
+      "loss": 0.1944,
+      "step": 5005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.856702531464941e-06,
+      "loss": 0.2096,
+      "step": 5010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.855275550101317e-06,
+      "loss": 0.2665,
+      "step": 5015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.853848568737694e-06,
+      "loss": 0.2119,
+      "step": 5020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.85242158737407e-06,
+      "loss": 0.1429,
+      "step": 5025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.850994606010446e-06,
+      "loss": 0.1091,
+      "step": 5030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.849567624646823e-06,
+      "loss": 0.1028,
+      "step": 5035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.848140643283199e-06,
+      "loss": 0.23,
+      "step": 5040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.846713661919575e-06,
+      "loss": 0.324,
+      "step": 5045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.845286680555952e-06,
+      "loss": 0.3115,
+      "step": 5050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.84385969919233e-06,
+      "loss": 0.1982,
+      "step": 5055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.842432717828706e-06,
+      "loss": 0.1659,
+      "step": 5060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.841005736465082e-06,
+      "loss": 0.1552,
+      "step": 5065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.839578755101459e-06,
+      "loss": 0.2341,
+      "step": 5070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.838151773737835e-06,
+      "loss": 0.2016,
+      "step": 5075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.836724792374213e-06,
+      "loss": 0.1199,
+      "step": 5080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.83529781101059e-06,
+      "loss": 0.0833,
+      "step": 5085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.833870829646966e-06,
+      "loss": 0.1925,
+      "step": 5090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.832443848283342e-06,
+      "loss": 0.2766,
+      "step": 5095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.831016866919718e-06,
+      "loss": 0.5481,
+      "step": 5100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.829589885556096e-06,
+      "loss": 0.192,
+      "step": 5105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.828162904192473e-06,
+      "loss": 0.1816,
+      "step": 5110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.826735922828849e-06,
+      "loss": 0.1952,
+      "step": 5115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.825308941465225e-06,
+      "loss": 0.1743,
+      "step": 5120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.823881960101602e-06,
+      "loss": 0.1717,
+      "step": 5125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.822454978737978e-06,
+      "loss": 0.0976,
+      "step": 5130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.821027997374354e-06,
+      "loss": 0.2031,
+      "step": 5135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.81960101601073e-06,
+      "loss": 0.1639,
+      "step": 5140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.818174034647107e-06,
+      "loss": 0.3788,
+      "step": 5145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.816747053283485e-06,
+      "loss": 0.6995,
+      "step": 5150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.815320071919861e-06,
+      "loss": 0.3716,
+      "step": 5155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.813893090556238e-06,
+      "loss": 0.1552,
+      "step": 5160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.812466109192614e-06,
+      "loss": 0.1855,
+      "step": 5165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.81103912782899e-06,
+      "loss": 0.1795,
+      "step": 5170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.809612146465368e-06,
+      "loss": 0.0744,
+      "step": 5175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.808185165101745e-06,
+      "loss": 0.1431,
+      "step": 5180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.806758183738121e-06,
+      "loss": 0.2072,
+      "step": 5185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.805331202374498e-06,
+      "loss": 0.1924,
+      "step": 5190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.803904221010874e-06,
+      "loss": 0.223,
+      "step": 5195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.802477239647252e-06,
+      "loss": 0.5766,
+      "step": 5200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.801050258283628e-06,
+      "loss": 0.1885,
+      "step": 5205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.799623276920005e-06,
+      "loss": 0.1874,
+      "step": 5210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.798196295556381e-06,
+      "loss": 0.1505,
+      "step": 5215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.796769314192757e-06,
+      "loss": 0.1784,
+      "step": 5220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.795342332829134e-06,
+      "loss": 0.18,
+      "step": 5225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.79391535146551e-06,
+      "loss": 0.1108,
+      "step": 5230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.792488370101886e-06,
+      "loss": 0.1662,
+      "step": 5235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.791061388738263e-06,
+      "loss": 0.1483,
+      "step": 5240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.78963440737464e-06,
+      "loss": 0.3551,
+      "step": 5245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.788207426011017e-06,
+      "loss": 0.262,
+      "step": 5250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.786780444647393e-06,
+      "loss": 0.1899,
+      "step": 5255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.78535346328377e-06,
+      "loss": 0.2994,
+      "step": 5260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.783926481920146e-06,
+      "loss": 0.1549,
+      "step": 5265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.782499500556524e-06,
+      "loss": 0.2618,
+      "step": 5270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.7810725191929e-06,
+      "loss": 0.2002,
+      "step": 5275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.779645537829277e-06,
+      "loss": 0.0888,
+      "step": 5280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.778218556465653e-06,
+      "loss": 0.2511,
+      "step": 5285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.77679157510203e-06,
+      "loss": 0.3803,
+      "step": 5290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.775364593738407e-06,
+      "loss": 0.5372,
+      "step": 5295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.773937612374784e-06,
+      "loss": 0.5779,
+      "step": 5300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.77251063101116e-06,
+      "loss": 0.215,
+      "step": 5305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.771083649647536e-06,
+      "loss": 0.259,
+      "step": 5310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.769656668283913e-06,
+      "loss": 0.1478,
+      "step": 5315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.768229686920289e-06,
+      "loss": 0.1971,
+      "step": 5320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.766802705556665e-06,
+      "loss": 0.2074,
+      "step": 5325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.765375724193042e-06,
+      "loss": 0.1503,
+      "step": 5330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.76394874282942e-06,
+      "loss": 0.1515,
+      "step": 5335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.762521761465796e-06,
+      "loss": 0.3774,
+      "step": 5340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.761094780102172e-06,
+      "loss": 0.2298,
+      "step": 5345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.759667798738549e-06,
+      "loss": 0.556,
+      "step": 5350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.758240817374925e-06,
+      "loss": 0.2015,
+      "step": 5355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.756813836011301e-06,
+      "loss": 0.148,
+      "step": 5360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.75538685464768e-06,
+      "loss": 0.2125,
+      "step": 5365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.753959873284056e-06,
+      "loss": 0.2282,
+      "step": 5370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.752532891920432e-06,
+      "loss": 0.2395,
+      "step": 5375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.751105910556809e-06,
+      "loss": 0.0962,
+      "step": 5380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.749678929193185e-06,
+      "loss": 0.1118,
+      "step": 5385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.748251947829563e-06,
+      "loss": 0.2768,
+      "step": 5390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.74682496646594e-06,
+      "loss": 0.3651,
+      "step": 5395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.745397985102316e-06,
+      "loss": 0.5803,
+      "step": 5400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.743971003738692e-06,
+      "loss": 0.1585,
+      "step": 5405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.742544022375068e-06,
+      "loss": 0.2192,
+      "step": 5410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.741117041011445e-06,
+      "loss": 0.2076,
+      "step": 5415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.739690059647823e-06,
+      "loss": 0.2241,
+      "step": 5420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.738263078284199e-06,
+      "loss": 0.1996,
+      "step": 5425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.736836096920575e-06,
+      "loss": 0.1283,
+      "step": 5430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.735409115556952e-06,
+      "loss": 0.2356,
+      "step": 5435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.733982134193328e-06,
+      "loss": 0.1745,
+      "step": 5440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.732555152829704e-06,
+      "loss": 0.2165,
+      "step": 5445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.73112817146608e-06,
+      "loss": 0.6278,
+      "step": 5450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.729701190102457e-06,
+      "loss": 0.1927,
+      "step": 5455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.728274208738835e-06,
+      "loss": 0.2593,
+      "step": 5460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.726847227375211e-06,
+      "loss": 0.1807,
+      "step": 5465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.725420246011588e-06,
+      "loss": 0.1971,
+      "step": 5470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.723993264647964e-06,
+      "loss": 0.1768,
+      "step": 5475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.72256628328434e-06,
+      "loss": 0.1762,
+      "step": 5480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.721139301920718e-06,
+      "loss": 0.1797,
+      "step": 5485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.719712320557095e-06,
+      "loss": 0.1995,
+      "step": 5490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.718285339193471e-06,
+      "loss": 0.2537,
+      "step": 5495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.716858357829847e-06,
+      "loss": 0.4993,
+      "step": 5500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.715431376466224e-06,
+      "loss": 0.1784,
+      "step": 5505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.714004395102602e-06,
+      "loss": 0.2367,
+      "step": 5510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.712577413738978e-06,
+      "loss": 0.1833,
+      "step": 5515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.711150432375354e-06,
+      "loss": 0.2138,
+      "step": 5520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.70972345101173e-06,
+      "loss": 0.1933,
+      "step": 5525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.708296469648107e-06,
+      "loss": 0.1948,
+      "step": 5530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.706869488284484e-06,
+      "loss": 0.0868,
+      "step": 5535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.70544250692086e-06,
+      "loss": 0.2228,
+      "step": 5540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.704015525557236e-06,
+      "loss": 0.263,
+      "step": 5545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.702588544193613e-06,
+      "loss": 0.4862,
+      "step": 5550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.70116156282999e-06,
+      "loss": 0.2016,
+      "step": 5555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.699734581466367e-06,
+      "loss": 0.1936,
+      "step": 5560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.698307600102743e-06,
+      "loss": 0.2058,
+      "step": 5565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.69688061873912e-06,
+      "loss": 0.4567,
+      "step": 5570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.695453637375496e-06,
+      "loss": 0.1977,
+      "step": 5575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.694026656011874e-06,
+      "loss": 0.0746,
+      "step": 5580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.69259967464825e-06,
+      "loss": 0.2863,
+      "step": 5585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.691172693284627e-06,
+      "loss": 0.1975,
+      "step": 5590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.689745711921003e-06,
+      "loss": 0.3318,
+      "step": 5595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.688318730557381e-06,
+      "loss": 0.283,
+      "step": 5600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.686891749193757e-06,
+      "loss": 0.173,
+      "step": 5605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.685464767830134e-06,
+      "loss": 0.1533,
+      "step": 5610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.68403778646651e-06,
+      "loss": 0.1697,
+      "step": 5615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.682610805102886e-06,
+      "loss": 0.1507,
+      "step": 5620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.681183823739263e-06,
+      "loss": 0.1705,
+      "step": 5625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.679756842375639e-06,
+      "loss": 0.0738,
+      "step": 5630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.678329861012015e-06,
+      "loss": 0.1915,
+      "step": 5635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.676902879648392e-06,
+      "loss": 0.2064,
+      "step": 5640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.675475898284768e-06,
+      "loss": 0.1501,
+      "step": 5645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.674048916921146e-06,
+      "loss": 0.4604,
+      "step": 5650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.672621935557522e-06,
+      "loss": 0.1099,
+      "step": 5655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.671194954193899e-06,
+      "loss": 0.1849,
+      "step": 5660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.669767972830275e-06,
+      "loss": 0.2311,
+      "step": 5665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.668340991466651e-06,
+      "loss": 0.1917,
+      "step": 5670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.66691401010303e-06,
+      "loss": 0.2114,
+      "step": 5675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.665487028739406e-06,
+      "loss": 0.0909,
+      "step": 5680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.664060047375782e-06,
+      "loss": 0.2297,
+      "step": 5685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.662633066012158e-06,
+      "loss": 0.1691,
+      "step": 5690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.661206084648537e-06,
+      "loss": 0.3327,
+      "step": 5695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.659779103284913e-06,
+      "loss": 0.772,
+      "step": 5700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.65835212192129e-06,
+      "loss": 0.1924,
+      "step": 5705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.656925140557666e-06,
+      "loss": 0.2268,
+      "step": 5710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.655498159194042e-06,
+      "loss": 0.2013,
+      "step": 5715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.654071177830418e-06,
+      "loss": 0.2103,
+      "step": 5720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.652644196466795e-06,
+      "loss": 0.2448,
+      "step": 5725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.651217215103171e-06,
+      "loss": 0.099,
+      "step": 5730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.649790233739547e-06,
+      "loss": 0.1983,
+      "step": 5735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.648363252375924e-06,
+      "loss": 0.0715,
+      "step": 5740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.646936271012302e-06,
+      "loss": 0.2699,
+      "step": 5745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.645509289648678e-06,
+      "loss": 0.4508,
+      "step": 5750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.644082308285054e-06,
+      "loss": 0.1935,
+      "step": 5755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.64265532692143e-06,
+      "loss": 0.1979,
+      "step": 5760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.641228345557807e-06,
+      "loss": 0.1516,
+      "step": 5765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.639801364194185e-06,
+      "loss": 0.1582,
+      "step": 5770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.638374382830561e-06,
+      "loss": 0.1515,
+      "step": 5775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.636947401466938e-06,
+      "loss": 0.0538,
+      "step": 5780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.635520420103314e-06,
+      "loss": 0.1432,
+      "step": 5785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.634093438739692e-06,
+      "loss": 0.266,
+      "step": 5790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.632666457376068e-06,
+      "loss": 0.5542,
+      "step": 5795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.631239476012445e-06,
+      "loss": 0.2882,
+      "step": 5800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.629812494648821e-06,
+      "loss": 0.2182,
+      "step": 5805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.628385513285197e-06,
+      "loss": 0.2666,
+      "step": 5810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.626958531921574e-06,
+      "loss": 0.1588,
+      "step": 5815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.62553155055795e-06,
+      "loss": 0.241,
+      "step": 5820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.624104569194326e-06,
+      "loss": 0.0942,
+      "step": 5825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.622677587830703e-06,
+      "loss": 0.0784,
+      "step": 5830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.621250606467079e-06,
+      "loss": 0.1611,
+      "step": 5835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.619823625103457e-06,
+      "loss": 0.1319,
+      "step": 5840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.618396643739833e-06,
+      "loss": 0.4533,
+      "step": 5845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.61696966237621e-06,
+      "loss": 0.4332,
+      "step": 5850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.615542681012586e-06,
+      "loss": 0.1972,
+      "step": 5855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.614115699648962e-06,
+      "loss": 0.2395,
+      "step": 5860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.61268871828534e-06,
+      "loss": 0.2174,
+      "step": 5865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.611261736921717e-06,
+      "loss": 0.2569,
+      "step": 5870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.609834755558093e-06,
+      "loss": 0.1612,
+      "step": 5875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.60840777419447e-06,
+      "loss": 0.1016,
+      "step": 5880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.606980792830848e-06,
+      "loss": 0.1382,
+      "step": 5885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.605553811467224e-06,
+      "loss": 0.2932,
+      "step": 5890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.6041268301036e-06,
+      "loss": 0.3552,
+      "step": 5895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.602699848739977e-06,
+      "loss": 0.6778,
+      "step": 5900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.601272867376353e-06,
+      "loss": 0.1395,
+      "step": 5905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.59984588601273e-06,
+      "loss": 0.1618,
+      "step": 5910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.598418904649106e-06,
+      "loss": 0.2916,
+      "step": 5915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.596991923285482e-06,
+      "loss": 0.2307,
+      "step": 5920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.595564941921858e-06,
+      "loss": 0.1712,
+      "step": 5925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.594137960558235e-06,
+      "loss": 0.1044,
+      "step": 5930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.592710979194613e-06,
+      "loss": 0.1749,
+      "step": 5935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.591283997830989e-06,
+      "loss": 0.2064,
+      "step": 5940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.589857016467365e-06,
+      "loss": 0.2992,
+      "step": 5945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.588430035103742e-06,
+      "loss": 0.3827,
+      "step": 5950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.587003053740118e-06,
+      "loss": 0.2054,
+      "step": 5955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.585576072376496e-06,
+      "loss": 0.1882,
+      "step": 5960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.584149091012872e-06,
+      "loss": 0.1752,
+      "step": 5965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.582722109649249e-06,
+      "loss": 0.2178,
+      "step": 5970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.581295128285625e-06,
+      "loss": 0.1532,
+      "step": 5975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.579868146922003e-06,
+      "loss": 0.162,
+      "step": 5980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.57844116555838e-06,
+      "loss": 0.2166,
+      "step": 5985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.577014184194756e-06,
+      "loss": 0.2177,
+      "step": 5990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.575587202831132e-06,
+      "loss": 0.2964,
+      "step": 5995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.574160221467508e-06,
+      "loss": 0.3765,
+      "step": 6000
+    },
+    {
+      "epoch": 0.17,
+      "eval_loss": 0.0795619785785675,
+      "eval_runtime": 3436.6447,
+      "eval_samples_per_second": 4.661,
+      "eval_steps_per_second": 1.165,
+      "eval_wer": 0.2294266137756942,
+      "step": 6000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.572733240103885e-06,
+      "loss": 0.1503,
+      "step": 6005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.571306258740261e-06,
+      "loss": 0.2646,
+      "step": 6010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.569879277376637e-06,
+      "loss": 0.2817,
+      "step": 6015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.568452296013014e-06,
+      "loss": 0.2402,
+      "step": 6020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.56702531464939e-06,
+      "loss": 0.1246,
+      "step": 6025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.565598333285768e-06,
+      "loss": 0.1411,
+      "step": 6030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.564171351922144e-06,
+      "loss": 0.2098,
+      "step": 6035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.56274437055852e-06,
+      "loss": 0.1542,
+      "step": 6040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.561317389194897e-06,
+      "loss": 0.4184,
+      "step": 6045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.559890407831273e-06,
+      "loss": 0.4777,
+      "step": 6050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.558463426467652e-06,
+      "loss": 0.1435,
+      "step": 6055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.557036445104028e-06,
+      "loss": 0.1859,
+      "step": 6060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.555609463740404e-06,
+      "loss": 0.1865,
+      "step": 6065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.55418248237678e-06,
+      "loss": 0.2156,
+      "step": 6070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.552755501013159e-06,
+      "loss": 0.1261,
+      "step": 6075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.551328519649535e-06,
+      "loss": 0.2086,
+      "step": 6080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.549901538285911e-06,
+      "loss": 0.0838,
+      "step": 6085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.548474556922288e-06,
+      "loss": 0.2231,
+      "step": 6090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.547047575558664e-06,
+      "loss": 0.2857,
+      "step": 6095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.545905990467766e-06,
+      "loss": 0.4935,
+      "step": 6100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.544479009104142e-06,
+      "loss": 0.2334,
+      "step": 6105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.543052027740519e-06,
+      "loss": 0.2609,
+      "step": 6110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.541625046376895e-06,
+      "loss": 0.2063,
+      "step": 6115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.540198065013271e-06,
+      "loss": 0.1624,
+      "step": 6120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.538771083649648e-06,
+      "loss": 0.2474,
+      "step": 6125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.537344102286024e-06,
+      "loss": 0.1688,
+      "step": 6130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.5359171209224e-06,
+      "loss": 0.102,
+      "step": 6135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.534490139558778e-06,
+      "loss": 0.1941,
+      "step": 6140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.533063158195155e-06,
+      "loss": 0.2409,
+      "step": 6145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.531636176831531e-06,
+      "loss": 0.5918,
+      "step": 6150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.530209195467907e-06,
+      "loss": 0.6058,
+      "step": 6155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.528782214104284e-06,
+      "loss": 0.2429,
+      "step": 6160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.52735523274066e-06,
+      "loss": 0.2443,
+      "step": 6165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.525928251377038e-06,
+      "loss": 0.198,
+      "step": 6170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.524501270013414e-06,
+      "loss": 0.1545,
+      "step": 6175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.52307428864979e-06,
+      "loss": 0.1433,
+      "step": 6180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.521647307286167e-06,
+      "loss": 0.0983,
+      "step": 6185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.520220325922545e-06,
+      "loss": 0.1364,
+      "step": 6190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.518793344558922e-06,
+      "loss": 0.1471,
+      "step": 6195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.517366363195298e-06,
+      "loss": 0.5086,
+      "step": 6200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.515939381831674e-06,
+      "loss": 0.1766,
+      "step": 6205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.51451240046805e-06,
+      "loss": 0.2234,
+      "step": 6210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.513085419104427e-06,
+      "loss": 0.2427,
+      "step": 6215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.511658437740803e-06,
+      "loss": 0.2137,
+      "step": 6220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.51023145637718e-06,
+      "loss": 0.0814,
+      "step": 6225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.508804475013558e-06,
+      "loss": 0.2149,
+      "step": 6230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.507377493649934e-06,
+      "loss": 0.1671,
+      "step": 6235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.50595051228631e-06,
+      "loss": 0.2632,
+      "step": 6240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.504523530922687e-06,
+      "loss": 0.3538,
+      "step": 6245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.503096549559063e-06,
+      "loss": 0.5223,
+      "step": 6250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.50166956819544e-06,
+      "loss": 0.1936,
+      "step": 6255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.500242586831816e-06,
+      "loss": 0.1584,
+      "step": 6260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.498815605468194e-06,
+      "loss": 0.1398,
+      "step": 6265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.49738862410457e-06,
+      "loss": 0.1843,
+      "step": 6270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.495961642740946e-06,
+      "loss": 0.1717,
+      "step": 6275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.494534661377323e-06,
+      "loss": 0.0737,
+      "step": 6280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.4931076800137e-06,
+      "loss": 0.0921,
+      "step": 6285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.491680698650077e-06,
+      "loss": 0.2542,
+      "step": 6290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.490253717286453e-06,
+      "loss": 0.2339,
+      "step": 6295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.48882673592283e-06,
+      "loss": 0.5075,
+      "step": 6300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.487399754559206e-06,
+      "loss": 0.1163,
+      "step": 6305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.485972773195582e-06,
+      "loss": 0.1872,
+      "step": 6310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.484545791831959e-06,
+      "loss": 0.1719,
+      "step": 6315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.483118810468337e-06,
+      "loss": 0.2333,
+      "step": 6320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.481691829104713e-06,
+      "loss": 0.1895,
+      "step": 6325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.48026484774109e-06,
+      "loss": 0.0974,
+      "step": 6330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.478837866377466e-06,
+      "loss": 0.1817,
+      "step": 6335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.477410885013842e-06,
+      "loss": 0.2273,
+      "step": 6340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.475983903650218e-06,
+      "loss": 0.3048,
+      "step": 6345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.474556922286595e-06,
+      "loss": 0.4877,
+      "step": 6350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.473129940922971e-06,
+      "loss": 0.1744,
+      "step": 6355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.47170295955935e-06,
+      "loss": 0.138,
+      "step": 6360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.470275978195726e-06,
+      "loss": 0.108,
+      "step": 6365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.468848996832102e-06,
+      "loss": 0.1577,
+      "step": 6370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.467422015468478e-06,
+      "loss": 0.2492,
+      "step": 6375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.465995034104856e-06,
+      "loss": 0.0848,
+      "step": 6380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.464568052741233e-06,
+      "loss": 0.1697,
+      "step": 6385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.463141071377609e-06,
+      "loss": 0.3038,
+      "step": 6390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.461714090013985e-06,
+      "loss": 0.4789,
+      "step": 6395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.460287108650362e-06,
+      "loss": 0.5222,
+      "step": 6400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.45886012728674e-06,
+      "loss": 0.1689,
+      "step": 6405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.457433145923116e-06,
+      "loss": 0.1895,
+      "step": 6410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.456006164559492e-06,
+      "loss": 0.3185,
+      "step": 6415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.454579183195869e-06,
+      "loss": 0.2007,
+      "step": 6420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.453152201832245e-06,
+      "loss": 0.0862,
+      "step": 6425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.451725220468621e-06,
+      "loss": 0.1575,
+      "step": 6430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.450298239104998e-06,
+      "loss": 0.1323,
+      "step": 6435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.448871257741374e-06,
+      "loss": 0.205,
+      "step": 6440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.44744427637775e-06,
+      "loss": 0.4507,
+      "step": 6445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.446017295014127e-06,
+      "loss": 0.6509,
+      "step": 6450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.444590313650505e-06,
+      "loss": 0.1532,
+      "step": 6455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.443163332286881e-06,
+      "loss": 0.1993,
+      "step": 6460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.441736350923257e-06,
+      "loss": 0.2075,
+      "step": 6465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.440309369559634e-06,
+      "loss": 0.2406,
+      "step": 6470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.438882388196012e-06,
+      "loss": 0.1678,
+      "step": 6475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.437455406832388e-06,
+      "loss": 0.0898,
+      "step": 6480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.436028425468764e-06,
+      "loss": 0.0728,
+      "step": 6485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.43460144410514e-06,
+      "loss": 0.178,
+      "step": 6490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.433174462741517e-06,
+      "loss": 0.328,
+      "step": 6495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.431747481377895e-06,
+      "loss": 0.5988,
+      "step": 6500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.430320500014271e-06,
+      "loss": 0.2281,
+      "step": 6505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.428893518650648e-06,
+      "loss": 0.1533,
+      "step": 6510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.427466537287024e-06,
+      "loss": 0.1882,
+      "step": 6515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.4260395559234e-06,
+      "loss": 0.1461,
+      "step": 6520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.424612574559777e-06,
+      "loss": 0.1125,
+      "step": 6525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.423185593196153e-06,
+      "loss": 0.2008,
+      "step": 6530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.42175861183253e-06,
+      "loss": 0.1642,
+      "step": 6535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.420331630468906e-06,
+      "loss": 0.3273,
+      "step": 6540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.418904649105282e-06,
+      "loss": 0.1172,
+      "step": 6545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.41747766774166e-06,
+      "loss": 0.5899,
+      "step": 6550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.416050686378037e-06,
+      "loss": 0.226,
+      "step": 6555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.414623705014413e-06,
+      "loss": 0.1787,
+      "step": 6560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.41319672365079e-06,
+      "loss": 0.1932,
+      "step": 6565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.411769742287167e-06,
+      "loss": 0.3157,
+      "step": 6570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.410342760923544e-06,
+      "loss": 0.1779,
+      "step": 6575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.40891577955992e-06,
+      "loss": 0.0548,
+      "step": 6580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.407488798196296e-06,
+      "loss": 0.2249,
+      "step": 6585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.406061816832673e-06,
+      "loss": 0.3006,
+      "step": 6590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.40463483546905e-06,
+      "loss": 0.4068,
+      "step": 6595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.403207854105427e-06,
+      "loss": 0.4126,
+      "step": 6600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.401780872741803e-06,
+      "loss": 0.2776,
+      "step": 6605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.40035389137818e-06,
+      "loss": 0.2043,
+      "step": 6610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.398926910014556e-06,
+      "loss": 0.202,
+      "step": 6615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.397499928650932e-06,
+      "loss": 0.2153,
+      "step": 6620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.396072947287309e-06,
+      "loss": 0.1042,
+      "step": 6625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.394645965923685e-06,
+      "loss": 0.1125,
+      "step": 6630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.393218984560061e-06,
+      "loss": 0.2851,
+      "step": 6635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.39179200319644e-06,
+      "loss": 0.2376,
+      "step": 6640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.390365021832816e-06,
+      "loss": 0.3222,
+      "step": 6645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.388938040469192e-06,
+      "loss": 0.4322,
+      "step": 6650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.387511059105568e-06,
+      "loss": 0.1321,
+      "step": 6655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.386084077741945e-06,
+      "loss": 0.2256,
+      "step": 6660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.384657096378323e-06,
+      "loss": 0.1782,
+      "step": 6665
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.383230115014699e-06,
+      "loss": 0.1718,
+      "step": 6670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.381803133651075e-06,
+      "loss": 0.241,
+      "step": 6675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.380376152287452e-06,
+      "loss": 0.2157,
+      "step": 6680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.378949170923828e-06,
+      "loss": 0.2164,
+      "step": 6685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.377522189560206e-06,
+      "loss": 0.1748,
+      "step": 6690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.376095208196583e-06,
+      "loss": 0.1148,
+      "step": 6695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.374668226832959e-06,
+      "loss": 0.3105,
+      "step": 6700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.373241245469335e-06,
+      "loss": 0.2214,
+      "step": 6705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.371814264105712e-06,
+      "loss": 0.1674,
+      "step": 6710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.370387282742088e-06,
+      "loss": 0.1968,
+      "step": 6715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.368960301378464e-06,
+      "loss": 0.2137,
+      "step": 6720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.36753332001484e-06,
+      "loss": 0.1645,
+      "step": 6725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.366106338651217e-06,
+      "loss": 0.1292,
+      "step": 6730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.364679357287595e-06,
+      "loss": 0.1719,
+      "step": 6735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.363252375923971e-06,
+      "loss": 0.2621,
+      "step": 6740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.361825394560348e-06,
+      "loss": 0.5572,
+      "step": 6745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.360398413196724e-06,
+      "loss": 0.4367,
+      "step": 6750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.3589714318331e-06,
+      "loss": 0.1601,
+      "step": 6755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.357544450469478e-06,
+      "loss": 0.2053,
+      "step": 6760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.356117469105855e-06,
+      "loss": 0.1942,
+      "step": 6765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.354690487742231e-06,
+      "loss": 0.1708,
+      "step": 6770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.353263506378607e-06,
+      "loss": 0.2491,
+      "step": 6775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.351836525014984e-06,
+      "loss": 0.1236,
+      "step": 6780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.350409543651362e-06,
+      "loss": 0.2253,
+      "step": 6785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.348982562287738e-06,
+      "loss": 0.2409,
+      "step": 6790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.347555580924114e-06,
+      "loss": 0.2078,
+      "step": 6795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.34612859956049e-06,
+      "loss": 0.4089,
+      "step": 6800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.344701618196867e-06,
+      "loss": 0.1342,
+      "step": 6805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.343274636833243e-06,
+      "loss": 0.1629,
+      "step": 6810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.34184765546962e-06,
+      "loss": 0.149,
+      "step": 6815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.340420674105996e-06,
+      "loss": 0.1979,
+      "step": 6820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.338993692742372e-06,
+      "loss": 0.1371,
+      "step": 6825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.33756671137875e-06,
+      "loss": 0.1345,
+      "step": 6830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.336139730015127e-06,
+      "loss": 0.3031,
+      "step": 6835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.334712748651503e-06,
+      "loss": 0.0934,
+      "step": 6840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.33328576728788e-06,
+      "loss": 0.3847,
+      "step": 6845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.331858785924256e-06,
+      "loss": 0.4262,
+      "step": 6850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.330431804560634e-06,
+      "loss": 0.2704,
+      "step": 6855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.32900482319701e-06,
+      "loss": 0.174,
+      "step": 6860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.327577841833386e-06,
+      "loss": 0.203,
+      "step": 6865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.326150860469763e-06,
+      "loss": 0.1766,
+      "step": 6870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.32472387910614e-06,
+      "loss": 0.1113,
+      "step": 6875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.323296897742517e-06,
+      "loss": 0.1789,
+      "step": 6880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.321869916378894e-06,
+      "loss": 0.1672,
+      "step": 6885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.32044293501527e-06,
+      "loss": 0.2633,
+      "step": 6890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.319015953651646e-06,
+      "loss": 0.3725,
+      "step": 6895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.317588972288023e-06,
+      "loss": 0.3129,
+      "step": 6900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.316161990924399e-06,
+      "loss": 0.2139,
+      "step": 6905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.314735009560775e-06,
+      "loss": 0.3623,
+      "step": 6910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.313308028197152e-06,
+      "loss": 0.1442,
+      "step": 6915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.311881046833528e-06,
+      "loss": 0.1482,
+      "step": 6920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.310454065469906e-06,
+      "loss": 0.2105,
+      "step": 6925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.309027084106282e-06,
+      "loss": 0.1828,
+      "step": 6930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.307600102742659e-06,
+      "loss": 0.1349,
+      "step": 6935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.306173121379035e-06,
+      "loss": 0.2098,
+      "step": 6940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.304746140015411e-06,
+      "loss": 0.3807,
+      "step": 6945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.30331915865179e-06,
+      "loss": 0.5789,
+      "step": 6950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.301892177288166e-06,
+      "loss": 0.1509,
+      "step": 6955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.300465195924542e-06,
+      "loss": 0.1647,
+      "step": 6960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.299038214560918e-06,
+      "loss": 0.1773,
+      "step": 6965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.297611233197295e-06,
+      "loss": 0.1803,
+      "step": 6970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.296184251833673e-06,
+      "loss": 0.2572,
+      "step": 6975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.294757270470049e-06,
+      "loss": 0.0578,
+      "step": 6980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.293330289106425e-06,
+      "loss": 0.194,
+      "step": 6985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.291903307742802e-06,
+      "loss": 0.2055,
+      "step": 6990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.290476326379178e-06,
+      "loss": 0.3183,
+      "step": 6995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.289049345015554e-06,
+      "loss": 0.5604,
+      "step": 7000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.28762236365193e-06,
+      "loss": 0.1832,
+      "step": 7005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.286195382288307e-06,
+      "loss": 0.229,
+      "step": 7010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.284768400924683e-06,
+      "loss": 0.1836,
+      "step": 7015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.283341419561061e-06,
+      "loss": 0.2354,
+      "step": 7020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.281914438197438e-06,
+      "loss": 0.1575,
+      "step": 7025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.280487456833814e-06,
+      "loss": 0.047,
+      "step": 7030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.27906047547019e-06,
+      "loss": 0.0861,
+      "step": 7035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.277633494106567e-06,
+      "loss": 0.2914,
+      "step": 7040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.276206512742945e-06,
+      "loss": 0.1699,
+      "step": 7045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.274779531379321e-06,
+      "loss": 0.3493,
+      "step": 7050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.273352550015698e-06,
+      "loss": 0.1849,
+      "step": 7055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.271925568652074e-06,
+      "loss": 0.1878,
+      "step": 7060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.27049858728845e-06,
+      "loss": 0.1817,
+      "step": 7065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.269071605924828e-06,
+      "loss": 0.1612,
+      "step": 7070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.267644624561205e-06,
+      "loss": 0.1838,
+      "step": 7075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.266217643197581e-06,
+      "loss": 0.1406,
+      "step": 7080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.264790661833957e-06,
+      "loss": 0.2158,
+      "step": 7085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.263363680470334e-06,
+      "loss": 0.3149,
+      "step": 7090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.26193669910671e-06,
+      "loss": 0.3768,
+      "step": 7095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.260509717743086e-06,
+      "loss": 0.4874,
+      "step": 7100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.259082736379464e-06,
+      "loss": 0.1532,
+      "step": 7105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.25765575501584e-06,
+      "loss": 0.1728,
+      "step": 7110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.256228773652217e-06,
+      "loss": 0.162,
+      "step": 7115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.254801792288593e-06,
+      "loss": 0.2831,
+      "step": 7120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.25337481092497e-06,
+      "loss": 0.1868,
+      "step": 7125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.251947829561346e-06,
+      "loss": 0.0658,
+      "step": 7130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.250520848197722e-06,
+      "loss": 0.2732,
+      "step": 7135
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.2490938668341e-06,
+      "loss": 0.1974,
+      "step": 7140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.247666885470477e-06,
+      "loss": 0.21,
+      "step": 7145
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.246239904106853e-06,
+      "loss": 0.3657,
+      "step": 7150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.24481292274323e-06,
+      "loss": 0.1833,
+      "step": 7155
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.243385941379606e-06,
+      "loss": 0.1515,
+      "step": 7160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.241958960015984e-06,
+      "loss": 0.2312,
+      "step": 7165
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.24053197865236e-06,
+      "loss": 0.1856,
+      "step": 7170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.239104997288736e-06,
+      "loss": 0.2191,
+      "step": 7175
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.237678015925113e-06,
+      "loss": 0.1461,
+      "step": 7180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.236251034561489e-06,
+      "loss": 0.1361,
+      "step": 7185
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.234824053197865e-06,
+      "loss": 0.1592,
+      "step": 7190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.233397071834243e-06,
+      "loss": 0.1711,
+      "step": 7195
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.23197009047062e-06,
+      "loss": 0.4492,
+      "step": 7200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.230543109106996e-06,
+      "loss": 0.2174,
+      "step": 7205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.229116127743372e-06,
+      "loss": 0.1999,
+      "step": 7210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.227689146379749e-06,
+      "loss": 0.1429,
+      "step": 7215
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.226262165016125e-06,
+      "loss": 0.2191,
+      "step": 7220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.224835183652502e-06,
+      "loss": 0.171,
+      "step": 7225
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.223408202288878e-06,
+      "loss": 0.1044,
+      "step": 7230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.221981220925256e-06,
+      "loss": 0.1673,
+      "step": 7235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.220554239561632e-06,
+      "loss": 0.1706,
+      "step": 7240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.219127258198009e-06,
+      "loss": 0.178,
+      "step": 7245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.217700276834385e-06,
+      "loss": 0.6589,
+      "step": 7250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.216273295470761e-06,
+      "loss": 0.1911,
+      "step": 7255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.21484631410714e-06,
+      "loss": 0.1498,
+      "step": 7260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.213419332743516e-06,
+      "loss": 0.1801,
+      "step": 7265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.211992351379892e-06,
+      "loss": 0.1972,
+      "step": 7270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.210565370016268e-06,
+      "loss": 0.1474,
+      "step": 7275
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.209138388652645e-06,
+      "loss": 0.1972,
+      "step": 7280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.207711407289023e-06,
+      "loss": 0.2453,
+      "step": 7285
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.206284425925399e-06,
+      "loss": 0.2086,
+      "step": 7290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.204857444561775e-06,
+      "loss": 0.1912,
+      "step": 7295
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.203430463198152e-06,
+      "loss": 0.3786,
+      "step": 7300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.202003481834528e-06,
+      "loss": 0.2067,
+      "step": 7305
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.200576500470904e-06,
+      "loss": 0.2041,
+      "step": 7310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.19914951910728e-06,
+      "loss": 0.244,
+      "step": 7315
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.197722537743657e-06,
+      "loss": 0.2493,
+      "step": 7320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.196295556380033e-06,
+      "loss": 0.0708,
+      "step": 7325
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.194868575016411e-06,
+      "loss": 0.0863,
+      "step": 7330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.193441593652788e-06,
+      "loss": 0.1413,
+      "step": 7335
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.192014612289164e-06,
+      "loss": 0.1141,
+      "step": 7340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.19058763092554e-06,
+      "loss": 0.4008,
+      "step": 7345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.189160649561917e-06,
+      "loss": 0.5528,
+      "step": 7350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.187733668198295e-06,
+      "loss": 0.1992,
+      "step": 7355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.186306686834671e-06,
+      "loss": 0.1314,
+      "step": 7360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.184879705471047e-06,
+      "loss": 0.1417,
+      "step": 7365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.183452724107424e-06,
+      "loss": 0.1609,
+      "step": 7370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.1820257427438e-06,
+      "loss": 0.1983,
+      "step": 7375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.180598761380178e-06,
+      "loss": 0.1147,
+      "step": 7380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.179171780016555e-06,
+      "loss": 0.1589,
+      "step": 7385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.17774479865293e-06,
+      "loss": 0.1664,
+      "step": 7390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.176317817289307e-06,
+      "loss": 0.4555,
+      "step": 7395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.174890835925684e-06,
+      "loss": 0.5475,
+      "step": 7400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.17346385456206e-06,
+      "loss": 0.1252,
+      "step": 7405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.172036873198436e-06,
+      "loss": 0.21,
+      "step": 7410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.170609891834813e-06,
+      "loss": 0.202,
+      "step": 7415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.169182910471189e-06,
+      "loss": 0.2249,
+      "step": 7420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.167755929107567e-06,
+      "loss": 0.2559,
+      "step": 7425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.166328947743943e-06,
+      "loss": 0.1324,
+      "step": 7430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.16490196638032e-06,
+      "loss": 0.2564,
+      "step": 7435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.163474985016696e-06,
+      "loss": 0.2318,
+      "step": 7440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.162048003653072e-06,
+      "loss": 0.2666,
+      "step": 7445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.16062102228945e-06,
+      "loss": 0.3466,
+      "step": 7450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.159194040925827e-06,
+      "loss": 0.2163,
+      "step": 7455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.157767059562203e-06,
+      "loss": 0.2175,
+      "step": 7460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.15634007819858e-06,
+      "loss": 0.2092,
+      "step": 7465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.154913096834956e-06,
+      "loss": 0.2197,
+      "step": 7470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.153486115471334e-06,
+      "loss": 0.0569,
+      "step": 7475
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.15205913410771e-06,
+      "loss": 0.0812,
+      "step": 7480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.150632152744086e-06,
+      "loss": 0.1031,
+      "step": 7485
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.149205171380463e-06,
+      "loss": 0.3571,
+      "step": 7490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.147778190016839e-06,
+      "loss": 0.1208,
+      "step": 7495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.146351208653215e-06,
+      "loss": 0.2976,
+      "step": 7500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.144924227289592e-06,
+      "loss": 0.1949,
+      "step": 7505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.143497245925968e-06,
+      "loss": 0.246,
+      "step": 7510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.142070264562344e-06,
+      "loss": 0.1885,
+      "step": 7515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.140643283198722e-06,
+      "loss": 0.2733,
+      "step": 7520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.139216301835099e-06,
+      "loss": 0.0634,
+      "step": 7525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.137789320471475e-06,
+      "loss": 0.1103,
+      "step": 7530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.136362339107851e-06,
+      "loss": 0.0912,
+      "step": 7535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.134935357744228e-06,
+      "loss": 0.2565,
+      "step": 7540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.133508376380606e-06,
+      "loss": 0.3867,
+      "step": 7545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.132081395016982e-06,
+      "loss": 0.4726,
+      "step": 7550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.130654413653358e-06,
+      "loss": 0.4183,
+      "step": 7555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.129227432289735e-06,
+      "loss": 0.1844,
+      "step": 7560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.127800450926111e-06,
+      "loss": 0.1948,
+      "step": 7565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.12637346956249e-06,
+      "loss": 0.2435,
+      "step": 7570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.124946488198866e-06,
+      "loss": 0.2355,
+      "step": 7575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.123519506835242e-06,
+      "loss": 0.1969,
+      "step": 7580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.122092525471618e-06,
+      "loss": 0.1432,
+      "step": 7585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.120665544107995e-06,
+      "loss": 0.2547,
+      "step": 7590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.119238562744371e-06,
+      "loss": 0.3437,
+      "step": 7595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.117811581380747e-06,
+      "loss": 0.6044,
+      "step": 7600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.116384600017124e-06,
+      "loss": 0.2163,
+      "step": 7605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.1149576186535e-06,
+      "loss": 0.1687,
+      "step": 7610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.113530637289878e-06,
+      "loss": 0.1551,
+      "step": 7615
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.112103655926254e-06,
+      "loss": 0.2271,
+      "step": 7620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.11067667456263e-06,
+      "loss": 0.1671,
+      "step": 7625
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.109249693199007e-06,
+      "loss": 0.0885,
+      "step": 7630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.107822711835383e-06,
+      "loss": 0.1219,
+      "step": 7635
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.106395730471761e-06,
+      "loss": 0.1902,
+      "step": 7640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.104968749108138e-06,
+      "loss": 0.2712,
+      "step": 7645
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.103541767744514e-06,
+      "loss": 0.4752,
+      "step": 7650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.10211478638089e-06,
+      "loss": 0.1733,
+      "step": 7655
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.100687805017267e-06,
+      "loss": 0.1289,
+      "step": 7660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.099260823653645e-06,
+      "loss": 0.2383,
+      "step": 7665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.097833842290021e-06,
+      "loss": 0.1959,
+      "step": 7670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.096406860926397e-06,
+      "loss": 0.1568,
+      "step": 7675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.094979879562774e-06,
+      "loss": 0.1829,
+      "step": 7680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.09355289819915e-06,
+      "loss": 0.1501,
+      "step": 7685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.092125916835526e-06,
+      "loss": 0.2989,
+      "step": 7690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.090698935471903e-06,
+      "loss": 0.2442,
+      "step": 7695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.089271954108279e-06,
+      "loss": 0.8114,
+      "step": 7700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.087844972744655e-06,
+      "loss": 0.1299,
+      "step": 7705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.086417991381033e-06,
+      "loss": 0.2395,
+      "step": 7710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.08499101001741e-06,
+      "loss": 0.1642,
+      "step": 7715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.083564028653786e-06,
+      "loss": 0.1507,
+      "step": 7720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.082137047290162e-06,
+      "loss": 0.1931,
+      "step": 7725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.080710065926539e-06,
+      "loss": 0.079,
+      "step": 7730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.079283084562917e-06,
+      "loss": 0.0929,
+      "step": 7735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.077856103199293e-06,
+      "loss": 0.1381,
+      "step": 7740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.07642912183567e-06,
+      "loss": 0.3335,
+      "step": 7745
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.075002140472046e-06,
+      "loss": 0.3351,
+      "step": 7750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.073575159108422e-06,
+      "loss": 0.1355,
+      "step": 7755
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.0721481777448e-06,
+      "loss": 0.1426,
+      "step": 7760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.070721196381177e-06,
+      "loss": 0.1558,
+      "step": 7765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.069294215017553e-06,
+      "loss": 0.1551,
+      "step": 7770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.06786723365393e-06,
+      "loss": 0.2197,
+      "step": 7775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.066440252290306e-06,
+      "loss": 0.0977,
+      "step": 7780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.065013270926682e-06,
+      "loss": 0.168,
+      "step": 7785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.063586289563058e-06,
+      "loss": 0.1557,
+      "step": 7790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.062159308199435e-06,
+      "loss": 0.1372,
+      "step": 7795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.060732326835811e-06,
+      "loss": 0.384,
+      "step": 7800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.059305345472189e-06,
+      "loss": 0.1653,
+      "step": 7805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.057878364108565e-06,
+      "loss": 0.1789,
+      "step": 7810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.056451382744942e-06,
+      "loss": 0.2406,
+      "step": 7815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.055024401381318e-06,
+      "loss": 0.1819,
+      "step": 7820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.053597420017694e-06,
+      "loss": 0.1301,
+      "step": 7825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.052170438654072e-06,
+      "loss": 0.1108,
+      "step": 7830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.050743457290449e-06,
+      "loss": 0.0467,
+      "step": 7835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.049316475926825e-06,
+      "loss": 0.3237,
+      "step": 7840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.047889494563201e-06,
+      "loss": 0.2159,
+      "step": 7845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.046462513199578e-06,
+      "loss": 0.4882,
+      "step": 7850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.045035531835956e-06,
+      "loss": 0.1892,
+      "step": 7855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.043608550472332e-06,
+      "loss": 0.1517,
+      "step": 7860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.042181569108708e-06,
+      "loss": 0.17,
+      "step": 7865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.040754587745085e-06,
+      "loss": 0.2468,
+      "step": 7870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.039327606381461e-06,
+      "loss": 0.1442,
+      "step": 7875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.037900625017837e-06,
+      "loss": 0.121,
+      "step": 7880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.036473643654214e-06,
+      "loss": 0.1657,
+      "step": 7885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.03504666229059e-06,
+      "loss": 0.1207,
+      "step": 7890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.033619680926968e-06,
+      "loss": 0.3081,
+      "step": 7895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.032192699563344e-06,
+      "loss": 0.4598,
+      "step": 7900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.03076571819972e-06,
+      "loss": 0.2183,
+      "step": 7905
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.029338736836097e-06,
+      "loss": 0.2336,
+      "step": 7910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.027911755472474e-06,
+      "loss": 0.1913,
+      "step": 7915
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.02648477410885e-06,
+      "loss": 0.2144,
+      "step": 7920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.025057792745228e-06,
+      "loss": 0.1693,
+      "step": 7925
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.023630811381604e-06,
+      "loss": 0.3176,
+      "step": 7930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.02220383001798e-06,
+      "loss": 0.2124,
+      "step": 7935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.020776848654357e-06,
+      "loss": 0.1698,
+      "step": 7940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.019349867290735e-06,
+      "loss": 0.1724,
+      "step": 7945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.017922885927111e-06,
+      "loss": 0.2481,
+      "step": 7950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.016495904563488e-06,
+      "loss": 0.1423,
+      "step": 7955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.015068923199864e-06,
+      "loss": 0.2494,
+      "step": 7960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.01364194183624e-06,
+      "loss": 0.2003,
+      "step": 7965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.012214960472617e-06,
+      "loss": 0.195,
+      "step": 7970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.010787979108993e-06,
+      "loss": 0.143,
+      "step": 7975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.00936099774537e-06,
+      "loss": 0.0434,
+      "step": 7980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.007934016381747e-06,
+      "loss": 0.2307,
+      "step": 7985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.006507035018124e-06,
+      "loss": 0.2078,
+      "step": 7990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.0050800536545e-06,
+      "loss": 0.316,
+      "step": 7995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.003653072290876e-06,
+      "loss": 0.2622,
+      "step": 8000
+    },
+    {
+      "epoch": 0.22,
+      "eval_loss": 0.08057957142591476,
+      "eval_runtime": 3022.947,
+      "eval_samples_per_second": 5.299,
+      "eval_steps_per_second": 1.325,
+      "eval_wer": 0.22843490804183195,
+      "step": 8000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.002226090927253e-06,
+      "loss": 0.2346,
+      "step": 8005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.000799109563629e-06,
+      "loss": 0.1882,
+      "step": 8010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.999372128200005e-06,
+      "loss": 0.1467,
+      "step": 8015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.997945146836383e-06,
+      "loss": 0.2359,
+      "step": 8020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.99651816547276e-06,
+      "loss": 0.1228,
+      "step": 8025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.995091184109136e-06,
+      "loss": 0.0632,
+      "step": 8030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.993664202745512e-06,
+      "loss": 0.1376,
+      "step": 8035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.99223722138189e-06,
+      "loss": 0.362,
+      "step": 8040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.990810240018267e-06,
+      "loss": 0.5213,
+      "step": 8045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.989383258654643e-06,
+      "loss": 0.5246,
+      "step": 8050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.98795627729102e-06,
+      "loss": 0.161,
+      "step": 8055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.986529295927396e-06,
+      "loss": 0.2175,
+      "step": 8060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.985102314563772e-06,
+      "loss": 0.1384,
+      "step": 8065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.983675333200148e-06,
+      "loss": 0.1738,
+      "step": 8070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.982248351836527e-06,
+      "loss": 0.2637,
+      "step": 8075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.980821370472903e-06,
+      "loss": 0.1429,
+      "step": 8080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.97939438910928e-06,
+      "loss": 0.189,
+      "step": 8085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.977967407745656e-06,
+      "loss": 0.1426,
+      "step": 8090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.976540426382032e-06,
+      "loss": 0.2112,
+      "step": 8095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.975113445018408e-06,
+      "loss": 0.2782,
+      "step": 8100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.973686463654785e-06,
+      "loss": 0.1385,
+      "step": 8105
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.972259482291161e-06,
+      "loss": 0.2095,
+      "step": 8110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.970832500927539e-06,
+      "loss": 0.1658,
+      "step": 8115
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.969405519563915e-06,
+      "loss": 0.1805,
+      "step": 8120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.967978538200292e-06,
+      "loss": 0.224,
+      "step": 8125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.966551556836668e-06,
+      "loss": 0.1345,
+      "step": 8130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.965124575473046e-06,
+      "loss": 0.091,
+      "step": 8135
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.963697594109422e-06,
+      "loss": 0.1689,
+      "step": 8140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.962270612745799e-06,
+      "loss": 0.3499,
+      "step": 8145
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.960843631382175e-06,
+      "loss": 0.3034,
+      "step": 8150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.959416650018551e-06,
+      "loss": 0.2078,
+      "step": 8155
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.95798966865493e-06,
+      "loss": 0.1228,
+      "step": 8160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.956562687291306e-06,
+      "loss": 0.1752,
+      "step": 8165
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.955135705927682e-06,
+      "loss": 0.1979,
+      "step": 8170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.953708724564058e-06,
+      "loss": 0.2338,
+      "step": 8175
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.952281743200435e-06,
+      "loss": 0.161,
+      "step": 8180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.950854761836811e-06,
+      "loss": 0.1555,
+      "step": 8185
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.949427780473187e-06,
+      "loss": 0.1691,
+      "step": 8190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.948000799109564e-06,
+      "loss": 0.2433,
+      "step": 8195
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.94657381774594e-06,
+      "loss": 0.35,
+      "step": 8200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.945146836382316e-06,
+      "loss": 0.183,
+      "step": 8205
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.943719855018694e-06,
+      "loss": 0.1599,
+      "step": 8210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.94229287365507e-06,
+      "loss": 0.1263,
+      "step": 8215
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.940865892291447e-06,
+      "loss": 0.2152,
+      "step": 8220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.939438910927823e-06,
+      "loss": 0.1271,
+      "step": 8225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.938011929564201e-06,
+      "loss": 0.0165,
+      "step": 8230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.936584948200578e-06,
+      "loss": 0.1579,
+      "step": 8235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.935157966836954e-06,
+      "loss": 0.3127,
+      "step": 8240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.93373098547333e-06,
+      "loss": 0.1991,
+      "step": 8245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.932304004109707e-06,
+      "loss": 0.5045,
+      "step": 8250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.930877022746085e-06,
+      "loss": 0.1249,
+      "step": 8255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.929450041382461e-06,
+      "loss": 0.1456,
+      "step": 8260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.928023060018838e-06,
+      "loss": 0.163,
+      "step": 8265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.926596078655214e-06,
+      "loss": 0.2551,
+      "step": 8270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.92516909729159e-06,
+      "loss": 0.0672,
+      "step": 8275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.923742115927967e-06,
+      "loss": 0.1371,
+      "step": 8280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.922315134564343e-06,
+      "loss": 0.1814,
+      "step": 8285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.92088815320072e-06,
+      "loss": 0.2584,
+      "step": 8290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.919461171837096e-06,
+      "loss": 0.3287,
+      "step": 8295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.918319586746198e-06,
+      "loss": 0.4046,
+      "step": 8300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.916892605382574e-06,
+      "loss": 0.1849,
+      "step": 8305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.91546562401895e-06,
+      "loss": 0.1787,
+      "step": 8310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.914038642655327e-06,
+      "loss": 0.233,
+      "step": 8315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.912611661291703e-06,
+      "loss": 0.1461,
+      "step": 8320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.911184679928081e-06,
+      "loss": 0.0856,
+      "step": 8325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.909757698564457e-06,
+      "loss": 0.2195,
+      "step": 8330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.908330717200834e-06,
+      "loss": 0.1154,
+      "step": 8335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.90690373583721e-06,
+      "loss": 0.2593,
+      "step": 8340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.905476754473588e-06,
+      "loss": 0.1829,
+      "step": 8345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.904049773109964e-06,
+      "loss": 0.5558,
+      "step": 8350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.90262279174634e-06,
+      "loss": 0.1854,
+      "step": 8355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.901195810382717e-06,
+      "loss": 0.2554,
+      "step": 8360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.899768829019093e-06,
+      "loss": 0.2087,
+      "step": 8365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.898341847655471e-06,
+      "loss": 0.18,
+      "step": 8370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.896914866291848e-06,
+      "loss": 0.2436,
+      "step": 8375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.895487884928224e-06,
+      "loss": 0.0874,
+      "step": 8380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.8940609035646e-06,
+      "loss": 0.1586,
+      "step": 8385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.892633922200977e-06,
+      "loss": 0.2233,
+      "step": 8390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.891206940837353e-06,
+      "loss": 0.3407,
+      "step": 8395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.88977995947373e-06,
+      "loss": 0.3159,
+      "step": 8400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.888352978110106e-06,
+      "loss": 0.1739,
+      "step": 8405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.886925996746482e-06,
+      "loss": 0.1841,
+      "step": 8410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.885499015382859e-06,
+      "loss": 0.1759,
+      "step": 8415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.884072034019237e-06,
+      "loss": 0.2541,
+      "step": 8420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.882645052655613e-06,
+      "loss": 0.0824,
+      "step": 8425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.88121807129199e-06,
+      "loss": 0.0575,
+      "step": 8430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.879791089928366e-06,
+      "loss": 0.1388,
+      "step": 8435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.878364108564744e-06,
+      "loss": 0.1756,
+      "step": 8440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.87693712720112e-06,
+      "loss": 0.2056,
+      "step": 8445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.875510145837496e-06,
+      "loss": 0.6129,
+      "step": 8450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.874083164473873e-06,
+      "loss": 0.1343,
+      "step": 8455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.872656183110249e-06,
+      "loss": 0.1692,
+      "step": 8460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.871229201746627e-06,
+      "loss": 0.193,
+      "step": 8465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.869802220383003e-06,
+      "loss": 0.1887,
+      "step": 8470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.86837523901938e-06,
+      "loss": 0.1236,
+      "step": 8475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.866948257655756e-06,
+      "loss": 0.1449,
+      "step": 8480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.865521276292132e-06,
+      "loss": 0.1807,
+      "step": 8485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.864094294928509e-06,
+      "loss": 0.0956,
+      "step": 8490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.862667313564885e-06,
+      "loss": 0.2542,
+      "step": 8495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.861240332201261e-06,
+      "loss": 0.4587,
+      "step": 8500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.859813350837638e-06,
+      "loss": 0.2121,
+      "step": 8505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.858386369474014e-06,
+      "loss": 0.3679,
+      "step": 8510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.856959388110392e-06,
+      "loss": 0.1392,
+      "step": 8515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.855532406746768e-06,
+      "loss": 0.1165,
+      "step": 8520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.854105425383145e-06,
+      "loss": 0.204,
+      "step": 8525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.852678444019521e-06,
+      "loss": 0.1654,
+      "step": 8530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.851251462655899e-06,
+      "loss": 0.1839,
+      "step": 8535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.849824481292275e-06,
+      "loss": 0.1727,
+      "step": 8540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.848397499928652e-06,
+      "loss": 0.2335,
+      "step": 8545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.846970518565028e-06,
+      "loss": 0.4425,
+      "step": 8550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.845543537201404e-06,
+      "loss": 0.1648,
+      "step": 8555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.844116555837783e-06,
+      "loss": 0.1482,
+      "step": 8560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.842689574474159e-06,
+      "loss": 0.1983,
+      "step": 8565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.841262593110535e-06,
+      "loss": 0.2705,
+      "step": 8570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.839835611746912e-06,
+      "loss": 0.1826,
+      "step": 8575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.838408630383288e-06,
+      "loss": 0.1382,
+      "step": 8580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.836981649019664e-06,
+      "loss": 0.2028,
+      "step": 8585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.83555466765604e-06,
+      "loss": 0.098,
+      "step": 8590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.834127686292417e-06,
+      "loss": 0.2188,
+      "step": 8595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.832700704928793e-06,
+      "loss": 0.6639,
+      "step": 8600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.83127372356517e-06,
+      "loss": 0.1766,
+      "step": 8605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.829846742201548e-06,
+      "loss": 0.1344,
+      "step": 8610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.828419760837924e-06,
+      "loss": 0.1565,
+      "step": 8615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.8269927794743e-06,
+      "loss": 0.2745,
+      "step": 8620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.825565798110677e-06,
+      "loss": 0.2157,
+      "step": 8625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.824138816747055e-06,
+      "loss": 0.095,
+      "step": 8630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.822711835383431e-06,
+      "loss": 0.1851,
+      "step": 8635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.821284854019807e-06,
+      "loss": 0.2995,
+      "step": 8640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.819857872656184e-06,
+      "loss": 0.3037,
+      "step": 8645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.81843089129256e-06,
+      "loss": 0.3917,
+      "step": 8650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.817003909928938e-06,
+      "loss": 0.1416,
+      "step": 8655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.815576928565314e-06,
+      "loss": 0.2278,
+      "step": 8660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.81414994720169e-06,
+      "loss": 0.2162,
+      "step": 8665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.812722965838067e-06,
+      "loss": 0.2055,
+      "step": 8670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.811295984474443e-06,
+      "loss": 0.0889,
+      "step": 8675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.80986900311082e-06,
+      "loss": 0.1144,
+      "step": 8680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.808442021747196e-06,
+      "loss": 0.2463,
+      "step": 8685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.807015040383572e-06,
+      "loss": 0.3082,
+      "step": 8690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.805588059019949e-06,
+      "loss": 0.1998,
+      "step": 8695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.804161077656325e-06,
+      "loss": 0.288,
+      "step": 8700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.802734096292703e-06,
+      "loss": 0.1421,
+      "step": 8705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.80130711492908e-06,
+      "loss": 0.1546,
+      "step": 8710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.799880133565456e-06,
+      "loss": 0.2216,
+      "step": 8715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.798453152201832e-06,
+      "loss": 0.2045,
+      "step": 8720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.79702617083821e-06,
+      "loss": 0.2076,
+      "step": 8725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.795599189474586e-06,
+      "loss": 0.0846,
+      "step": 8730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.794172208110963e-06,
+      "loss": 0.2256,
+      "step": 8735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.79274522674734e-06,
+      "loss": 0.1954,
+      "step": 8740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.791318245383716e-06,
+      "loss": 0.0833,
+      "step": 8745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.789891264020094e-06,
+      "loss": 0.6029,
+      "step": 8750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.78846428265647e-06,
+      "loss": 0.1641,
+      "step": 8755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.787037301292846e-06,
+      "loss": 0.2114,
+      "step": 8760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.785610319929223e-06,
+      "loss": 0.1914,
+      "step": 8765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.784183338565599e-06,
+      "loss": 0.2398,
+      "step": 8770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.782756357201975e-06,
+      "loss": 0.1619,
+      "step": 8775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.781329375838352e-06,
+      "loss": 0.1529,
+      "step": 8780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.779902394474728e-06,
+      "loss": 0.2231,
+      "step": 8785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.778475413111104e-06,
+      "loss": 0.1318,
+      "step": 8790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.777048431747482e-06,
+      "loss": 0.1566,
+      "step": 8795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.775621450383859e-06,
+      "loss": 0.3991,
+      "step": 8800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.774194469020235e-06,
+      "loss": 0.146,
+      "step": 8805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.772767487656611e-06,
+      "loss": 0.2099,
+      "step": 8810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.771340506292988e-06,
+      "loss": 0.2104,
+      "step": 8815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.769913524929366e-06,
+      "loss": 0.1561,
+      "step": 8820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.768486543565742e-06,
+      "loss": 0.1779,
+      "step": 8825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.767059562202118e-06,
+      "loss": 0.1237,
+      "step": 8830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.765632580838495e-06,
+      "loss": 0.212,
+      "step": 8835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.764205599474871e-06,
+      "loss": 0.2135,
+      "step": 8840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.762778618111249e-06,
+      "loss": 0.3727,
+      "step": 8845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.761351636747625e-06,
+      "loss": 0.9279,
+      "step": 8850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.759924655384002e-06,
+      "loss": 0.1871,
+      "step": 8855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.758497674020378e-06,
+      "loss": 0.2423,
+      "step": 8860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.757070692656754e-06,
+      "loss": 0.1663,
+      "step": 8865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.75564371129313e-06,
+      "loss": 0.252,
+      "step": 8870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.754216729929507e-06,
+      "loss": 0.0367,
+      "step": 8875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.752789748565885e-06,
+      "loss": 0.0966,
+      "step": 8880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.751362767202261e-06,
+      "loss": 0.186,
+      "step": 8885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.749935785838638e-06,
+      "loss": 0.3719,
+      "step": 8890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.748508804475014e-06,
+      "loss": 0.5401,
+      "step": 8895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.74708182311139e-06,
+      "loss": 0.5217,
+      "step": 8900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.745654841747767e-06,
+      "loss": 0.1836,
+      "step": 8905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.744227860384143e-06,
+      "loss": 0.199,
+      "step": 8910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.742800879020521e-06,
+      "loss": 0.2465,
+      "step": 8915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.741373897656898e-06,
+      "loss": 0.2377,
+      "step": 8920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.739946916293274e-06,
+      "loss": 0.2297,
+      "step": 8925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.73851993492965e-06,
+      "loss": 0.072,
+      "step": 8930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.737092953566027e-06,
+      "loss": 0.1414,
+      "step": 8935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.735665972202405e-06,
+      "loss": 0.1911,
+      "step": 8940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.734238990838781e-06,
+      "loss": 0.2331,
+      "step": 8945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.732812009475157e-06,
+      "loss": 0.5353,
+      "step": 8950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.731385028111534e-06,
+      "loss": 0.2056,
+      "step": 8955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.72995804674791e-06,
+      "loss": 0.1993,
+      "step": 8960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.728531065384286e-06,
+      "loss": 0.2409,
+      "step": 8965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.727104084020664e-06,
+      "loss": 0.2607,
+      "step": 8970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.72567710265704e-06,
+      "loss": 0.202,
+      "step": 8975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.724250121293417e-06,
+      "loss": 0.098,
+      "step": 8980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.722823139929793e-06,
+      "loss": 0.1181,
+      "step": 8985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.72139615856617e-06,
+      "loss": 0.1399,
+      "step": 8990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.719969177202546e-06,
+      "loss": 0.3393,
+      "step": 8995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.718542195838922e-06,
+      "loss": 0.2566,
+      "step": 9000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.717115214475299e-06,
+      "loss": 0.1742,
+      "step": 9005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.715688233111677e-06,
+      "loss": 0.1669,
+      "step": 9010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.714261251748053e-06,
+      "loss": 0.1532,
+      "step": 9015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.71283427038443e-06,
+      "loss": 0.1997,
+      "step": 9020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.711407289020806e-06,
+      "loss": 0.102,
+      "step": 9025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.709980307657182e-06,
+      "loss": 0.1047,
+      "step": 9030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.70855332629356e-06,
+      "loss": 0.1207,
+      "step": 9035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.707126344929936e-06,
+      "loss": 0.1498,
+      "step": 9040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.705699363566313e-06,
+      "loss": 0.3169,
+      "step": 9045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.704272382202689e-06,
+      "loss": 0.5967,
+      "step": 9050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.702845400839065e-06,
+      "loss": 0.1251,
+      "step": 9055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.701418419475443e-06,
+      "loss": 0.1762,
+      "step": 9060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.69999143811182e-06,
+      "loss": 0.2053,
+      "step": 9065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.698564456748196e-06,
+      "loss": 0.223,
+      "step": 9070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.697137475384573e-06,
+      "loss": 0.1064,
+      "step": 9075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.695710494020949e-06,
+      "loss": 0.1265,
+      "step": 9080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.694283512657325e-06,
+      "loss": 0.0834,
+      "step": 9085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.692856531293702e-06,
+      "loss": 0.3055,
+      "step": 9090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.691429549930078e-06,
+      "loss": 0.3791,
+      "step": 9095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.690002568566454e-06,
+      "loss": 0.7226,
+      "step": 9100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.688575587202832e-06,
+      "loss": 0.2248,
+      "step": 9105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.687148605839209e-06,
+      "loss": 0.167,
+      "step": 9110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.685721624475585e-06,
+      "loss": 0.157,
+      "step": 9115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.684294643111961e-06,
+      "loss": 0.1776,
+      "step": 9120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.682867661748338e-06,
+      "loss": 0.1557,
+      "step": 9125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.681440680384716e-06,
+      "loss": 0.1552,
+      "step": 9130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.680013699021092e-06,
+      "loss": 0.1635,
+      "step": 9135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.678586717657468e-06,
+      "loss": 0.2556,
+      "step": 9140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.677159736293845e-06,
+      "loss": 0.3,
+      "step": 9145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.675732754930221e-06,
+      "loss": 0.4252,
+      "step": 9150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.674305773566599e-06,
+      "loss": 0.1934,
+      "step": 9155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.672878792202975e-06,
+      "loss": 0.1446,
+      "step": 9160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.671451810839352e-06,
+      "loss": 0.1904,
+      "step": 9165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.670024829475728e-06,
+      "loss": 0.2697,
+      "step": 9170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.668597848112104e-06,
+      "loss": 0.1665,
+      "step": 9175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.66717086674848e-06,
+      "loss": 0.0452,
+      "step": 9180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.665743885384857e-06,
+      "loss": 0.213,
+      "step": 9185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.664316904021233e-06,
+      "loss": 0.2696,
+      "step": 9190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.66288992265761e-06,
+      "loss": 0.254,
+      "step": 9195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.661462941293988e-06,
+      "loss": 0.34,
+      "step": 9200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.660035959930364e-06,
+      "loss": 0.2095,
+      "step": 9205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.65860897856674e-06,
+      "loss": 0.1563,
+      "step": 9210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.657181997203117e-06,
+      "loss": 0.1154,
+      "step": 9215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.655755015839493e-06,
+      "loss": 0.1982,
+      "step": 9220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.654328034475871e-06,
+      "loss": 0.1001,
+      "step": 9225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.652901053112247e-06,
+      "loss": 0.1257,
+      "step": 9230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.651474071748624e-06,
+      "loss": 0.1364,
+      "step": 9235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.650047090385e-06,
+      "loss": 0.1823,
+      "step": 9240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.648620109021376e-06,
+      "loss": 0.3176,
+      "step": 9245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.647193127657755e-06,
+      "loss": 0.411,
+      "step": 9250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.645766146294131e-06,
+      "loss": 0.168,
+      "step": 9255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.644339164930507e-06,
+      "loss": 0.2204,
+      "step": 9260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.642912183566884e-06,
+      "loss": 0.2458,
+      "step": 9265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.64148520220326e-06,
+      "loss": 0.1735,
+      "step": 9270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.640058220839636e-06,
+      "loss": 0.0257,
+      "step": 9275
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.638631239476013e-06,
+      "loss": 0.2665,
+      "step": 9280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.637204258112389e-06,
+      "loss": 0.1184,
+      "step": 9285
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.635777276748765e-06,
+      "loss": 0.1373,
+      "step": 9290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.634350295385143e-06,
+      "loss": 0.1819,
+      "step": 9295
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.63292331402152e-06,
+      "loss": 0.226,
+      "step": 9300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.631496332657896e-06,
+      "loss": 0.1845,
+      "step": 9305
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.630069351294272e-06,
+      "loss": 0.1571,
+      "step": 9310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.628642369930649e-06,
+      "loss": 0.1804,
+      "step": 9315
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.627215388567027e-06,
+      "loss": 0.2218,
+      "step": 9320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.625788407203403e-06,
+      "loss": 0.1621,
+      "step": 9325
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.624361425839779e-06,
+      "loss": 0.1616,
+      "step": 9330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.622934444476156e-06,
+      "loss": 0.1291,
+      "step": 9335
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.621507463112532e-06,
+      "loss": 0.0476,
+      "step": 9340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.620080481748909e-06,
+      "loss": 0.1954,
+      "step": 9345
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.6186535003852855e-06,
+      "loss": 0.4141,
+      "step": 9350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.617226519021662e-06,
+      "loss": 0.1552,
+      "step": 9355
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.615799537658038e-06,
+      "loss": 0.1397,
+      "step": 9360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.614372556294416e-06,
+      "loss": 0.1373,
+      "step": 9365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.6129455749307926e-06,
+      "loss": 0.1499,
+      "step": 9370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.611518593567169e-06,
+      "loss": 0.1765,
+      "step": 9375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.610091612203545e-06,
+      "loss": 0.0806,
+      "step": 9380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.608664630839922e-06,
+      "loss": 0.0987,
+      "step": 9385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.607237649476299e-06,
+      "loss": 0.2782,
+      "step": 9390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.605810668112675e-06,
+      "loss": 0.4899,
+      "step": 9395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.6043836867490515e-06,
+      "loss": 0.4337,
+      "step": 9400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.602956705385428e-06,
+      "loss": 0.192,
+      "step": 9405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.601529724021804e-06,
+      "loss": 0.1428,
+      "step": 9410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.600102742658182e-06,
+      "loss": 0.2142,
+      "step": 9415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.5986757612945585e-06,
+      "loss": 0.2392,
+      "step": 9420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.597248779930935e-06,
+      "loss": 0.1141,
+      "step": 9425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.595821798567311e-06,
+      "loss": 0.0959,
+      "step": 9430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.5943948172036875e-06,
+      "loss": 0.3744,
+      "step": 9435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.592967835840065e-06,
+      "loss": 0.2595,
+      "step": 9440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.591540854476441e-06,
+      "loss": 0.1794,
+      "step": 9445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.590113873112817e-06,
+      "loss": 0.4224,
+      "step": 9450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.588686891749194e-06,
+      "loss": 0.2906,
+      "step": 9455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.587259910385572e-06,
+      "loss": 0.1409,
+      "step": 9460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.585832929021948e-06,
+      "loss": 0.1964,
+      "step": 9465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.5844059476583244e-06,
+      "loss": 0.2423,
+      "step": 9470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.582978966294701e-06,
+      "loss": 0.0667,
+      "step": 9475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.581551984931077e-06,
+      "loss": 0.1139,
+      "step": 9480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.580125003567454e-06,
+      "loss": 0.2279,
+      "step": 9485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.578698022203831e-06,
+      "loss": 0.3152,
+      "step": 9490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.577271040840207e-06,
+      "loss": 0.265,
+      "step": 9495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.575844059476583e-06,
+      "loss": 0.482,
+      "step": 9500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.57441707811296e-06,
+      "loss": 0.159,
+      "step": 9505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.572990096749338e-06,
+      "loss": 0.1347,
+      "step": 9510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.571563115385714e-06,
+      "loss": 0.1418,
+      "step": 9515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.57013613402209e-06,
+      "loss": 0.1773,
+      "step": 9520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.568709152658467e-06,
+      "loss": 0.0984,
+      "step": 9525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.567282171294843e-06,
+      "loss": 0.0493,
+      "step": 9530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.56585518993122e-06,
+      "loss": 0.3436,
+      "step": 9535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.5644282085675966e-06,
+      "loss": 0.2792,
+      "step": 9540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.563001227203973e-06,
+      "loss": 0.2537,
+      "step": 9545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.561574245840349e-06,
+      "loss": 0.2774,
+      "step": 9550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.560147264476727e-06,
+      "loss": 0.0976,
+      "step": 9555
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.558720283113104e-06,
+      "loss": 0.2648,
+      "step": 9560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.55729330174948e-06,
+      "loss": 0.2325,
+      "step": 9565
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.555866320385856e-06,
+      "loss": 0.2104,
+      "step": 9570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.554439339022233e-06,
+      "loss": 0.1456,
+      "step": 9575
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.55301235765861e-06,
+      "loss": 0.1227,
+      "step": 9580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.551585376294986e-06,
+      "loss": 0.1161,
+      "step": 9585
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.5501583949313625e-06,
+      "loss": 0.2457,
+      "step": 9590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.548731413567739e-06,
+      "loss": 0.3123,
+      "step": 9595
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.547304432204115e-06,
+      "loss": 0.4145,
+      "step": 9600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.545877450840493e-06,
+      "loss": 0.1799,
+      "step": 9605
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.5444504694768695e-06,
+      "loss": 0.1297,
+      "step": 9610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.543023488113246e-06,
+      "loss": 0.1053,
+      "step": 9615
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.541596506749622e-06,
+      "loss": 0.2045,
+      "step": 9620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.5401695253859985e-06,
+      "loss": 0.1469,
+      "step": 9625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.538742544022376e-06,
+      "loss": 0.1846,
+      "step": 9630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.537315562658752e-06,
+      "loss": 0.1,
+      "step": 9635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.535888581295128e-06,
+      "loss": 0.1658,
+      "step": 9640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.534461599931505e-06,
+      "loss": 0.1718,
+      "step": 9645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.533034618567883e-06,
+      "loss": 0.263,
+      "step": 9650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.531607637204259e-06,
+      "loss": 0.1255,
+      "step": 9655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.5301806558406355e-06,
+      "loss": 0.2833,
+      "step": 9660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.528753674477012e-06,
+      "loss": 0.1569,
+      "step": 9665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.527326693113388e-06,
+      "loss": 0.153,
+      "step": 9670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.525899711749765e-06,
+      "loss": 0.237,
+      "step": 9675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.524472730386142e-06,
+      "loss": 0.0772,
+      "step": 9680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.523045749022518e-06,
+      "loss": 0.1243,
+      "step": 9685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.521618767658894e-06,
+      "loss": 0.1303,
+      "step": 9690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.5201917862952715e-06,
+      "loss": 0.2455,
+      "step": 9695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.518764804931649e-06,
+      "loss": 0.4695,
+      "step": 9700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.517337823568025e-06,
+      "loss": 0.1337,
+      "step": 9705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.515910842204401e-06,
+      "loss": 0.1306,
+      "step": 9710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.514483860840778e-06,
+      "loss": 0.1741,
+      "step": 9715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.513056879477154e-06,
+      "loss": 0.203,
+      "step": 9720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.511629898113531e-06,
+      "loss": 0.1038,
+      "step": 9725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.510202916749908e-06,
+      "loss": 0.1263,
+      "step": 9730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.508775935386285e-06,
+      "loss": 0.1612,
+      "step": 9735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.507348954022661e-06,
+      "loss": 0.2431,
+      "step": 9740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.505921972659038e-06,
+      "loss": 0.1577,
+      "step": 9745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.504494991295415e-06,
+      "loss": 0.3567,
+      "step": 9750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.503068009931791e-06,
+      "loss": 0.2375,
+      "step": 9755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.501641028568167e-06,
+      "loss": 0.1828,
+      "step": 9760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.500214047204544e-06,
+      "loss": 0.1248,
+      "step": 9765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.498787065840921e-06,
+      "loss": 0.2088,
+      "step": 9770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.497360084477297e-06,
+      "loss": 0.1981,
+      "step": 9775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.495933103113674e-06,
+      "loss": 0.139,
+      "step": 9780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.494506121750051e-06,
+      "loss": 0.1037,
+      "step": 9785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.493079140386427e-06,
+      "loss": 0.1607,
+      "step": 9790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.491652159022804e-06,
+      "loss": 0.3597,
+      "step": 9795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.4902251776591806e-06,
+      "loss": 0.2919,
+      "step": 9800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.488798196295557e-06,
+      "loss": 0.1854,
+      "step": 9805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.487371214931933e-06,
+      "loss": 0.1637,
+      "step": 9810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.48594423356831e-06,
+      "loss": 0.2263,
+      "step": 9815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.484517252204687e-06,
+      "loss": 0.1544,
+      "step": 9820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.483090270841064e-06,
+      "loss": 0.2028,
+      "step": 9825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.48166328947744e-06,
+      "loss": 0.12,
+      "step": 9830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.480236308113817e-06,
+      "loss": 0.1801,
+      "step": 9835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.478809326750194e-06,
+      "loss": 0.2422,
+      "step": 9840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.47738234538657e-06,
+      "loss": 0.1783,
+      "step": 9845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.4759553640229465e-06,
+      "loss": 0.6002,
+      "step": 9850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.474528382659323e-06,
+      "loss": 0.176,
+      "step": 9855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.473101401295699e-06,
+      "loss": 0.1873,
+      "step": 9860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.471674419932076e-06,
+      "loss": 0.2043,
+      "step": 9865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.4702474385684535e-06,
+      "loss": 0.1526,
+      "step": 9870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.46882045720483e-06,
+      "loss": 0.202,
+      "step": 9875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.467393475841206e-06,
+      "loss": 0.155,
+      "step": 9880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.4659664944775826e-06,
+      "loss": 0.1549,
+      "step": 9885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.46453951311396e-06,
+      "loss": 0.2724,
+      "step": 9890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.463112531750336e-06,
+      "loss": 0.2492,
+      "step": 9895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.461685550386712e-06,
+      "loss": 0.397,
+      "step": 9900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.460258569023089e-06,
+      "loss": 0.2147,
+      "step": 9905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.458831587659465e-06,
+      "loss": 0.1992,
+      "step": 9910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.457404606295843e-06,
+      "loss": 0.1494,
+      "step": 9915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.4559776249322195e-06,
+      "loss": 0.2049,
+      "step": 9920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.454550643568596e-06,
+      "loss": 0.1952,
+      "step": 9925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.453123662204972e-06,
+      "loss": 0.0743,
+      "step": 9930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.451696680841349e-06,
+      "loss": 0.1328,
+      "step": 9935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.450269699477726e-06,
+      "loss": 0.2897,
+      "step": 9940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.448842718114102e-06,
+      "loss": 0.2717,
+      "step": 9945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.447415736750478e-06,
+      "loss": 0.3851,
+      "step": 9950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.445988755386855e-06,
+      "loss": 0.2401,
+      "step": 9955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.444561774023233e-06,
+      "loss": 0.2105,
+      "step": 9960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.443134792659609e-06,
+      "loss": 0.1656,
+      "step": 9965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.441707811295985e-06,
+      "loss": 0.2268,
+      "step": 9970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.440280829932362e-06,
+      "loss": 0.1913,
+      "step": 9975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.438853848568738e-06,
+      "loss": 0.0646,
+      "step": 9980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.437426867205115e-06,
+      "loss": 0.143,
+      "step": 9985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.435999885841492e-06,
+      "loss": 0.1793,
+      "step": 9990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.434572904477868e-06,
+      "loss": 0.1764,
+      "step": 9995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.433145923114244e-06,
+      "loss": 0.2205,
+      "step": 10000
+    },
+    {
+      "epoch": 0.28,
+      "eval_loss": 0.07491611689329147,
+      "eval_runtime": 2332.1386,
+      "eval_samples_per_second": 6.868,
+      "eval_steps_per_second": 1.717,
+      "eval_wer": 0.2068698160836639,
+      "step": 10000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.431718941750621e-06,
+      "loss": 0.2567,
+      "step": 10005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.430291960386999e-06,
+      "loss": 0.1234,
+      "step": 10010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.428864979023375e-06,
+      "loss": 0.1736,
+      "step": 10015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.427437997659751e-06,
+      "loss": 0.2017,
+      "step": 10020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.426011016296128e-06,
+      "loss": 0.1199,
+      "step": 10025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.424584034932505e-06,
+      "loss": 0.0675,
+      "step": 10030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.423157053568881e-06,
+      "loss": 0.1714,
+      "step": 10035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.4217300722052575e-06,
+      "loss": 0.1121,
+      "step": 10040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.420303090841634e-06,
+      "loss": 0.2543,
+      "step": 10045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.41887610947801e-06,
+      "loss": 0.7649,
+      "step": 10050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.417449128114388e-06,
+      "loss": 0.1539,
+      "step": 10055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.4160221467507646e-06,
+      "loss": 0.1907,
+      "step": 10060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.414595165387141e-06,
+      "loss": 0.1908,
+      "step": 10065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.413168184023517e-06,
+      "loss": 0.2002,
+      "step": 10070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.411741202659894e-06,
+      "loss": 0.1605,
+      "step": 10075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.410314221296271e-06,
+      "loss": 0.1371,
+      "step": 10080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.408887239932647e-06,
+      "loss": 0.227,
+      "step": 10085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.4074602585690235e-06,
+      "loss": 0.3145,
+      "step": 10090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.4060332772054e-06,
+      "loss": 0.3825,
+      "step": 10095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.404606295841776e-06,
+      "loss": 0.4545,
+      "step": 10100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.403179314478154e-06,
+      "loss": 0.1605,
+      "step": 10105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.4017523331145305e-06,
+      "loss": 0.1218,
+      "step": 10110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.400325351750907e-06,
+      "loss": 0.1969,
+      "step": 10115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.398898370387283e-06,
+      "loss": 0.159,
+      "step": 10120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.39747138902366e-06,
+      "loss": 0.2698,
+      "step": 10125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.396044407660037e-06,
+      "loss": 0.1793,
+      "step": 10130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.394617426296413e-06,
+      "loss": 0.1331,
+      "step": 10135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.393190444932789e-06,
+      "loss": 0.2675,
+      "step": 10140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.391763463569166e-06,
+      "loss": 0.3275,
+      "step": 10145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.390336482205544e-06,
+      "loss": 0.3463,
+      "step": 10150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.38890950084192e-06,
+      "loss": 0.1608,
+      "step": 10155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.3874825194782964e-06,
+      "loss": 0.1301,
+      "step": 10160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.386055538114673e-06,
+      "loss": 0.1259,
+      "step": 10165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.384628556751049e-06,
+      "loss": 0.2146,
+      "step": 10170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.383201575387426e-06,
+      "loss": 0.1014,
+      "step": 10175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.381774594023803e-06,
+      "loss": 0.1242,
+      "step": 10180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.380347612660179e-06,
+      "loss": 0.0413,
+      "step": 10185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.378920631296555e-06,
+      "loss": 0.2255,
+      "step": 10190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.377493649932932e-06,
+      "loss": 0.291,
+      "step": 10195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.37606666856931e-06,
+      "loss": 0.7039,
+      "step": 10200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.374639687205686e-06,
+      "loss": 0.1755,
+      "step": 10205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.373212705842062e-06,
+      "loss": 0.248,
+      "step": 10210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.371785724478439e-06,
+      "loss": 0.1566,
+      "step": 10215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.370358743114816e-06,
+      "loss": 0.1532,
+      "step": 10220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.368931761751192e-06,
+      "loss": 0.0946,
+      "step": 10225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.3675047803875686e-06,
+      "loss": 0.1811,
+      "step": 10230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.366077799023945e-06,
+      "loss": 0.2768,
+      "step": 10235
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.364650817660321e-06,
+      "loss": 0.2201,
+      "step": 10240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.363223836296699e-06,
+      "loss": 0.238,
+      "step": 10245
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.361796854933076e-06,
+      "loss": 0.4387,
+      "step": 10250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.360369873569452e-06,
+      "loss": 0.1397,
+      "step": 10255
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.358942892205828e-06,
+      "loss": 0.1958,
+      "step": 10260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.357515910842205e-06,
+      "loss": 0.1626,
+      "step": 10265
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.356088929478582e-06,
+      "loss": 0.2076,
+      "step": 10270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.354661948114958e-06,
+      "loss": 0.2224,
+      "step": 10275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.3532349667513345e-06,
+      "loss": 0.0584,
+      "step": 10280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.351807985387711e-06,
+      "loss": 0.1163,
+      "step": 10285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.350381004024087e-06,
+      "loss": 0.3164,
+      "step": 10290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.348954022660465e-06,
+      "loss": 0.2504,
+      "step": 10295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.3475270412968415e-06,
+      "loss": 0.3822,
+      "step": 10300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.346100059933218e-06,
+      "loss": 0.1097,
+      "step": 10305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.344673078569594e-06,
+      "loss": 0.2248,
+      "step": 10310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.343246097205971e-06,
+      "loss": 0.1898,
+      "step": 10315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.341819115842348e-06,
+      "loss": 0.2194,
+      "step": 10320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.340392134478724e-06,
+      "loss": 0.0948,
+      "step": 10325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.3389651531151e-06,
+      "loss": 0.16,
+      "step": 10330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.337538171751477e-06,
+      "loss": 0.1801,
+      "step": 10335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.336111190387855e-06,
+      "loss": 0.2755,
+      "step": 10340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.334684209024231e-06,
+      "loss": 0.3869,
+      "step": 10345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.3332572276606075e-06,
+      "loss": 0.3723,
+      "step": 10350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.331830246296984e-06,
+      "loss": 0.1728,
+      "step": 10355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.33040326493336e-06,
+      "loss": 0.1652,
+      "step": 10360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.328976283569737e-06,
+      "loss": 0.2152,
+      "step": 10365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.327549302206114e-06,
+      "loss": 0.2127,
+      "step": 10370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.32612232084249e-06,
+      "loss": 0.0725,
+      "step": 10375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.324695339478866e-06,
+      "loss": 0.1122,
+      "step": 10380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.323268358115243e-06,
+      "loss": 0.098,
+      "step": 10385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.321841376751621e-06,
+      "loss": 0.1685,
+      "step": 10390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.320414395387997e-06,
+      "loss": 0.2073,
+      "step": 10395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.318987414024373e-06,
+      "loss": 0.3399,
+      "step": 10400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.31756043266075e-06,
+      "loss": 0.1522,
+      "step": 10405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.316133451297127e-06,
+      "loss": 0.2102,
+      "step": 10410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.314706469933503e-06,
+      "loss": 0.2329,
+      "step": 10415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.31327948856988e-06,
+      "loss": 0.227,
+      "step": 10420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.312137903478981e-06,
+      "loss": 1.2355,
+      "step": 10425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.310710922115358e-06,
+      "loss": 0.1112,
+      "step": 10430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.309283940751734e-06,
+      "loss": 0.1159,
+      "step": 10435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.307856959388111e-06,
+      "loss": 0.1421,
+      "step": 10440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.306429978024487e-06,
+      "loss": 0.36,
+      "step": 10445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.305002996660863e-06,
+      "loss": 0.415,
+      "step": 10450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.303576015297241e-06,
+      "loss": 0.3397,
+      "step": 10455
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.302149033933618e-06,
+      "loss": 0.1816,
+      "step": 10460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.300722052569994e-06,
+      "loss": 0.1479,
+      "step": 10465
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.29929507120637e-06,
+      "loss": 0.176,
+      "step": 10470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.297868089842747e-06,
+      "loss": 0.1749,
+      "step": 10475
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.296441108479124e-06,
+      "loss": 0.0827,
+      "step": 10480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.2950141271155e-06,
+      "loss": 0.1076,
+      "step": 10485
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.293587145751877e-06,
+      "loss": 0.2105,
+      "step": 10490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.292160164388253e-06,
+      "loss": 0.3473,
+      "step": 10495
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.290733183024631e-06,
+      "loss": 0.6358,
+      "step": 10500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.289306201661007e-06,
+      "loss": 0.1205,
+      "step": 10505
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.287879220297384e-06,
+      "loss": 0.1851,
+      "step": 10510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.28645223893376e-06,
+      "loss": 0.1954,
+      "step": 10515
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.285025257570136e-06,
+      "loss": 0.2525,
+      "step": 10520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.2835982762065135e-06,
+      "loss": 0.2418,
+      "step": 10525
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.28217129484289e-06,
+      "loss": 0.0711,
+      "step": 10530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.280744313479266e-06,
+      "loss": 0.1908,
+      "step": 10535
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.2793173321156425e-06,
+      "loss": 0.2853,
+      "step": 10540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.27789035075202e-06,
+      "loss": 0.3841,
+      "step": 10545
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.276463369388397e-06,
+      "loss": 0.6,
+      "step": 10550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.275036388024773e-06,
+      "loss": 0.11,
+      "step": 10555
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.27360940666115e-06,
+      "loss": 0.1714,
+      "step": 10560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.272182425297526e-06,
+      "loss": 0.0995,
+      "step": 10565
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.270755443933902e-06,
+      "loss": 0.1887,
+      "step": 10570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.2693284625702795e-06,
+      "loss": 0.1676,
+      "step": 10575
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.267901481206656e-06,
+      "loss": 0.1543,
+      "step": 10580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.266474499843032e-06,
+      "loss": 0.0817,
+      "step": 10585
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.265047518479409e-06,
+      "loss": 0.1569,
+      "step": 10590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.2636205371157865e-06,
+      "loss": 0.2467,
+      "step": 10595
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.262193555752163e-06,
+      "loss": 0.3902,
+      "step": 10600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.260766574388539e-06,
+      "loss": 0.1886,
+      "step": 10605
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.2593395930249155e-06,
+      "loss": 0.2343,
+      "step": 10610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.257912611661292e-06,
+      "loss": 0.1152,
+      "step": 10615
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.256485630297669e-06,
+      "loss": 0.2069,
+      "step": 10620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.255058648934045e-06,
+      "loss": 0.1517,
+      "step": 10625
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.253631667570422e-06,
+      "loss": 0.1141,
+      "step": 10630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.252204686206799e-06,
+      "loss": 0.1,
+      "step": 10635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.250777704843175e-06,
+      "loss": 0.2799,
+      "step": 10640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2493507234795524e-06,
+      "loss": 0.3713,
+      "step": 10645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.247923742115929e-06,
+      "loss": 0.2662,
+      "step": 10650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.246496760752305e-06,
+      "loss": 0.1017,
+      "step": 10655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2450697793886815e-06,
+      "loss": 0.1982,
+      "step": 10660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.243642798025058e-06,
+      "loss": 0.183,
+      "step": 10665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.242215816661435e-06,
+      "loss": 0.1771,
+      "step": 10670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.240788835297811e-06,
+      "loss": 0.155,
+      "step": 10675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2393618539341885e-06,
+      "loss": 0.077,
+      "step": 10680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.237934872570565e-06,
+      "loss": 0.0747,
+      "step": 10685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.236507891206942e-06,
+      "loss": 0.4591,
+      "step": 10690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.235080909843318e-06,
+      "loss": 0.3071,
+      "step": 10695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.233653928479695e-06,
+      "loss": 0.4575,
+      "step": 10700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.232226947116071e-06,
+      "loss": 0.5442,
+      "step": 10705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.230799965752447e-06,
+      "loss": 0.2193,
+      "step": 10710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2293729843888246e-06,
+      "loss": 0.225,
+      "step": 10715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.227946003025202e-06,
+      "loss": 0.3045,
+      "step": 10720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.226519021661578e-06,
+      "loss": 0.1232,
+      "step": 10725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2250920402979544e-06,
+      "loss": 0.1332,
+      "step": 10730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.223665058934331e-06,
+      "loss": 0.0676,
+      "step": 10735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.222238077570708e-06,
+      "loss": 0.2668,
+      "step": 10740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.220811096207084e-06,
+      "loss": 0.3183,
+      "step": 10745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.219384114843461e-06,
+      "loss": 0.3311,
+      "step": 10750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.217957133479837e-06,
+      "loss": 0.1576,
+      "step": 10755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.216530152116213e-06,
+      "loss": 0.1797,
+      "step": 10760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.215103170752591e-06,
+      "loss": 0.1971,
+      "step": 10765
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.213676189388968e-06,
+      "loss": 0.231,
+      "step": 10770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.212249208025344e-06,
+      "loss": 0.1203,
+      "step": 10775
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.21082222666172e-06,
+      "loss": 0.1624,
+      "step": 10780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2093952452980975e-06,
+      "loss": 0.1252,
+      "step": 10785
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.207968263934474e-06,
+      "loss": 0.1593,
+      "step": 10790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.20654128257085e-06,
+      "loss": 0.1432,
+      "step": 10795
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.2051143012072266e-06,
+      "loss": 0.4471,
+      "step": 10800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.203687319843603e-06,
+      "loss": 0.154,
+      "step": 10805
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.202260338479981e-06,
+      "loss": 0.2305,
+      "step": 10810
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.200833357116357e-06,
+      "loss": 0.2097,
+      "step": 10815
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.199406375752734e-06,
+      "loss": 0.13,
+      "step": 10820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.19797939438911e-06,
+      "loss": 0.1908,
+      "step": 10825
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.196552413025486e-06,
+      "loss": 0.0708,
+      "step": 10830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.1951254316618635e-06,
+      "loss": 0.172,
+      "step": 10835
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.19369845029824e-06,
+      "loss": 0.2536,
+      "step": 10840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.192271468934616e-06,
+      "loss": 0.1223,
+      "step": 10845
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.1908444875709925e-06,
+      "loss": 0.533,
+      "step": 10850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.189417506207369e-06,
+      "loss": 0.1416,
+      "step": 10855
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.187990524843747e-06,
+      "loss": 0.1807,
+      "step": 10860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.186563543480123e-06,
+      "loss": 0.2,
+      "step": 10865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.1851365621164995e-06,
+      "loss": 0.1764,
+      "step": 10870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.183709580752876e-06,
+      "loss": 0.1812,
+      "step": 10875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.182282599389253e-06,
+      "loss": 0.0932,
+      "step": 10880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.180855618025629e-06,
+      "loss": 0.2631,
+      "step": 10885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.179428636662006e-06,
+      "loss": 0.0859,
+      "step": 10890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.178001655298382e-06,
+      "loss": 0.1994,
+      "step": 10895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.176574673934758e-06,
+      "loss": 0.3887,
+      "step": 10900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.1751476925711364e-06,
+      "loss": 0.1399,
+      "step": 10905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.173720711207513e-06,
+      "loss": 0.2339,
+      "step": 10910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.172293729843889e-06,
+      "loss": 0.2061,
+      "step": 10915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.1708667484802655e-06,
+      "loss": 0.1486,
+      "step": 10920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.169439767116642e-06,
+      "loss": 0.1816,
+      "step": 10925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.168012785753019e-06,
+      "loss": 0.13,
+      "step": 10930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.166585804389395e-06,
+      "loss": 0.155,
+      "step": 10935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.165158823025772e-06,
+      "loss": 0.1147,
+      "step": 10940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.163731841662148e-06,
+      "loss": 0.2045,
+      "step": 10945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.162304860298524e-06,
+      "loss": 0.2344,
+      "step": 10950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.160877878934902e-06,
+      "loss": 0.301,
+      "step": 10955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.159450897571279e-06,
+      "loss": 0.213,
+      "step": 10960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.158023916207655e-06,
+      "loss": 0.1779,
+      "step": 10965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.156596934844031e-06,
+      "loss": 0.2003,
+      "step": 10970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.1551699534804086e-06,
+      "loss": 0.1772,
+      "step": 10975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.153742972116785e-06,
+      "loss": 0.0908,
+      "step": 10980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.152315990753161e-06,
+      "loss": 0.1826,
+      "step": 10985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.150889009389538e-06,
+      "loss": 0.2637,
+      "step": 10990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.149462028025914e-06,
+      "loss": 0.4362,
+      "step": 10995
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.148035046662292e-06,
+      "loss": 0.5842,
+      "step": 11000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.146608065298668e-06,
+      "loss": 0.1434,
+      "step": 11005
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.145181083935045e-06,
+      "loss": 0.106,
+      "step": 11010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.143754102571421e-06,
+      "loss": 0.1779,
+      "step": 11015
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.142327121207797e-06,
+      "loss": 0.219,
+      "step": 11020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1409001398441745e-06,
+      "loss": 0.1088,
+      "step": 11025
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.139473158480551e-06,
+      "loss": 0.0586,
+      "step": 11030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.138046177116927e-06,
+      "loss": 0.2313,
+      "step": 11035
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1366191957533035e-06,
+      "loss": 0.2773,
+      "step": 11040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.135477610662406e-06,
+      "loss": 1.0288,
+      "step": 11045
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.134050629298782e-06,
+      "loss": 0.2824,
+      "step": 11050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.132623647935158e-06,
+      "loss": 0.1649,
+      "step": 11055
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.131196666571535e-06,
+      "loss": 0.1973,
+      "step": 11060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.129769685207911e-06,
+      "loss": 0.1333,
+      "step": 11065
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.128342703844289e-06,
+      "loss": 0.16,
+      "step": 11070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.126915722480665e-06,
+      "loss": 0.1792,
+      "step": 11075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.125488741117042e-06,
+      "loss": 0.1677,
+      "step": 11080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.124061759753418e-06,
+      "loss": 0.0933,
+      "step": 11085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.122634778389795e-06,
+      "loss": 0.0849,
+      "step": 11090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1212077970261715e-06,
+      "loss": 0.2138,
+      "step": 11095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.119780815662548e-06,
+      "loss": 0.4769,
+      "step": 11100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.118353834298924e-06,
+      "loss": 0.1347,
+      "step": 11105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1169268529353005e-06,
+      "loss": 0.1429,
+      "step": 11110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.115499871571679e-06,
+      "loss": 0.1614,
+      "step": 11115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.114072890208055e-06,
+      "loss": 0.1527,
+      "step": 11120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.112645908844431e-06,
+      "loss": 0.1226,
+      "step": 11125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.111218927480808e-06,
+      "loss": 0.1522,
+      "step": 11130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.109791946117184e-06,
+      "loss": 0.2053,
+      "step": 11135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.108364964753561e-06,
+      "loss": 0.2242,
+      "step": 11140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1069379833899375e-06,
+      "loss": 0.2609,
+      "step": 11145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.105511002026314e-06,
+      "loss": 0.2835,
+      "step": 11150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.10408402066269e-06,
+      "loss": 0.1587,
+      "step": 11155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1026570392990665e-06,
+      "loss": 0.1726,
+      "step": 11160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.1012300579354445e-06,
+      "loss": 0.1555,
+      "step": 11165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.099803076571821e-06,
+      "loss": 0.2174,
+      "step": 11170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.098376095208197e-06,
+      "loss": 0.2005,
+      "step": 11175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0969491138445735e-06,
+      "loss": 0.0682,
+      "step": 11180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.095522132480951e-06,
+      "loss": 0.0635,
+      "step": 11185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.094095151117327e-06,
+      "loss": 0.1272,
+      "step": 11190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.092668169753703e-06,
+      "loss": 0.1967,
+      "step": 11195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.09124118839008e-06,
+      "loss": 0.3992,
+      "step": 11200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.089814207026456e-06,
+      "loss": 0.1422,
+      "step": 11205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.088387225662834e-06,
+      "loss": 0.1684,
+      "step": 11210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0869602442992104e-06,
+      "loss": 0.1779,
+      "step": 11215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.085533262935587e-06,
+      "loss": 0.1859,
+      "step": 11220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.084106281571963e-06,
+      "loss": 0.1575,
+      "step": 11225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0826793002083395e-06,
+      "loss": 0.0327,
+      "step": 11230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.081252318844717e-06,
+      "loss": 0.1897,
+      "step": 11235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.079825337481093e-06,
+      "loss": 0.2325,
+      "step": 11240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.078398356117469e-06,
+      "loss": 0.2526,
+      "step": 11245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.076971374753846e-06,
+      "loss": 0.3376,
+      "step": 11250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.075544393390222e-06,
+      "loss": 0.1749,
+      "step": 11255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0741174120266e-06,
+      "loss": 0.1364,
+      "step": 11260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.072690430662976e-06,
+      "loss": 0.1842,
+      "step": 11265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.071263449299353e-06,
+      "loss": 0.2257,
+      "step": 11270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.069836467935729e-06,
+      "loss": 0.2042,
+      "step": 11275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.068409486572106e-06,
+      "loss": 0.196,
+      "step": 11280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0669825052084826e-06,
+      "loss": 0.2094,
+      "step": 11285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.065555523844859e-06,
+      "loss": 0.1393,
+      "step": 11290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.064128542481235e-06,
+      "loss": 0.284,
+      "step": 11295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.062701561117612e-06,
+      "loss": 0.5269,
+      "step": 11300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.06127457975399e-06,
+      "loss": 0.1672,
+      "step": 11305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.059847598390366e-06,
+      "loss": 0.1597,
+      "step": 11310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.058420617026742e-06,
+      "loss": 0.2241,
+      "step": 11315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.056993635663119e-06,
+      "loss": 0.1999,
+      "step": 11320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.055566654299495e-06,
+      "loss": 0.1571,
+      "step": 11325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.054139672935872e-06,
+      "loss": 0.2032,
+      "step": 11330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0527126915722485e-06,
+      "loss": 0.1385,
+      "step": 11335
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.051285710208625e-06,
+      "loss": 0.2935,
+      "step": 11340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.049858728845001e-06,
+      "loss": 0.3487,
+      "step": 11345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0484317474813775e-06,
+      "loss": 0.6493,
+      "step": 11350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.0470047661177555e-06,
+      "loss": 0.1724,
+      "step": 11355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.045577784754132e-06,
+      "loss": 0.2007,
+      "step": 11360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.044150803390508e-06,
+      "loss": 0.1461,
+      "step": 11365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.0427238220268846e-06,
+      "loss": 0.2509,
+      "step": 11370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.041296840663262e-06,
+      "loss": 0.1743,
+      "step": 11375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.039869859299638e-06,
+      "loss": 0.1169,
+      "step": 11380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.038442877936014e-06,
+      "loss": 0.1445,
+      "step": 11385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.037015896572391e-06,
+      "loss": 0.232,
+      "step": 11390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.035588915208768e-06,
+      "loss": 0.3013,
+      "step": 11395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.034161933845145e-06,
+      "loss": 0.5159,
+      "step": 11400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.0327349524815215e-06,
+      "loss": 0.1103,
+      "step": 11405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.031307971117898e-06,
+      "loss": 0.18,
+      "step": 11410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.029880989754274e-06,
+      "loss": 0.1824,
+      "step": 11415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.0284540083906505e-06,
+      "loss": 0.1726,
+      "step": 11420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.027027027027028e-06,
+      "loss": 0.1194,
+      "step": 11425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.025600045663404e-06,
+      "loss": 0.1098,
+      "step": 11430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.02417306429978e-06,
+      "loss": 0.2506,
+      "step": 11435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.0227460829361575e-06,
+      "loss": 0.1701,
+      "step": 11440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.021319101572534e-06,
+      "loss": 0.2646,
+      "step": 11445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.019892120208911e-06,
+      "loss": 0.4503,
+      "step": 11450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.018465138845287e-06,
+      "loss": 0.1358,
+      "step": 11455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.017038157481664e-06,
+      "loss": 0.1648,
+      "step": 11460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.01561117611804e-06,
+      "loss": 0.1177,
+      "step": 11465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.014184194754417e-06,
+      "loss": 0.2052,
+      "step": 11470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.012757213390794e-06,
+      "loss": 0.047,
+      "step": 11475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.01133023202717e-06,
+      "loss": 0.1613,
+      "step": 11480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.009903250663547e-06,
+      "loss": 0.1665,
+      "step": 11485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.0084762692999235e-06,
+      "loss": 0.1467,
+      "step": 11490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.007049287936301e-06,
+      "loss": 0.317,
+      "step": 11495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.005622306572677e-06,
+      "loss": 0.5236,
+      "step": 11500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.004195325209053e-06,
+      "loss": 0.1651,
+      "step": 11505
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.00276834384543e-06,
+      "loss": 0.1625,
+      "step": 11510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.001341362481806e-06,
+      "loss": 0.2025,
+      "step": 11515
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.999914381118183e-06,
+      "loss": 0.1807,
+      "step": 11520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9984873997545595e-06,
+      "loss": 0.2067,
+      "step": 11525
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.997060418390937e-06,
+      "loss": 0.0538,
+      "step": 11530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.995633437027313e-06,
+      "loss": 0.1535,
+      "step": 11535
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.994206455663689e-06,
+      "loss": 0.2358,
+      "step": 11540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9927794743000666e-06,
+      "loss": 0.1317,
+      "step": 11545
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.991352492936443e-06,
+      "loss": 0.3778,
+      "step": 11550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.989925511572819e-06,
+      "loss": 0.2059,
+      "step": 11555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.988498530209196e-06,
+      "loss": 0.1814,
+      "step": 11560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.987071548845573e-06,
+      "loss": 0.1676,
+      "step": 11565
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.985644567481949e-06,
+      "loss": 0.1986,
+      "step": 11570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.984217586118326e-06,
+      "loss": 0.1007,
+      "step": 11575
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.982790604754703e-06,
+      "loss": 0.1667,
+      "step": 11580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.981363623391079e-06,
+      "loss": 0.1611,
+      "step": 11585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.979936642027456e-06,
+      "loss": 0.2181,
+      "step": 11590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9785096606638325e-06,
+      "loss": 0.2542,
+      "step": 11595
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.977082679300209e-06,
+      "loss": 0.4274,
+      "step": 11600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.975655697936585e-06,
+      "loss": 0.1906,
+      "step": 11605
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9742287165729615e-06,
+      "loss": 0.1839,
+      "step": 11610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.972801735209339e-06,
+      "loss": 0.1624,
+      "step": 11615
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.971374753845716e-06,
+      "loss": 0.2312,
+      "step": 11620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.969947772482092e-06,
+      "loss": 0.1718,
+      "step": 11625
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9685207911184686e-06,
+      "loss": 0.0537,
+      "step": 11630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.967093809754845e-06,
+      "loss": 0.1676,
+      "step": 11635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.965666828391222e-06,
+      "loss": 0.1831,
+      "step": 11640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9642398470275984e-06,
+      "loss": 0.3179,
+      "step": 11645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.962812865663975e-06,
+      "loss": 0.3606,
+      "step": 11650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.961385884300351e-06,
+      "loss": 0.1715,
+      "step": 11655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.959958902936728e-06,
+      "loss": 0.1673,
+      "step": 11660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9585319215731055e-06,
+      "loss": 0.154,
+      "step": 11665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.957104940209482e-06,
+      "loss": 0.187,
+      "step": 11670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.955677958845858e-06,
+      "loss": 0.1616,
+      "step": 11675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.9542509774822345e-06,
+      "loss": 0.15,
+      "step": 11680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.952823996118612e-06,
+      "loss": 0.2719,
+      "step": 11685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.951397014754988e-06,
+      "loss": 0.1173,
+      "step": 11690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.949970033391364e-06,
+      "loss": 0.2932,
+      "step": 11695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.948543052027741e-06,
+      "loss": 0.6715,
+      "step": 11700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.947116070664117e-06,
+      "loss": 0.148,
+      "step": 11705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.945689089300495e-06,
+      "loss": 0.1502,
+      "step": 11710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.944262107936871e-06,
+      "loss": 0.2129,
+      "step": 11715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.942835126573248e-06,
+      "loss": 0.1811,
+      "step": 11720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.941408145209624e-06,
+      "loss": 0.2013,
+      "step": 11725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.939981163846001e-06,
+      "loss": 0.0523,
+      "step": 11730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.938554182482378e-06,
+      "loss": 0.1874,
+      "step": 11735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.937127201118754e-06,
+      "loss": 0.0713,
+      "step": 11740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.93570021975513e-06,
+      "loss": 0.2806,
+      "step": 11745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.934273238391507e-06,
+      "loss": 0.2965,
+      "step": 11750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.932846257027885e-06,
+      "loss": 0.1685,
+      "step": 11755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.931419275664261e-06,
+      "loss": 0.1686,
+      "step": 11760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.929992294300637e-06,
+      "loss": 0.1532,
+      "step": 11765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.928565312937014e-06,
+      "loss": 0.1886,
+      "step": 11770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.92713833157339e-06,
+      "loss": 0.1128,
+      "step": 11775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.925711350209767e-06,
+      "loss": 0.0301,
+      "step": 11780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.9242843688461435e-06,
+      "loss": 0.1901,
+      "step": 11785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.92285738748252e-06,
+      "loss": 0.2202,
+      "step": 11790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.921430406118896e-06,
+      "loss": 0.3021,
+      "step": 11795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.9200034247552726e-06,
+      "loss": 0.5823,
+      "step": 11800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.918576443391651e-06,
+      "loss": 0.4491,
+      "step": 11805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.917149462028027e-06,
+      "loss": 0.1753,
+      "step": 11810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.915722480664403e-06,
+      "loss": 0.1456,
+      "step": 11815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.91429549930078e-06,
+      "loss": 0.1782,
+      "step": 11820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.912868517937157e-06,
+      "loss": 0.0875,
+      "step": 11825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.911441536573533e-06,
+      "loss": 0.0452,
+      "step": 11830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.9100145552099095e-06,
+      "loss": 0.2525,
+      "step": 11835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.908587573846286e-06,
+      "loss": 0.2504,
+      "step": 11840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.907160592482662e-06,
+      "loss": 0.2647,
+      "step": 11845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.90573361111904e-06,
+      "loss": 0.4513,
+      "step": 11850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.9043066297554165e-06,
+      "loss": 0.1579,
+      "step": 11855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.902879648391793e-06,
+      "loss": 0.2213,
+      "step": 11860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.901452667028169e-06,
+      "loss": 0.2816,
+      "step": 11865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.9000256856645455e-06,
+      "loss": 0.1949,
+      "step": 11870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.898598704300923e-06,
+      "loss": 0.1781,
+      "step": 11875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.897171722937299e-06,
+      "loss": 0.1304,
+      "step": 11880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.895744741573675e-06,
+      "loss": 0.1648,
+      "step": 11885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.894317760210052e-06,
+      "loss": 0.1741,
+      "step": 11890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.892890778846428e-06,
+      "loss": 0.1889,
+      "step": 11895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.891463797482806e-06,
+      "loss": 0.5496,
+      "step": 11900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.8900368161191824e-06,
+      "loss": 0.2701,
+      "step": 11905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.888609834755559e-06,
+      "loss": 0.2145,
+      "step": 11910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.887182853391935e-06,
+      "loss": 0.2208,
+      "step": 11915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.885755872028312e-06,
+      "loss": 0.1022,
+      "step": 11920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.884328890664689e-06,
+      "loss": 0.1382,
+      "step": 11925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.882901909301065e-06,
+      "loss": 0.1889,
+      "step": 11930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.881474927937441e-06,
+      "loss": 0.1475,
+      "step": 11935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.880047946573818e-06,
+      "loss": 0.2216,
+      "step": 11940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.878620965210196e-06,
+      "loss": 0.233,
+      "step": 11945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.877193983846572e-06,
+      "loss": 0.5957,
+      "step": 11950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.875767002482948e-06,
+      "loss": 0.1498,
+      "step": 11955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.874340021119325e-06,
+      "loss": 0.1514,
+      "step": 11960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.872913039755701e-06,
+      "loss": 0.1007,
+      "step": 11965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.871486058392078e-06,
+      "loss": 0.1524,
+      "step": 11970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.8700590770284546e-06,
+      "loss": 0.2239,
+      "step": 11975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.868632095664831e-06,
+      "loss": 0.097,
+      "step": 11980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.867205114301207e-06,
+      "loss": 0.1536,
+      "step": 11985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.865778132937584e-06,
+      "loss": 0.2596,
+      "step": 11990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.864351151573962e-06,
+      "loss": 0.2747,
+      "step": 11995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.862924170210338e-06,
+      "loss": 0.3562,
+      "step": 12000
+    },
+    {
+      "epoch": 0.33,
+      "eval_loss": 0.07923547923564911,
+      "eval_runtime": 2106.0892,
+      "eval_samples_per_second": 7.606,
+      "eval_steps_per_second": 1.902,
+      "eval_wer": 0.22589253516047603,
+      "step": 12000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.861497188846714e-06,
+      "loss": 0.1722,
+      "step": 12005
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.860070207483091e-06,
+      "loss": 0.169,
+      "step": 12010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.858643226119468e-06,
+      "loss": 0.1367,
+      "step": 12015
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.857216244755844e-06,
+      "loss": 0.1788,
+      "step": 12020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.8557892633922205e-06,
+      "loss": 0.1067,
+      "step": 12025
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.854362282028597e-06,
+      "loss": 0.086,
+      "step": 12030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.852935300664973e-06,
+      "loss": 0.154,
+      "step": 12035
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.851508319301351e-06,
+      "loss": 0.2277,
+      "step": 12040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.8500813379377275e-06,
+      "loss": 0.2513,
+      "step": 12045
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.848654356574104e-06,
+      "loss": 0.2739,
+      "step": 12050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.84722737521048e-06,
+      "loss": 0.1541,
+      "step": 12055
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.8458003938468566e-06,
+      "loss": 0.1395,
+      "step": 12060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.844373412483234e-06,
+      "loss": 0.1674,
+      "step": 12065
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.84294643111961e-06,
+      "loss": 0.2734,
+      "step": 12070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.841519449755986e-06,
+      "loss": 0.0343,
+      "step": 12075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.840092468392363e-06,
+      "loss": 0.0648,
+      "step": 12080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.838665487028739e-06,
+      "loss": 0.0999,
+      "step": 12085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.837238505665117e-06,
+      "loss": 0.1609,
+      "step": 12090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.8358115243014935e-06,
+      "loss": 0.2068,
+      "step": 12095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.83438454293787e-06,
+      "loss": 0.5631,
+      "step": 12100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.832957561574246e-06,
+      "loss": 0.2012,
+      "step": 12105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.831530580210623e-06,
+      "loss": 0.1435,
+      "step": 12110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.830103598847e-06,
+      "loss": 0.1954,
+      "step": 12115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.828676617483376e-06,
+      "loss": 0.1535,
+      "step": 12120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.827249636119752e-06,
+      "loss": 0.1071,
+      "step": 12125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.825822654756129e-06,
+      "loss": 0.0774,
+      "step": 12130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.824395673392507e-06,
+      "loss": 0.1752,
+      "step": 12135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.822968692028883e-06,
+      "loss": 0.111,
+      "step": 12140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.821541710665259e-06,
+      "loss": 0.2101,
+      "step": 12145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.820114729301636e-06,
+      "loss": 0.6161,
+      "step": 12150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.818687747938012e-06,
+      "loss": 0.1055,
+      "step": 12155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.817260766574389e-06,
+      "loss": 0.1056,
+      "step": 12160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.815833785210766e-06,
+      "loss": 0.1723,
+      "step": 12165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.814406803847142e-06,
+      "loss": 0.1849,
+      "step": 12170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.812979822483518e-06,
+      "loss": 0.1547,
+      "step": 12175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.811552841119895e-06,
+      "loss": 0.0779,
+      "step": 12180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.810125859756273e-06,
+      "loss": 0.0533,
+      "step": 12185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.808698878392649e-06,
+      "loss": 0.1573,
+      "step": 12190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.807271897029025e-06,
+      "loss": 0.2096,
+      "step": 12195
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.805844915665402e-06,
+      "loss": 0.4607,
+      "step": 12200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.804417934301779e-06,
+      "loss": 0.1499,
+      "step": 12205
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.802990952938155e-06,
+      "loss": 0.196,
+      "step": 12210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.8015639715745315e-06,
+      "loss": 0.2675,
+      "step": 12215
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.800136990210908e-06,
+      "loss": 0.1761,
+      "step": 12220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.798710008847284e-06,
+      "loss": 0.1389,
+      "step": 12225
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.797283027483662e-06,
+      "loss": 0.0576,
+      "step": 12230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7958560461200386e-06,
+      "loss": 0.2832,
+      "step": 12235
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.794429064756415e-06,
+      "loss": 0.1552,
+      "step": 12240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.793002083392791e-06,
+      "loss": 0.2106,
+      "step": 12245
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.791575102029168e-06,
+      "loss": 0.3959,
+      "step": 12250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.790148120665545e-06,
+      "loss": 0.1223,
+      "step": 12255
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.788721139301921e-06,
+      "loss": 0.1802,
+      "step": 12260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7872941579382975e-06,
+      "loss": 0.1227,
+      "step": 12265
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.785867176574674e-06,
+      "loss": 0.1722,
+      "step": 12270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.784440195211051e-06,
+      "loss": 0.1879,
+      "step": 12275
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.783013213847428e-06,
+      "loss": 0.1466,
+      "step": 12280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7815862324838045e-06,
+      "loss": 0.0777,
+      "step": 12285
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.780159251120181e-06,
+      "loss": 0.0767,
+      "step": 12290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.778732269756557e-06,
+      "loss": 0.209,
+      "step": 12295
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.777305288392934e-06,
+      "loss": 0.4856,
+      "step": 12300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.775878307029311e-06,
+      "loss": 0.2182,
+      "step": 12305
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.774451325665687e-06,
+      "loss": 0.193,
+      "step": 12310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.773024344302063e-06,
+      "loss": 0.1729,
+      "step": 12315
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7715973629384406e-06,
+      "loss": 0.1722,
+      "step": 12320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.770170381574818e-06,
+      "loss": 0.1712,
+      "step": 12325
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.768743400211194e-06,
+      "loss": 0.1361,
+      "step": 12330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7673164188475704e-06,
+      "loss": 0.0951,
+      "step": 12335
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.765889437483947e-06,
+      "loss": 0.1864,
+      "step": 12340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.764462456120323e-06,
+      "loss": 0.3715,
+      "step": 12345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7630354747567e-06,
+      "loss": 0.4201,
+      "step": 12350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.761608493393077e-06,
+      "loss": 0.1058,
+      "step": 12355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.760181512029453e-06,
+      "loss": 0.1139,
+      "step": 12360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.75875453066583e-06,
+      "loss": 0.1212,
+      "step": 12365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7573275493022065e-06,
+      "loss": 0.1982,
+      "step": 12370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.755900567938584e-06,
+      "loss": 0.2394,
+      "step": 12375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.75447358657496e-06,
+      "loss": 0.3017,
+      "step": 12380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.753046605211336e-06,
+      "loss": 0.1209,
+      "step": 12385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.751619623847713e-06,
+      "loss": 0.284,
+      "step": 12390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.75019264248409e-06,
+      "loss": 0.1577,
+      "step": 12395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.748765661120466e-06,
+      "loss": 0.3294,
+      "step": 12400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7473386797568426e-06,
+      "loss": 0.1664,
+      "step": 12405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.74591169839322e-06,
+      "loss": 0.1291,
+      "step": 12410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.744484717029596e-06,
+      "loss": 0.1548,
+      "step": 12415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.743057735665973e-06,
+      "loss": 0.1263,
+      "step": 12420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.74163075430235e-06,
+      "loss": 0.1347,
+      "step": 12425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.740203772938726e-06,
+      "loss": 0.0752,
+      "step": 12430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.738776791575102e-06,
+      "loss": 0.1343,
+      "step": 12435
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.737349810211479e-06,
+      "loss": 0.3127,
+      "step": 12440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.735922828847856e-06,
+      "loss": 0.2375,
+      "step": 12445
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.734495847484233e-06,
+      "loss": 0.4197,
+      "step": 12450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.733068866120609e-06,
+      "loss": 0.1688,
+      "step": 12455
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.731641884756986e-06,
+      "loss": 0.1575,
+      "step": 12460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.730214903393362e-06,
+      "loss": 0.1481,
+      "step": 12465
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.728787922029739e-06,
+      "loss": 0.1359,
+      "step": 12470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.7273609406661155e-06,
+      "loss": 0.1666,
+      "step": 12475
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.725933959302492e-06,
+      "loss": 0.1549,
+      "step": 12480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.724506977938868e-06,
+      "loss": 0.2111,
+      "step": 12485
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.723079996575245e-06,
+      "loss": 0.0749,
+      "step": 12490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.721653015211623e-06,
+      "loss": 0.3546,
+      "step": 12495
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.720226033847999e-06,
+      "loss": 0.533,
+      "step": 12500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.718799052484375e-06,
+      "loss": 0.2164,
+      "step": 12505
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.717372071120752e-06,
+      "loss": 0.178,
+      "step": 12510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.715945089757129e-06,
+      "loss": 0.1629,
+      "step": 12515
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.714518108393505e-06,
+      "loss": 0.156,
+      "step": 12520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.7130911270298815e-06,
+      "loss": 0.1892,
+      "step": 12525
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.711664145666258e-06,
+      "loss": 0.1118,
+      "step": 12530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.710237164302634e-06,
+      "loss": 0.1347,
+      "step": 12535
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.708810182939012e-06,
+      "loss": 0.2332,
+      "step": 12540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.7073832015753885e-06,
+      "loss": 0.3007,
+      "step": 12545
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.705956220211765e-06,
+      "loss": 0.3231,
+      "step": 12550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.704529238848141e-06,
+      "loss": 0.1384,
+      "step": 12555
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.7031022574845175e-06,
+      "loss": 0.1172,
+      "step": 12560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.701675276120895e-06,
+      "loss": 0.2045,
+      "step": 12565
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.700248294757271e-06,
+      "loss": 0.188,
+      "step": 12570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.698821313393647e-06,
+      "loss": 0.1805,
+      "step": 12575
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.697394332030024e-06,
+      "loss": 0.1306,
+      "step": 12580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.695967350666402e-06,
+      "loss": 0.2015,
+      "step": 12585
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.694540369302778e-06,
+      "loss": 0.318,
+      "step": 12590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.6931133879391544e-06,
+      "loss": 0.22,
+      "step": 12595
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.691686406575531e-06,
+      "loss": 0.4261,
+      "step": 12600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.690259425211907e-06,
+      "loss": 0.1451,
+      "step": 12605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.688832443848284e-06,
+      "loss": 0.2177,
+      "step": 12610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.687405462484661e-06,
+      "loss": 0.1622,
+      "step": 12615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.685978481121037e-06,
+      "loss": 0.1724,
+      "step": 12620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.684551499757413e-06,
+      "loss": 0.0654,
+      "step": 12625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.68312451839379e-06,
+      "loss": 0.056,
+      "step": 12630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.681697537030168e-06,
+      "loss": 0.1806,
+      "step": 12635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.680270555666544e-06,
+      "loss": 0.1486,
+      "step": 12640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.67884357430292e-06,
+      "loss": 0.4049,
+      "step": 12645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.677416592939297e-06,
+      "loss": 0.2504,
+      "step": 12650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.675989611575673e-06,
+      "loss": 0.2093,
+      "step": 12655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.67456263021205e-06,
+      "loss": 0.1685,
+      "step": 12660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.6731356488484266e-06,
+      "loss": 0.1728,
+      "step": 12665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.671708667484803e-06,
+      "loss": 0.1841,
+      "step": 12670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.670281686121179e-06,
+      "loss": 0.056,
+      "step": 12675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.668854704757557e-06,
+      "loss": 0.1487,
+      "step": 12680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.667427723393934e-06,
+      "loss": 0.1941,
+      "step": 12685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.66600074203031e-06,
+      "loss": 0.1175,
+      "step": 12690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.664573760666686e-06,
+      "loss": 0.3037,
+      "step": 12695
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.663146779303063e-06,
+      "loss": 0.5031,
+      "step": 12700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.66171979793944e-06,
+      "loss": 0.1533,
+      "step": 12705
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.660292816575816e-06,
+      "loss": 0.1927,
+      "step": 12710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.6588658352121925e-06,
+      "loss": 0.1464,
+      "step": 12715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.657438853848569e-06,
+      "loss": 0.1705,
+      "step": 12720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.656011872484945e-06,
+      "loss": 0.153,
+      "step": 12725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.654584891121323e-06,
+      "loss": 0.075,
+      "step": 12730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.6531579097576995e-06,
+      "loss": 0.1802,
+      "step": 12735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.651730928394076e-06,
+      "loss": 0.1223,
+      "step": 12740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.650303947030452e-06,
+      "loss": 0.1696,
+      "step": 12745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.6488769656668286e-06,
+      "loss": 0.2838,
+      "step": 12750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.647449984303206e-06,
+      "loss": 0.1788,
+      "step": 12755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.646023002939582e-06,
+      "loss": 0.1858,
+      "step": 12760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.644596021575958e-06,
+      "loss": 0.1803,
+      "step": 12765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.643169040212335e-06,
+      "loss": 0.2407,
+      "step": 12770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.641742058848713e-06,
+      "loss": 0.1448,
+      "step": 12775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.640315077485089e-06,
+      "loss": 0.0726,
+      "step": 12780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.6388880961214655e-06,
+      "loss": 0.149,
+      "step": 12785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.637461114757842e-06,
+      "loss": 0.1544,
+      "step": 12790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.636034133394218e-06,
+      "loss": 0.2843,
+      "step": 12795
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.634607152030595e-06,
+      "loss": 0.4658,
+      "step": 12800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.633180170666972e-06,
+      "loss": 0.1841,
+      "step": 12805
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.631753189303348e-06,
+      "loss": 0.1337,
+      "step": 12810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.630326207939724e-06,
+      "loss": 0.1977,
+      "step": 12815
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.628899226576101e-06,
+      "loss": 0.1626,
+      "step": 12820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.627472245212479e-06,
+      "loss": 0.1412,
+      "step": 12825
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.626045263848855e-06,
+      "loss": 0.0621,
+      "step": 12830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.624618282485231e-06,
+      "loss": 0.1898,
+      "step": 12835
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.623191301121608e-06,
+      "loss": 0.2279,
+      "step": 12840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.621764319757984e-06,
+      "loss": 0.1943,
+      "step": 12845
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.620337338394361e-06,
+      "loss": 0.3011,
+      "step": 12850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.618910357030738e-06,
+      "loss": 0.1656,
+      "step": 12855
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.617483375667114e-06,
+      "loss": 0.1859,
+      "step": 12860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.61605639430349e-06,
+      "loss": 0.166,
+      "step": 12865
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.614629412939868e-06,
+      "loss": 0.1987,
+      "step": 12870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.613202431576245e-06,
+      "loss": 0.1146,
+      "step": 12875
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.611775450212621e-06,
+      "loss": 0.0917,
+      "step": 12880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.610348468848997e-06,
+      "loss": 0.1346,
+      "step": 12885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.608921487485374e-06,
+      "loss": 0.172,
+      "step": 12890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.607494506121751e-06,
+      "loss": 0.2051,
+      "step": 12895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.606067524758127e-06,
+      "loss": 0.3892,
+      "step": 12900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.6046405433945035e-06,
+      "loss": 0.1457,
+      "step": 12905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.60321356203088e-06,
+      "loss": 0.1606,
+      "step": 12910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.601786580667256e-06,
+      "loss": 0.1592,
+      "step": 12915
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.600359599303634e-06,
+      "loss": 0.1946,
+      "step": 12920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.598932617940011e-06,
+      "loss": 0.1297,
+      "step": 12925
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.597505636576387e-06,
+      "loss": 0.1505,
+      "step": 12930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.596078655212763e-06,
+      "loss": 0.1718,
+      "step": 12935
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5946516738491404e-06,
+      "loss": 0.4016,
+      "step": 12940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.593224692485517e-06,
+      "loss": 0.2401,
+      "step": 12945
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.591797711121893e-06,
+      "loss": 0.3069,
+      "step": 12950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5903707297582695e-06,
+      "loss": 0.228,
+      "step": 12955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.588943748394646e-06,
+      "loss": 0.1626,
+      "step": 12960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.587516767031024e-06,
+      "loss": 0.1761,
+      "step": 12965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5860897856674e-06,
+      "loss": 0.1507,
+      "step": 12970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5846628043037765e-06,
+      "loss": 0.1849,
+      "step": 12975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.583235822940153e-06,
+      "loss": 0.1138,
+      "step": 12980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.581808841576529e-06,
+      "loss": 0.2012,
+      "step": 12985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.580381860212906e-06,
+      "loss": 0.163,
+      "step": 12990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.578954878849283e-06,
+      "loss": 0.2199,
+      "step": 12995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.577527897485659e-06,
+      "loss": 0.525,
+      "step": 13000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.576100916122035e-06,
+      "loss": 0.1355,
+      "step": 13005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.574673934758412e-06,
+      "loss": 0.0978,
+      "step": 13010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.57324695339479e-06,
+      "loss": 0.1562,
+      "step": 13015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.571819972031166e-06,
+      "loss": 0.2119,
+      "step": 13020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5703929906675424e-06,
+      "loss": 0.0624,
+      "step": 13025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.568966009303919e-06,
+      "loss": 0.0974,
+      "step": 13030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.567539027940296e-06,
+      "loss": 0.2115,
+      "step": 13035
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.566112046576672e-06,
+      "loss": 0.1932,
+      "step": 13040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.564685065213049e-06,
+      "loss": 0.3689,
+      "step": 13045
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.563258083849425e-06,
+      "loss": 0.4729,
+      "step": 13050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.561831102485801e-06,
+      "loss": 0.1639,
+      "step": 13055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.560404121122179e-06,
+      "loss": 0.1029,
+      "step": 13060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.558977139758556e-06,
+      "loss": 0.1764,
+      "step": 13065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.557550158394932e-06,
+      "loss": 0.2053,
+      "step": 13070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.556123177031308e-06,
+      "loss": 0.1633,
+      "step": 13075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.554696195667685e-06,
+      "loss": 0.2154,
+      "step": 13080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.553269214304062e-06,
+      "loss": 0.1189,
+      "step": 13085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.551842232940438e-06,
+      "loss": 0.1119,
+      "step": 13090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5504152515768146e-06,
+      "loss": 0.298,
+      "step": 13095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.548988270213191e-06,
+      "loss": 0.3527,
+      "step": 13100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.547561288849567e-06,
+      "loss": 0.2411,
+      "step": 13105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.546134307485945e-06,
+      "loss": 0.1694,
+      "step": 13110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.544707326122322e-06,
+      "loss": 0.129,
+      "step": 13115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.543280344758698e-06,
+      "loss": 0.2187,
+      "step": 13120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.541853363395074e-06,
+      "loss": 0.1709,
+      "step": 13125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5404263820314515e-06,
+      "loss": 0.1164,
+      "step": 13130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.538999400667828e-06,
+      "loss": 0.0528,
+      "step": 13135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.537572419304204e-06,
+      "loss": 0.1298,
+      "step": 13140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.5361454379405805e-06,
+      "loss": 0.2292,
+      "step": 13145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.534718456576957e-06,
+      "loss": 0.2503,
+      "step": 13150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.533291475213335e-06,
+      "loss": 0.1328,
+      "step": 13155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.531864493849711e-06,
+      "loss": 0.1546,
+      "step": 13160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.5304375124860875e-06,
+      "loss": 0.202,
+      "step": 13165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.529010531122464e-06,
+      "loss": 0.2743,
+      "step": 13170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.52758354975884e-06,
+      "loss": 0.0807,
+      "step": 13175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.526156568395217e-06,
+      "loss": 0.0131,
+      "step": 13180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.524729587031594e-06,
+      "loss": 0.0778,
+      "step": 13185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.52330260566797e-06,
+      "loss": 0.1794,
+      "step": 13190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.521875624304347e-06,
+      "loss": 0.3173,
+      "step": 13195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.520448642940724e-06,
+      "loss": 0.3544,
+      "step": 13200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.519021661577101e-06,
+      "loss": 0.2196,
+      "step": 13205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.517594680213477e-06,
+      "loss": 0.1745,
+      "step": 13210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.5161676988498535e-06,
+      "loss": 0.1374,
+      "step": 13215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.51474071748623e-06,
+      "loss": 0.2678,
+      "step": 13220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.513313736122607e-06,
+      "loss": 0.1534,
+      "step": 13225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.511886754758983e-06,
+      "loss": 0.0927,
+      "step": 13230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.51045977339536e-06,
+      "loss": 0.1013,
+      "step": 13235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.509032792031737e-06,
+      "loss": 0.2501,
+      "step": 13240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.507605810668113e-06,
+      "loss": 0.2229,
+      "step": 13245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.50617882930449e-06,
+      "loss": 0.2752,
+      "step": 13250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.504751847940867e-06,
+      "loss": 0.1636,
+      "step": 13255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.503324866577243e-06,
+      "loss": 0.136,
+      "step": 13260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.501897885213619e-06,
+      "loss": 0.14,
+      "step": 13265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.500470903849996e-06,
+      "loss": 0.1113,
+      "step": 13270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.499043922486373e-06,
+      "loss": 0.165,
+      "step": 13275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.497616941122749e-06,
+      "loss": 0.0431,
+      "step": 13280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4961899597591264e-06,
+      "loss": 0.0915,
+      "step": 13285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.494762978395503e-06,
+      "loss": 0.0957,
+      "step": 13290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.493335997031879e-06,
+      "loss": 0.3148,
+      "step": 13295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.491909015668256e-06,
+      "loss": 0.2711,
+      "step": 13300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.490482034304633e-06,
+      "loss": 0.2059,
+      "step": 13305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.489055052941009e-06,
+      "loss": 0.1473,
+      "step": 13310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.487628071577385e-06,
+      "loss": 0.2038,
+      "step": 13315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4862010902137625e-06,
+      "loss": 0.2306,
+      "step": 13320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.484774108850139e-06,
+      "loss": 0.1228,
+      "step": 13325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.483347127486516e-06,
+      "loss": 0.0809,
+      "step": 13330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.481920146122892e-06,
+      "loss": 0.1194,
+      "step": 13335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.480493164759269e-06,
+      "loss": 0.3542,
+      "step": 13340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.479066183395646e-06,
+      "loss": 0.3603,
+      "step": 13345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.477639202032022e-06,
+      "loss": 0.4615,
+      "step": 13350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4762122206683986e-06,
+      "loss": 0.2122,
+      "step": 13355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.474785239304775e-06,
+      "loss": 0.1875,
+      "step": 13360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.473358257941151e-06,
+      "loss": 0.1534,
+      "step": 13365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4719312765775284e-06,
+      "loss": 0.1891,
+      "step": 13370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.470504295213906e-06,
+      "loss": 0.2287,
+      "step": 13375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.469077313850282e-06,
+      "loss": 0.1701,
+      "step": 13380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.467650332486658e-06,
+      "loss": 0.0908,
+      "step": 13385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.466223351123035e-06,
+      "loss": 0.1964,
+      "step": 13390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.464796369759412e-06,
+      "loss": 0.2054,
+      "step": 13395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.463369388395788e-06,
+      "loss": 0.5634,
+      "step": 13400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4619424070321645e-06,
+      "loss": 0.1152,
+      "step": 13405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.460515425668541e-06,
+      "loss": 0.2444,
+      "step": 13410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.459088444304918e-06,
+      "loss": 0.1514,
+      "step": 13415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.457661462941295e-06,
+      "loss": 0.1599,
+      "step": 13420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4562344815776715e-06,
+      "loss": 0.188,
+      "step": 13425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.454807500214048e-06,
+      "loss": 0.0559,
+      "step": 13430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.453380518850424e-06,
+      "loss": 0.022,
+      "step": 13435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.451953537486801e-06,
+      "loss": 0.1629,
+      "step": 13440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.450526556123178e-06,
+      "loss": 0.295,
+      "step": 13445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.449099574759554e-06,
+      "loss": 0.6125,
+      "step": 13450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4476725933959304e-06,
+      "loss": 0.184,
+      "step": 13455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.446245612032307e-06,
+      "loss": 0.2965,
+      "step": 13460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.444818630668685e-06,
+      "loss": 0.1644,
+      "step": 13465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.443391649305061e-06,
+      "loss": 0.1635,
+      "step": 13470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.4419646679414375e-06,
+      "loss": 0.1359,
+      "step": 13475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.440537686577814e-06,
+      "loss": 0.1287,
+      "step": 13480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.43911070521419e-06,
+      "loss": 0.1645,
+      "step": 13485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.437683723850567e-06,
+      "loss": 0.1742,
+      "step": 13490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.436256742486944e-06,
+      "loss": 0.1094,
+      "step": 13495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.43482976112332e-06,
+      "loss": 0.3841,
+      "step": 13500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.433402779759696e-06,
+      "loss": 0.1358,
+      "step": 13505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.431975798396074e-06,
+      "loss": 0.1486,
+      "step": 13510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.430548817032451e-06,
+      "loss": 0.1687,
+      "step": 13515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.429121835668827e-06,
+      "loss": 0.2705,
+      "step": 13520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.427694854305203e-06,
+      "loss": 0.0766,
+      "step": 13525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.42626787294158e-06,
+      "loss": 0.1227,
+      "step": 13530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.424840891577957e-06,
+      "loss": 0.1874,
+      "step": 13535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.423413910214333e-06,
+      "loss": 0.1504,
+      "step": 13540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.42198692885071e-06,
+      "loss": 0.1187,
+      "step": 13545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.420559947487086e-06,
+      "loss": 0.3289,
+      "step": 13550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.419132966123462e-06,
+      "loss": 0.1386,
+      "step": 13555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.41770598475984e-06,
+      "loss": 0.1542,
+      "step": 13560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.416279003396217e-06,
+      "loss": 0.16,
+      "step": 13565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.414852022032593e-06,
+      "loss": 0.2421,
+      "step": 13570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.413425040668969e-06,
+      "loss": 0.1571,
+      "step": 13575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.411998059305346e-06,
+      "loss": 0.0579,
+      "step": 13580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.410571077941723e-06,
+      "loss": 0.1032,
+      "step": 13585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.409144096578099e-06,
+      "loss": 0.0885,
+      "step": 13590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.4077171152144755e-06,
+      "loss": 0.2286,
+      "step": 13595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.406290133850852e-06,
+      "loss": 0.6907,
+      "step": 13600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.40486315248723e-06,
+      "loss": 0.1718,
+      "step": 13605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.403436171123606e-06,
+      "loss": 0.1999,
+      "step": 13610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.402009189759983e-06,
+      "loss": 0.2339,
+      "step": 13615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.400582208396359e-06,
+      "loss": 0.1723,
+      "step": 13620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.399155227032735e-06,
+      "loss": 0.0938,
+      "step": 13625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3977282456691124e-06,
+      "loss": 0.0504,
+      "step": 13630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.396301264305489e-06,
+      "loss": 0.1868,
+      "step": 13635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.394874282941865e-06,
+      "loss": 0.2351,
+      "step": 13640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3934473015782415e-06,
+      "loss": 0.3668,
+      "step": 13645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.392020320214618e-06,
+      "loss": 0.6481,
+      "step": 13650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.390593338850996e-06,
+      "loss": 0.1458,
+      "step": 13655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.389166357487372e-06,
+      "loss": 0.2637,
+      "step": 13660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3877393761237485e-06,
+      "loss": 0.2027,
+      "step": 13665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.386312394760125e-06,
+      "loss": 0.2086,
+      "step": 13670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.384885413396501e-06,
+      "loss": 0.0929,
+      "step": 13675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.383458432032878e-06,
+      "loss": 0.0966,
+      "step": 13680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.382031450669255e-06,
+      "loss": 0.2651,
+      "step": 13685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.380604469305631e-06,
+      "loss": 0.2294,
+      "step": 13690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.379177487942007e-06,
+      "loss": 0.406,
+      "step": 13695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.377750506578385e-06,
+      "loss": 0.4792,
+      "step": 13700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.376323525214762e-06,
+      "loss": 0.1531,
+      "step": 13705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.374896543851138e-06,
+      "loss": 0.1431,
+      "step": 13710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3734695624875144e-06,
+      "loss": 0.1865,
+      "step": 13715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.372042581123891e-06,
+      "loss": 0.1561,
+      "step": 13720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.370615599760268e-06,
+      "loss": 0.1485,
+      "step": 13725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.369188618396644e-06,
+      "loss": 0.0859,
+      "step": 13730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.367761637033021e-06,
+      "loss": 0.2794,
+      "step": 13735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.366334655669397e-06,
+      "loss": 0.2333,
+      "step": 13740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.364907674305773e-06,
+      "loss": 0.3636,
+      "step": 13745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.363480692942151e-06,
+      "loss": 0.5979,
+      "step": 13750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.362053711578528e-06,
+      "loss": 0.1171,
+      "step": 13755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.360626730214904e-06,
+      "loss": 0.1514,
+      "step": 13760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.35919974885128e-06,
+      "loss": 0.1609,
+      "step": 13765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.357772767487657e-06,
+      "loss": 0.1385,
+      "step": 13770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.356345786124034e-06,
+      "loss": 0.0788,
+      "step": 13775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.35491880476041e-06,
+      "loss": 0.1974,
+      "step": 13780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3534918233967866e-06,
+      "loss": 0.1626,
+      "step": 13785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.352064842033163e-06,
+      "loss": 0.3104,
+      "step": 13790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.350637860669541e-06,
+      "loss": 0.2074,
+      "step": 13795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.349210879305917e-06,
+      "loss": 0.2424,
+      "step": 13800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.347783897942294e-06,
+      "loss": 0.1079,
+      "step": 13805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.34635691657867e-06,
+      "loss": 0.1808,
+      "step": 13810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.344929935215046e-06,
+      "loss": 0.1682,
+      "step": 13815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3435029538514235e-06,
+      "loss": 0.2335,
+      "step": 13820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3420759724878e-06,
+      "loss": 0.2089,
+      "step": 13825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.340648991124176e-06,
+      "loss": 0.1013,
+      "step": 13830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3392220097605525e-06,
+      "loss": 0.1075,
+      "step": 13835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.337795028396929e-06,
+      "loss": 0.204,
+      "step": 13840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.336368047033307e-06,
+      "loss": 0.1115,
+      "step": 13845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.334941065669683e-06,
+      "loss": 0.2524,
+      "step": 13850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.3335140843060595e-06,
+      "loss": 0.1643,
+      "step": 13855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.332087102942436e-06,
+      "loss": 0.1864,
+      "step": 13860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.330660121578812e-06,
+      "loss": 0.1753,
+      "step": 13865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.329233140215189e-06,
+      "loss": 0.2333,
+      "step": 13870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.327806158851566e-06,
+      "loss": 0.2355,
+      "step": 13875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.326379177487942e-06,
+      "loss": 0.0981,
+      "step": 13880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.324952196124318e-06,
+      "loss": 0.1081,
+      "step": 13885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.3235252147606964e-06,
+      "loss": 0.2217,
+      "step": 13890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.322098233397073e-06,
+      "loss": 0.2834,
+      "step": 13895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.320671252033449e-06,
+      "loss": 0.3532,
+      "step": 13900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.3192442706698255e-06,
+      "loss": 0.1661,
+      "step": 13905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.317817289306202e-06,
+      "loss": 0.1203,
+      "step": 13910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.316390307942579e-06,
+      "loss": 0.2317,
+      "step": 13915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.314963326578955e-06,
+      "loss": 0.216,
+      "step": 13920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.313536345215332e-06,
+      "loss": 0.1415,
+      "step": 13925
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.312109363851708e-06,
+      "loss": 0.2143,
+      "step": 13930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.310682382488084e-06,
+      "loss": 0.0619,
+      "step": 13935
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.309255401124462e-06,
+      "loss": 0.2325,
+      "step": 13940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.307828419760839e-06,
+      "loss": 0.1864,
+      "step": 13945
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.306401438397215e-06,
+      "loss": 0.4923,
+      "step": 13950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.304974457033591e-06,
+      "loss": 0.167,
+      "step": 13955
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.303547475669968e-06,
+      "loss": 0.2789,
+      "step": 13960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.302120494306345e-06,
+      "loss": 0.3597,
+      "step": 13965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.300693512942721e-06,
+      "loss": 0.2088,
+      "step": 13970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.299266531579098e-06,
+      "loss": 0.2258,
+      "step": 13975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.297839550215474e-06,
+      "loss": 0.0746,
+      "step": 13980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.296412568851852e-06,
+      "loss": 0.1711,
+      "step": 13985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.294985587488228e-06,
+      "loss": 0.1465,
+      "step": 13990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.293558606124605e-06,
+      "loss": 0.277,
+      "step": 13995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.292131624760981e-06,
+      "loss": 0.3886,
+      "step": 14000
+    },
+    {
+      "epoch": 0.39,
+      "eval_loss": 0.06621355563402176,
+      "eval_runtime": 2033.5992,
+      "eval_samples_per_second": 7.877,
+      "eval_steps_per_second": 1.969,
+      "eval_wer": 0.19680851063829788,
+      "step": 14000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.290704643397357e-06,
+      "loss": 0.1386,
+      "step": 14005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2892776620337345e-06,
+      "loss": 0.2327,
+      "step": 14010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.287850680670111e-06,
+      "loss": 0.2392,
+      "step": 14015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.286423699306487e-06,
+      "loss": 0.2466,
+      "step": 14020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2849967179428635e-06,
+      "loss": 0.1271,
+      "step": 14025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.283569736579241e-06,
+      "loss": 0.092,
+      "step": 14030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.282142755215618e-06,
+      "loss": 0.1979,
+      "step": 14035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.280715773851994e-06,
+      "loss": 0.1962,
+      "step": 14040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2792887924883706e-06,
+      "loss": 0.3097,
+      "step": 14045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.277861811124747e-06,
+      "loss": 0.4639,
+      "step": 14050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.276434829761123e-06,
+      "loss": 0.1918,
+      "step": 14055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2750078483975004e-06,
+      "loss": 0.2204,
+      "step": 14060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.273580867033877e-06,
+      "loss": 0.1276,
+      "step": 14065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.272153885670253e-06,
+      "loss": 0.125,
+      "step": 14070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.27072690430663e-06,
+      "loss": 0.1187,
+      "step": 14075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2692999229430075e-06,
+      "loss": 0.1567,
+      "step": 14080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.267872941579384e-06,
+      "loss": 0.1703,
+      "step": 14085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.26644596021576e-06,
+      "loss": 0.2286,
+      "step": 14090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2650189788521365e-06,
+      "loss": 0.2252,
+      "step": 14095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.263591997488513e-06,
+      "loss": 0.5425,
+      "step": 14100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.26216501612489e-06,
+      "loss": 0.1377,
+      "step": 14105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.260738034761266e-06,
+      "loss": 0.2165,
+      "step": 14110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.259311053397643e-06,
+      "loss": 0.1183,
+      "step": 14115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.25788407203402e-06,
+      "loss": 0.1774,
+      "step": 14120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.256457090670396e-06,
+      "loss": 0.196,
+      "step": 14125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.255030109306773e-06,
+      "loss": 0.0132,
+      "step": 14130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.25360312794315e-06,
+      "loss": 0.1714,
+      "step": 14135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.252176146579526e-06,
+      "loss": 0.2184,
+      "step": 14140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2507491652159024e-06,
+      "loss": 0.2794,
+      "step": 14145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.24932218385228e-06,
+      "loss": 0.3971,
+      "step": 14150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.247895202488656e-06,
+      "loss": 0.1343,
+      "step": 14155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.246468221125032e-06,
+      "loss": 0.1995,
+      "step": 14160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2450412397614095e-06,
+      "loss": 0.1216,
+      "step": 14165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.243614258397786e-06,
+      "loss": 0.2421,
+      "step": 14170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.242187277034163e-06,
+      "loss": 0.1313,
+      "step": 14175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.240760295670539e-06,
+      "loss": 0.1285,
+      "step": 14180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.239333314306916e-06,
+      "loss": 0.2933,
+      "step": 14185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.237906332943292e-06,
+      "loss": 0.1455,
+      "step": 14190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.236479351579668e-06,
+      "loss": 0.2936,
+      "step": 14195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.2350523702160455e-06,
+      "loss": 0.6418,
+      "step": 14200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.233625388852422e-06,
+      "loss": 0.1547,
+      "step": 14205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.232198407488799e-06,
+      "loss": 0.1088,
+      "step": 14210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.230771426125175e-06,
+      "loss": 0.1535,
+      "step": 14215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.229344444761552e-06,
+      "loss": 0.1816,
+      "step": 14220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.227917463397929e-06,
+      "loss": 0.2131,
+      "step": 14225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.226490482034305e-06,
+      "loss": 0.0358,
+      "step": 14230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.225063500670682e-06,
+      "loss": 0.116,
+      "step": 14235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.223636519307058e-06,
+      "loss": 0.0447,
+      "step": 14240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.222209537943435e-06,
+      "loss": 0.2341,
+      "step": 14245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.220782556579812e-06,
+      "loss": 0.2686,
+      "step": 14250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.219355575216189e-06,
+      "loss": 0.1419,
+      "step": 14255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.217928593852565e-06,
+      "loss": 0.1696,
+      "step": 14260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.216501612488941e-06,
+      "loss": 0.2146,
+      "step": 14265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.2150746311253185e-06,
+      "loss": 0.1784,
+      "step": 14270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.213647649761695e-06,
+      "loss": 0.1043,
+      "step": 14275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.212220668398071e-06,
+      "loss": 0.1636,
+      "step": 14280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.2107936870344475e-06,
+      "loss": 0.0623,
+      "step": 14285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.209366705670824e-06,
+      "loss": 0.2764,
+      "step": 14290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.207939724307202e-06,
+      "loss": 0.3821,
+      "step": 14295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.206512742943578e-06,
+      "loss": 0.3635,
+      "step": 14300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.205085761579955e-06,
+      "loss": 0.1758,
+      "step": 14305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.203658780216331e-06,
+      "loss": 0.2334,
+      "step": 14310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.202231798852707e-06,
+      "loss": 0.2056,
+      "step": 14315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.2008048174890844e-06,
+      "loss": 0.205,
+      "step": 14320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.199377836125461e-06,
+      "loss": 0.0321,
+      "step": 14325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.197950854761837e-06,
+      "loss": 0.1652,
+      "step": 14330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1965238733982135e-06,
+      "loss": 0.0713,
+      "step": 14335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1950968920345915e-06,
+      "loss": 0.2637,
+      "step": 14340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.193669910670968e-06,
+      "loss": 0.1655,
+      "step": 14345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.192242929307344e-06,
+      "loss": 0.5425,
+      "step": 14350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1908159479437205e-06,
+      "loss": 0.1133,
+      "step": 14355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.189388966580097e-06,
+      "loss": 0.1184,
+      "step": 14360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.187961985216474e-06,
+      "loss": 0.1982,
+      "step": 14365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.18653500385285e-06,
+      "loss": 0.1289,
+      "step": 14370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.185108022489227e-06,
+      "loss": 0.158,
+      "step": 14375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.183681041125603e-06,
+      "loss": 0.0574,
+      "step": 14380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.182254059761979e-06,
+      "loss": 0.1613,
+      "step": 14385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.180827078398357e-06,
+      "loss": 0.0882,
+      "step": 14390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.179400097034734e-06,
+      "loss": 0.285,
+      "step": 14395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.17797311567111e-06,
+      "loss": 0.4831,
+      "step": 14400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1765461343074864e-06,
+      "loss": 0.1271,
+      "step": 14405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.175119152943863e-06,
+      "loss": 0.1416,
+      "step": 14410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.17369217158024e-06,
+      "loss": 0.2117,
+      "step": 14415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.172265190216616e-06,
+      "loss": 0.217,
+      "step": 14420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.170838208852993e-06,
+      "loss": 0.1049,
+      "step": 14425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.169411227489369e-06,
+      "loss": 0.0806,
+      "step": 14430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.167984246125747e-06,
+      "loss": 0.165,
+      "step": 14435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.166557264762123e-06,
+      "loss": 0.2446,
+      "step": 14440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1651302833985e-06,
+      "loss": 0.3714,
+      "step": 14445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.163703302034876e-06,
+      "loss": 0.4461,
+      "step": 14450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.162276320671252e-06,
+      "loss": 0.1162,
+      "step": 14455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1608493393076295e-06,
+      "loss": 0.148,
+      "step": 14460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.159422357944006e-06,
+      "loss": 0.1423,
+      "step": 14465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.157995376580382e-06,
+      "loss": 0.2353,
+      "step": 14470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1565683952167586e-06,
+      "loss": 0.212,
+      "step": 14475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.155141413853135e-06,
+      "loss": 0.1511,
+      "step": 14480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.153714432489513e-06,
+      "loss": 0.1371,
+      "step": 14485
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.152287451125889e-06,
+      "loss": 0.189,
+      "step": 14490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.150860469762266e-06,
+      "loss": 0.2548,
+      "step": 14495
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.149433488398642e-06,
+      "loss": 0.5477,
+      "step": 14500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.148006507035018e-06,
+      "loss": 0.1403,
+      "step": 14505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1465795256713955e-06,
+      "loss": 0.1262,
+      "step": 14510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.145152544307772e-06,
+      "loss": 0.1666,
+      "step": 14515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.143725562944148e-06,
+      "loss": 0.1582,
+      "step": 14520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1422985815805245e-06,
+      "loss": 0.0996,
+      "step": 14525
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1408716002169025e-06,
+      "loss": 0.0793,
+      "step": 14530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.139444618853279e-06,
+      "loss": 0.1115,
+      "step": 14535
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.138017637489655e-06,
+      "loss": 0.1565,
+      "step": 14540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1365906561260315e-06,
+      "loss": 0.3243,
+      "step": 14545
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.135163674762408e-06,
+      "loss": 0.4306,
+      "step": 14550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.133736693398785e-06,
+      "loss": 0.1278,
+      "step": 14555
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.132309712035161e-06,
+      "loss": 0.1131,
+      "step": 14560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.130882730671538e-06,
+      "loss": 0.1418,
+      "step": 14565
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.129455749307914e-06,
+      "loss": 0.195,
+      "step": 14570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.12802876794429e-06,
+      "loss": 0.1337,
+      "step": 14575
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1266017865806685e-06,
+      "loss": 0.1816,
+      "step": 14580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.125174805217045e-06,
+      "loss": 0.0926,
+      "step": 14585
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.123747823853421e-06,
+      "loss": 0.1594,
+      "step": 14590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.1223208424897975e-06,
+      "loss": 0.2272,
+      "step": 14595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.120893861126174e-06,
+      "loss": 0.5619,
+      "step": 14600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.119466879762551e-06,
+      "loss": 0.1702,
+      "step": 14605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.118039898398927e-06,
+      "loss": 0.1968,
+      "step": 14610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.116612917035304e-06,
+      "loss": 0.15,
+      "step": 14615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.11518593567168e-06,
+      "loss": 0.1767,
+      "step": 14620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.113758954308058e-06,
+      "loss": 0.1723,
+      "step": 14625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.112331972944434e-06,
+      "loss": 0.1547,
+      "step": 14630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.110904991580811e-06,
+      "loss": 0.1129,
+      "step": 14635
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.109478010217187e-06,
+      "loss": 0.1855,
+      "step": 14640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.108051028853563e-06,
+      "loss": 0.2278,
+      "step": 14645
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.106624047489941e-06,
+      "loss": 0.4278,
+      "step": 14650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.105197066126317e-06,
+      "loss": 0.1545,
+      "step": 14655
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.103770084762693e-06,
+      "loss": 0.1949,
+      "step": 14660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.10234310339907e-06,
+      "loss": 0.1673,
+      "step": 14665
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.100916122035446e-06,
+      "loss": 0.2014,
+      "step": 14670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.099489140671824e-06,
+      "loss": 0.1868,
+      "step": 14675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0980621593082e-06,
+      "loss": 0.0941,
+      "step": 14680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.096635177944577e-06,
+      "loss": 0.1168,
+      "step": 14685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.095208196580953e-06,
+      "loss": 0.1158,
+      "step": 14690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.093781215217329e-06,
+      "loss": 0.3806,
+      "step": 14695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0923542338537065e-06,
+      "loss": 0.4403,
+      "step": 14700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.090927252490083e-06,
+      "loss": 0.147,
+      "step": 14705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.089500271126459e-06,
+      "loss": 0.1279,
+      "step": 14710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0880732897628355e-06,
+      "loss": 0.1222,
+      "step": 14715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0866463083992136e-06,
+      "loss": 0.2057,
+      "step": 14720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.08521932703559e-06,
+      "loss": 0.1129,
+      "step": 14725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.083792345671966e-06,
+      "loss": 0.2365,
+      "step": 14730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0823653643083426e-06,
+      "loss": 0.1269,
+      "step": 14735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.080938382944719e-06,
+      "loss": 0.1231,
+      "step": 14740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.079511401581096e-06,
+      "loss": 0.2604,
+      "step": 14745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0780844202174724e-06,
+      "loss": 0.4422,
+      "step": 14750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.076657438853849e-06,
+      "loss": 0.174,
+      "step": 14755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.075230457490225e-06,
+      "loss": 0.113,
+      "step": 14760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0738034761266015e-06,
+      "loss": 0.2471,
+      "step": 14765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0723764947629795e-06,
+      "loss": 0.1751,
+      "step": 14770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.070949513399356e-06,
+      "loss": 0.1614,
+      "step": 14775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.069522532035732e-06,
+      "loss": 0.1217,
+      "step": 14780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0680955506721085e-06,
+      "loss": 0.1163,
+      "step": 14785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.066668569308485e-06,
+      "loss": 0.089,
+      "step": 14790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.065241587944862e-06,
+      "loss": 0.1482,
+      "step": 14795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.063814606581238e-06,
+      "loss": 0.4254,
+      "step": 14800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.062387625217615e-06,
+      "loss": 0.148,
+      "step": 14805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.060960643853991e-06,
+      "loss": 0.1652,
+      "step": 14810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.059533662490369e-06,
+      "loss": 0.173,
+      "step": 14815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.058106681126745e-06,
+      "loss": 0.1958,
+      "step": 14820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.056679699763122e-06,
+      "loss": 0.1786,
+      "step": 14825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.055252718399498e-06,
+      "loss": 0.1066,
+      "step": 14830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0538257370358744e-06,
+      "loss": 0.0715,
+      "step": 14835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.052398755672252e-06,
+      "loss": 0.1148,
+      "step": 14840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.050971774308628e-06,
+      "loss": 0.2405,
+      "step": 14845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.049544792945004e-06,
+      "loss": 0.2661,
+      "step": 14850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.048117811581381e-06,
+      "loss": 0.1637,
+      "step": 14855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.046690830217757e-06,
+      "loss": 0.1238,
+      "step": 14860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.045263848854135e-06,
+      "loss": 0.2242,
+      "step": 14865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.043836867490511e-06,
+      "loss": 0.2049,
+      "step": 14870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.042409886126888e-06,
+      "loss": 0.0851,
+      "step": 14875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.040982904763264e-06,
+      "loss": 0.0644,
+      "step": 14880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.03955592339964e-06,
+      "loss": 0.0894,
+      "step": 14885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0381289420360175e-06,
+      "loss": 0.2002,
+      "step": 14890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.036701960672394e-06,
+      "loss": 0.2101,
+      "step": 14895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.03527497930877e-06,
+      "loss": 0.3475,
+      "step": 14900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0338479979451466e-06,
+      "loss": 0.1352,
+      "step": 14905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.032421016581525e-06,
+      "loss": 0.1464,
+      "step": 14910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.030994035217901e-06,
+      "loss": 0.1251,
+      "step": 14915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.029567053854277e-06,
+      "loss": 0.1559,
+      "step": 14920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.028140072490654e-06,
+      "loss": 0.1228,
+      "step": 14925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.02671309112703e-06,
+      "loss": 0.1002,
+      "step": 14930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.025286109763407e-06,
+      "loss": 0.1621,
+      "step": 14935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0238591283997835e-06,
+      "loss": 0.1759,
+      "step": 14940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.02243214703616e-06,
+      "loss": 0.3777,
+      "step": 14945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.021005165672536e-06,
+      "loss": 0.2646,
+      "step": 14950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.019578184308913e-06,
+      "loss": 0.1493,
+      "step": 14955
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.0181512029452905e-06,
+      "loss": 0.1904,
+      "step": 14960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.016724221581667e-06,
+      "loss": 0.1436,
+      "step": 14965
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.015297240218043e-06,
+      "loss": 0.1792,
+      "step": 14970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.0138702588544195e-06,
+      "loss": 0.1413,
+      "step": 14975
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.012443277490796e-06,
+      "loss": 0.0761,
+      "step": 14980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.011016296127173e-06,
+      "loss": 0.1626,
+      "step": 14985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.009589314763549e-06,
+      "loss": 0.1427,
+      "step": 14990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.008162333399927e-06,
+      "loss": 0.1857,
+      "step": 14995
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.006735352036303e-06,
+      "loss": 0.3772,
+      "step": 15000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.00530837067268e-06,
+      "loss": 0.1449,
+      "step": 15005
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.0038813893090564e-06,
+      "loss": 0.1873,
+      "step": 15010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.002454407945433e-06,
+      "loss": 0.1683,
+      "step": 15015
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.001027426581809e-06,
+      "loss": 0.1978,
+      "step": 15020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9996004452181855e-06,
+      "loss": 0.1495,
+      "step": 15025
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.998173463854563e-06,
+      "loss": 0.1911,
+      "step": 15030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.996746482490939e-06,
+      "loss": 0.0522,
+      "step": 15035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.995319501127316e-06,
+      "loss": 0.1913,
+      "step": 15040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9938925197636925e-06,
+      "loss": 0.0795,
+      "step": 15045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.992465538400069e-06,
+      "loss": 0.6079,
+      "step": 15050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.991038557036446e-06,
+      "loss": 0.1595,
+      "step": 15055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.989611575672822e-06,
+      "loss": 0.1979,
+      "step": 15060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.988184594309199e-06,
+      "loss": 0.1023,
+      "step": 15065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.986757612945575e-06,
+      "loss": 0.1596,
+      "step": 15070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.985330631581951e-06,
+      "loss": 0.1979,
+      "step": 15075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9839036502183286e-06,
+      "loss": 0.0847,
+      "step": 15080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.982476668854706e-06,
+      "loss": 0.1253,
+      "step": 15085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.981049687491082e-06,
+      "loss": 0.1647,
+      "step": 15090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9796227061274584e-06,
+      "loss": 0.1517,
+      "step": 15095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.978195724763836e-06,
+      "loss": 0.5746,
+      "step": 15100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.976768743400212e-06,
+      "loss": 0.3669,
+      "step": 15105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.975341762036588e-06,
+      "loss": 0.1328,
+      "step": 15110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.973914780672965e-06,
+      "loss": 0.1376,
+      "step": 15115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.972487799309341e-06,
+      "loss": 0.1686,
+      "step": 15120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.971060817945718e-06,
+      "loss": 0.0865,
+      "step": 15125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.969633836582095e-06,
+      "loss": 0.1565,
+      "step": 15130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.968206855218472e-06,
+      "loss": 0.1184,
+      "step": 15135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.966779873854848e-06,
+      "loss": 0.0771,
+      "step": 15140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.965352892491224e-06,
+      "loss": 0.4435,
+      "step": 15145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9639259111276015e-06,
+      "loss": 0.4307,
+      "step": 15150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.962498929763978e-06,
+      "loss": 0.0955,
+      "step": 15155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.961071948400354e-06,
+      "loss": 0.13,
+      "step": 15160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9596449670367306e-06,
+      "loss": 0.1701,
+      "step": 15165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.958217985673107e-06,
+      "loss": 0.2054,
+      "step": 15170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.956791004309485e-06,
+      "loss": 0.1288,
+      "step": 15175
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.955364022945861e-06,
+      "loss": 0.1721,
+      "step": 15180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.953937041582238e-06,
+      "loss": 0.0594,
+      "step": 15185
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.952510060218614e-06,
+      "loss": 0.2867,
+      "step": 15190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.951083078854991e-06,
+      "loss": 0.4329,
+      "step": 15195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9496560974913675e-06,
+      "loss": 0.4032,
+      "step": 15200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.948229116127744e-06,
+      "loss": 0.1789,
+      "step": 15205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.94680213476412e-06,
+      "loss": 0.1517,
+      "step": 15210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9453751534004965e-06,
+      "loss": 0.1671,
+      "step": 15215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9439481720368745e-06,
+      "loss": 0.1864,
+      "step": 15220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.942521190673251e-06,
+      "loss": 0.1615,
+      "step": 15225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.941094209309627e-06,
+      "loss": 0.054,
+      "step": 15230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9396672279460035e-06,
+      "loss": 0.1439,
+      "step": 15235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.93824024658238e-06,
+      "loss": 0.1283,
+      "step": 15240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.936813265218757e-06,
+      "loss": 0.2536,
+      "step": 15245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.935386283855133e-06,
+      "loss": 0.2707,
+      "step": 15250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.93395930249151e-06,
+      "loss": 0.2927,
+      "step": 15255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.932532321127886e-06,
+      "loss": 0.1679,
+      "step": 15260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.931105339764262e-06,
+      "loss": 0.2068,
+      "step": 15265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9296783584006405e-06,
+      "loss": 0.1637,
+      "step": 15270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.928251377037017e-06,
+      "loss": 0.0461,
+      "step": 15275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.926824395673393e-06,
+      "loss": 0.0905,
+      "step": 15280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.9253974143097695e-06,
+      "loss": 0.2088,
+      "step": 15285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.923970432946147e-06,
+      "loss": 0.1959,
+      "step": 15290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.922543451582523e-06,
+      "loss": 0.3654,
+      "step": 15295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.921116470218899e-06,
+      "loss": 0.4821,
+      "step": 15300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.919689488855276e-06,
+      "loss": 0.1287,
+      "step": 15305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.918262507491652e-06,
+      "loss": 0.1532,
+      "step": 15310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.91683552612803e-06,
+      "loss": 0.1781,
+      "step": 15315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.915408544764406e-06,
+      "loss": 0.1431,
+      "step": 15320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.913981563400783e-06,
+      "loss": 0.1377,
+      "step": 15325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.912554582037159e-06,
+      "loss": 0.1645,
+      "step": 15330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.911127600673535e-06,
+      "loss": 0.119,
+      "step": 15335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.909700619309913e-06,
+      "loss": 0.187,
+      "step": 15340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.908273637946289e-06,
+      "loss": 0.2611,
+      "step": 15345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.906846656582665e-06,
+      "loss": 0.4396,
+      "step": 15350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.905419675219042e-06,
+      "loss": 0.1585,
+      "step": 15355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.903992693855418e-06,
+      "loss": 0.1651,
+      "step": 15360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.902565712491796e-06,
+      "loss": 0.266,
+      "step": 15365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.901138731128172e-06,
+      "loss": 0.1742,
+      "step": 15370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.899711749764549e-06,
+      "loss": 0.1705,
+      "step": 15375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.898284768400925e-06,
+      "loss": 0.0397,
+      "step": 15380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.896857787037302e-06,
+      "loss": 0.2106,
+      "step": 15385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8954308056736785e-06,
+      "loss": 0.1104,
+      "step": 15390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.894003824310055e-06,
+      "loss": 0.25,
+      "step": 15395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.892576842946431e-06,
+      "loss": 0.1658,
+      "step": 15400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8911498615828075e-06,
+      "loss": 0.2023,
+      "step": 15405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8897228802191856e-06,
+      "loss": 0.1889,
+      "step": 15410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.888295898855562e-06,
+      "loss": 0.1971,
+      "step": 15415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.886868917491938e-06,
+      "loss": 0.186,
+      "step": 15420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8854419361283146e-06,
+      "loss": 0.1272,
+      "step": 15425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.884014954764691e-06,
+      "loss": 0.0889,
+      "step": 15430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.882587973401068e-06,
+      "loss": 0.0736,
+      "step": 15435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8811609920374444e-06,
+      "loss": 0.2415,
+      "step": 15440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.879734010673821e-06,
+      "loss": 0.2244,
+      "step": 15445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.878307029310197e-06,
+      "loss": 0.3602,
+      "step": 15450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.876880047946575e-06,
+      "loss": 0.1511,
+      "step": 15455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8754530665829515e-06,
+      "loss": 0.1409,
+      "step": 15460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.874026085219328e-06,
+      "loss": 0.1107,
+      "step": 15465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.872599103855704e-06,
+      "loss": 0.1823,
+      "step": 15470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8711721224920805e-06,
+      "loss": 0.0253,
+      "step": 15475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.869745141128458e-06,
+      "loss": 0.2005,
+      "step": 15480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.868318159764834e-06,
+      "loss": 0.2155,
+      "step": 15485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.86689117840121e-06,
+      "loss": 0.1663,
+      "step": 15490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.865464197037587e-06,
+      "loss": 0.1827,
+      "step": 15495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.864037215673963e-06,
+      "loss": 0.5004,
+      "step": 15500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.862610234310341e-06,
+      "loss": 0.1969,
+      "step": 15505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.861183252946717e-06,
+      "loss": 0.167,
+      "step": 15510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.859756271583094e-06,
+      "loss": 0.1858,
+      "step": 15515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.85832929021947e-06,
+      "loss": 0.1566,
+      "step": 15520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8569023088558464e-06,
+      "loss": 0.1181,
+      "step": 15525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.855475327492224e-06,
+      "loss": 0.139,
+      "step": 15530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8540483461286e-06,
+      "loss": 0.1463,
+      "step": 15535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.852621364764976e-06,
+      "loss": 0.1682,
+      "step": 15540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.851194383401353e-06,
+      "loss": 0.3414,
+      "step": 15545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.849767402037731e-06,
+      "loss": 0.4187,
+      "step": 15550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.848340420674107e-06,
+      "loss": 0.1858,
+      "step": 15555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.846913439310483e-06,
+      "loss": 0.1252,
+      "step": 15560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.84548645794686e-06,
+      "loss": 0.1411,
+      "step": 15565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.844059476583236e-06,
+      "loss": 0.1414,
+      "step": 15570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.842632495219613e-06,
+      "loss": 0.1748,
+      "step": 15575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8412055138559895e-06,
+      "loss": 0.0769,
+      "step": 15580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.839778532492366e-06,
+      "loss": 0.0694,
+      "step": 15585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.838351551128742e-06,
+      "loss": 0.194,
+      "step": 15590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8369245697651186e-06,
+      "loss": 0.2519,
+      "step": 15595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.835497588401497e-06,
+      "loss": 0.288,
+      "step": 15600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.834070607037873e-06,
+      "loss": 0.1731,
+      "step": 15605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.832643625674249e-06,
+      "loss": 0.1078,
+      "step": 15610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.831216644310626e-06,
+      "loss": 0.1198,
+      "step": 15615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.829789662947002e-06,
+      "loss": 0.2483,
+      "step": 15620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.828362681583379e-06,
+      "loss": 0.0385,
+      "step": 15625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8269357002197555e-06,
+      "loss": 0.1998,
+      "step": 15630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.825508718856132e-06,
+      "loss": 0.3765,
+      "step": 15635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.824081737492508e-06,
+      "loss": 0.1923,
+      "step": 15640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.822654756128886e-06,
+      "loss": 0.1936,
+      "step": 15645
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8212277747652625e-06,
+      "loss": 0.406,
+      "step": 15650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.819800793401639e-06,
+      "loss": 0.1442,
+      "step": 15655
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.818373812038015e-06,
+      "loss": 0.105,
+      "step": 15660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.8169468306743915e-06,
+      "loss": 0.1852,
+      "step": 15665
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.815519849310769e-06,
+      "loss": 0.1685,
+      "step": 15670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.814092867947145e-06,
+      "loss": 0.2012,
+      "step": 15675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.812665886583521e-06,
+      "loss": 0.108,
+      "step": 15680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.811238905219898e-06,
+      "loss": 0.1475,
+      "step": 15685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.809811923856274e-06,
+      "loss": 0.0816,
+      "step": 15690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.808384942492652e-06,
+      "loss": 0.3827,
+      "step": 15695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8069579611290284e-06,
+      "loss": 0.5066,
+      "step": 15700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.805530979765405e-06,
+      "loss": 0.1954,
+      "step": 15705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.804103998401781e-06,
+      "loss": 0.1015,
+      "step": 15710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8026770170381575e-06,
+      "loss": 0.1712,
+      "step": 15715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.801250035674535e-06,
+      "loss": 0.1948,
+      "step": 15720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.799823054310911e-06,
+      "loss": 0.19,
+      "step": 15725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.798396072947287e-06,
+      "loss": 0.072,
+      "step": 15730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.796969091583664e-06,
+      "loss": 0.1734,
+      "step": 15735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.795542110220042e-06,
+      "loss": 0.0767,
+      "step": 15740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.794115128856418e-06,
+      "loss": 0.1765,
+      "step": 15745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.792688147492794e-06,
+      "loss": 0.4795,
+      "step": 15750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.791261166129171e-06,
+      "loss": 0.174,
+      "step": 15755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.789834184765547e-06,
+      "loss": 0.1905,
+      "step": 15760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.788407203401924e-06,
+      "loss": 0.1317,
+      "step": 15765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7869802220383006e-06,
+      "loss": 0.1573,
+      "step": 15770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.785553240674677e-06,
+      "loss": 0.1188,
+      "step": 15775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.784126259311053e-06,
+      "loss": 0.1563,
+      "step": 15780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7826992779474304e-06,
+      "loss": 0.0544,
+      "step": 15785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.781272296583808e-06,
+      "loss": 0.2061,
+      "step": 15790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.779845315220184e-06,
+      "loss": 0.2543,
+      "step": 15795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.77841833385656e-06,
+      "loss": 0.4642,
+      "step": 15800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.776991352492937e-06,
+      "loss": 0.2213,
+      "step": 15805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.775564371129313e-06,
+      "loss": 0.1508,
+      "step": 15810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.77413738976569e-06,
+      "loss": 0.1495,
+      "step": 15815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7727104084020665e-06,
+      "loss": 0.1711,
+      "step": 15820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.771283427038443e-06,
+      "loss": 0.063,
+      "step": 15825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.76985644567482e-06,
+      "loss": 0.0513,
+      "step": 15830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.768429464311197e-06,
+      "loss": 0.3326,
+      "step": 15835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7670024829475735e-06,
+      "loss": 0.1372,
+      "step": 15840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.76557550158395e-06,
+      "loss": 0.2871,
+      "step": 15845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.764148520220326e-06,
+      "loss": 0.5247,
+      "step": 15850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7627215388567026e-06,
+      "loss": 0.1836,
+      "step": 15855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.76129455749308e-06,
+      "loss": 0.142,
+      "step": 15860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.759867576129456e-06,
+      "loss": 0.2079,
+      "step": 15865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7584405947658324e-06,
+      "loss": 0.1359,
+      "step": 15870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.75701361340221e-06,
+      "loss": 0.1529,
+      "step": 15875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.755586632038586e-06,
+      "loss": 0.1607,
+      "step": 15880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.754159650674963e-06,
+      "loss": 0.179,
+      "step": 15885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7527326693113395e-06,
+      "loss": 0.1321,
+      "step": 15890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.751305687947716e-06,
+      "loss": 0.2263,
+      "step": 15895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.749878706584092e-06,
+      "loss": 0.2971,
+      "step": 15900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7484517252204685e-06,
+      "loss": 0.1456,
+      "step": 15905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.747024743856846e-06,
+      "loss": 0.2264,
+      "step": 15910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.745597762493222e-06,
+      "loss": 0.1319,
+      "step": 15915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.744170781129599e-06,
+      "loss": 0.1284,
+      "step": 15920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7427437997659755e-06,
+      "loss": 0.1671,
+      "step": 15925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.741316818402353e-06,
+      "loss": 0.1931,
+      "step": 15930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.739889837038729e-06,
+      "loss": 0.3126,
+      "step": 15935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.738462855675105e-06,
+      "loss": 0.3046,
+      "step": 15940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.737035874311482e-06,
+      "loss": 0.3687,
+      "step": 15945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.735608892947858e-06,
+      "loss": 0.4295,
+      "step": 15950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.734181911584235e-06,
+      "loss": 0.1389,
+      "step": 15955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.732754930220612e-06,
+      "loss": 0.2207,
+      "step": 15960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.731327948856989e-06,
+      "loss": 0.1332,
+      "step": 15965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.729900967493365e-06,
+      "loss": 0.1509,
+      "step": 15970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.7284739861297415e-06,
+      "loss": 0.1135,
+      "step": 15975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.727047004766119e-06,
+      "loss": 0.138,
+      "step": 15980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.725620023402495e-06,
+      "loss": 0.1944,
+      "step": 15985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.724193042038871e-06,
+      "loss": 0.1586,
+      "step": 15990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.722766060675248e-06,
+      "loss": 0.1768,
+      "step": 15995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.721339079311624e-06,
+      "loss": 0.5374,
+      "step": 16000
+    },
+    {
+      "epoch": 0.44,
+      "eval_loss": 0.0686265304684639,
+      "eval_runtime": 1948.2008,
+      "eval_samples_per_second": 8.222,
+      "eval_steps_per_second": 2.056,
+      "eval_wer": 0.185971871619185,
+      "step": 16000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.719912097948002e-06,
+      "loss": 0.1114,
+      "step": 16005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.718485116584378e-06,
+      "loss": 0.2,
+      "step": 16010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.717058135220755e-06,
+      "loss": 0.1075,
+      "step": 16015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.715631153857131e-06,
+      "loss": 0.1529,
+      "step": 16020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.714204172493508e-06,
+      "loss": 0.0363,
+      "step": 16025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.712777191129885e-06,
+      "loss": 0.2312,
+      "step": 16030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.711350209766261e-06,
+      "loss": 0.056,
+      "step": 16035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.709923228402637e-06,
+      "loss": 0.1475,
+      "step": 16040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.708496247039014e-06,
+      "loss": 0.2889,
+      "step": 16045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.707069265675392e-06,
+      "loss": 0.5634,
+      "step": 16050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.705642284311768e-06,
+      "loss": 0.1819,
+      "step": 16055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.704215302948144e-06,
+      "loss": 0.1169,
+      "step": 16060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.702788321584521e-06,
+      "loss": 0.2161,
+      "step": 16065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.701361340220897e-06,
+      "loss": 0.2054,
+      "step": 16070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.699934358857274e-06,
+      "loss": 0.1147,
+      "step": 16075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6985073774936505e-06,
+      "loss": 0.1058,
+      "step": 16080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.697080396130027e-06,
+      "loss": 0.1971,
+      "step": 16085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.695653414766403e-06,
+      "loss": 0.1929,
+      "step": 16090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6942264334027795e-06,
+      "loss": 0.2481,
+      "step": 16095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6927994520391576e-06,
+      "loss": 0.6187,
+      "step": 16100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.691372470675534e-06,
+      "loss": 0.1115,
+      "step": 16105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.68994548931191e-06,
+      "loss": 0.1638,
+      "step": 16110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6885185079482866e-06,
+      "loss": 0.1671,
+      "step": 16115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.687091526584664e-06,
+      "loss": 0.1476,
+      "step": 16120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.68566454522104e-06,
+      "loss": 0.1183,
+      "step": 16125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6842375638574164e-06,
+      "loss": 0.0607,
+      "step": 16130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.682810582493793e-06,
+      "loss": 0.097,
+      "step": 16135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.681383601130169e-06,
+      "loss": 0.1784,
+      "step": 16140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.679956619766547e-06,
+      "loss": 0.2754,
+      "step": 16145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6785296384029235e-06,
+      "loss": 0.4347,
+      "step": 16150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6771026570393e-06,
+      "loss": 0.2081,
+      "step": 16155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.675675675675676e-06,
+      "loss": 0.1189,
+      "step": 16160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6742486943120525e-06,
+      "loss": 0.1428,
+      "step": 16165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.67282171294843e-06,
+      "loss": 0.1191,
+      "step": 16170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.671394731584806e-06,
+      "loss": 0.1557,
+      "step": 16175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.669967750221182e-06,
+      "loss": 0.0787,
+      "step": 16180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.668540768857559e-06,
+      "loss": 0.0814,
+      "step": 16185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.667113787493935e-06,
+      "loss": 0.4326,
+      "step": 16190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.665686806130313e-06,
+      "loss": 0.3891,
+      "step": 16195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.664259824766689e-06,
+      "loss": 0.3938,
+      "step": 16200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.662832843403066e-06,
+      "loss": 0.0934,
+      "step": 16205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.661405862039442e-06,
+      "loss": 0.1079,
+      "step": 16210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.659978880675819e-06,
+      "loss": 0.1906,
+      "step": 16215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.658551899312196e-06,
+      "loss": 0.1347,
+      "step": 16220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.657124917948572e-06,
+      "loss": 0.1073,
+      "step": 16225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.655697936584948e-06,
+      "loss": 0.0906,
+      "step": 16230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.654270955221325e-06,
+      "loss": 0.0965,
+      "step": 16235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.652843973857703e-06,
+      "loss": 0.1537,
+      "step": 16240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.651416992494079e-06,
+      "loss": 0.1443,
+      "step": 16245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.649990011130455e-06,
+      "loss": 0.3198,
+      "step": 16250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.648563029766832e-06,
+      "loss": 0.1844,
+      "step": 16255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.647136048403208e-06,
+      "loss": 0.1468,
+      "step": 16260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.645709067039585e-06,
+      "loss": 0.2023,
+      "step": 16265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6442820856759615e-06,
+      "loss": 0.1989,
+      "step": 16270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.642855104312338e-06,
+      "loss": 0.0832,
+      "step": 16275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.641428122948714e-06,
+      "loss": 0.0124,
+      "step": 16280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6400011415850906e-06,
+      "loss": 0.1311,
+      "step": 16285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.638574160221469e-06,
+      "loss": 0.1941,
+      "step": 16290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.637147178857845e-06,
+      "loss": 0.0791,
+      "step": 16295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.635720197494221e-06,
+      "loss": 0.3192,
+      "step": 16300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.634293216130598e-06,
+      "loss": 0.1634,
+      "step": 16305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.632866234766975e-06,
+      "loss": 0.2161,
+      "step": 16310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.631439253403351e-06,
+      "loss": 0.1839,
+      "step": 16315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6300122720397275e-06,
+      "loss": 0.1424,
+      "step": 16320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.628585290676104e-06,
+      "loss": 0.1061,
+      "step": 16325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.62715830931248e-06,
+      "loss": 0.0574,
+      "step": 16330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.625731327948858e-06,
+      "loss": 0.1294,
+      "step": 16335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6243043465852345e-06,
+      "loss": 0.1222,
+      "step": 16340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.622877365221611e-06,
+      "loss": 0.2917,
+      "step": 16345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.621450383857987e-06,
+      "loss": 0.2507,
+      "step": 16350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6200234024943635e-06,
+      "loss": 0.1206,
+      "step": 16355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.618596421130741e-06,
+      "loss": 0.1327,
+      "step": 16360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.617169439767117e-06,
+      "loss": 0.1302,
+      "step": 16365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.615742458403493e-06,
+      "loss": 0.1795,
+      "step": 16370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.61431547703987e-06,
+      "loss": 0.1121,
+      "step": 16375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.612888495676246e-06,
+      "loss": 0.1555,
+      "step": 16380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.611461514312624e-06,
+      "loss": 0.0623,
+      "step": 16385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.6100345329490004e-06,
+      "loss": 0.1516,
+      "step": 16390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.608607551585377e-06,
+      "loss": 0.2326,
+      "step": 16395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.607180570221753e-06,
+      "loss": 0.4471,
+      "step": 16400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.60575358885813e-06,
+      "loss": 0.149,
+      "step": 16405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.604326607494507e-06,
+      "loss": 0.1888,
+      "step": 16410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.602899626130883e-06,
+      "loss": 0.2294,
+      "step": 16415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.601472644767259e-06,
+      "loss": 0.1127,
+      "step": 16420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.600045663403636e-06,
+      "loss": 0.0707,
+      "step": 16425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.598618682040014e-06,
+      "loss": 0.0558,
+      "step": 16430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.59719170067639e-06,
+      "loss": 0.134,
+      "step": 16435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.595764719312766e-06,
+      "loss": 0.1683,
+      "step": 16440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.594337737949143e-06,
+      "loss": 0.3001,
+      "step": 16445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.592910756585519e-06,
+      "loss": 0.445,
+      "step": 16450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.591483775221896e-06,
+      "loss": 0.1413,
+      "step": 16455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5900567938582726e-06,
+      "loss": 0.0874,
+      "step": 16460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.588629812494649e-06,
+      "loss": 0.1695,
+      "step": 16465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.587202831131025e-06,
+      "loss": 0.1552,
+      "step": 16470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.585775849767402e-06,
+      "loss": 0.171,
+      "step": 16475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.58434886840378e-06,
+      "loss": 0.0791,
+      "step": 16480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.582921887040156e-06,
+      "loss": 0.1119,
+      "step": 16485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.581494905676532e-06,
+      "loss": 0.2402,
+      "step": 16490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.580067924312909e-06,
+      "loss": 0.2793,
+      "step": 16495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.578640942949286e-06,
+      "loss": 0.5294,
+      "step": 16500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.577213961585662e-06,
+      "loss": 0.1235,
+      "step": 16505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5757869802220385e-06,
+      "loss": 0.1802,
+      "step": 16510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.574359998858415e-06,
+      "loss": 0.112,
+      "step": 16515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.572933017494791e-06,
+      "loss": 0.1402,
+      "step": 16520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.571506036131169e-06,
+      "loss": 0.2075,
+      "step": 16525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.57036445104027e-06,
+      "loss": 0.9067,
+      "step": 16530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.568937469676646e-06,
+      "loss": 0.0746,
+      "step": 16535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.567510488313022e-06,
+      "loss": 0.3122,
+      "step": 16540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5660835069494e-06,
+      "loss": 0.3623,
+      "step": 16545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.564656525585777e-06,
+      "loss": 0.4181,
+      "step": 16550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.563229544222153e-06,
+      "loss": 0.2052,
+      "step": 16555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.561802562858529e-06,
+      "loss": 0.1638,
+      "step": 16560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.560375581494906e-06,
+      "loss": 0.1692,
+      "step": 16565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.558948600131283e-06,
+      "loss": 0.1753,
+      "step": 16570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.557521618767659e-06,
+      "loss": 0.1365,
+      "step": 16575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5560946374040355e-06,
+      "loss": 0.11,
+      "step": 16580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.554667656040412e-06,
+      "loss": 0.0666,
+      "step": 16585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.553240674676788e-06,
+      "loss": 0.1271,
+      "step": 16590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.551813693313166e-06,
+      "loss": 0.2357,
+      "step": 16595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.550386711949543e-06,
+      "loss": 0.5136,
+      "step": 16600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.548959730585919e-06,
+      "loss": 0.0782,
+      "step": 16605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.547532749222295e-06,
+      "loss": 0.1825,
+      "step": 16610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5461057678586724e-06,
+      "loss": 0.2018,
+      "step": 16615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.544678786495049e-06,
+      "loss": 0.1425,
+      "step": 16620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.543251805131425e-06,
+      "loss": 0.0838,
+      "step": 16625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5418248237678015e-06,
+      "loss": 0.1659,
+      "step": 16630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.540397842404178e-06,
+      "loss": 0.112,
+      "step": 16635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.538970861040556e-06,
+      "loss": 0.3056,
+      "step": 16640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.537543879676932e-06,
+      "loss": 0.303,
+      "step": 16645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5361168983133085e-06,
+      "loss": 0.4717,
+      "step": 16650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.534689916949685e-06,
+      "loss": 0.2241,
+      "step": 16655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.533262935586061e-06,
+      "loss": 0.1074,
+      "step": 16660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.531835954222438e-06,
+      "loss": 0.1684,
+      "step": 16665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.530408972858815e-06,
+      "loss": 0.1234,
+      "step": 16670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.528981991495191e-06,
+      "loss": 0.1283,
+      "step": 16675
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.527555010131567e-06,
+      "loss": 0.0778,
+      "step": 16680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.526128028767945e-06,
+      "loss": 0.1442,
+      "step": 16685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.524701047404322e-06,
+      "loss": 0.1708,
+      "step": 16690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.523274066040698e-06,
+      "loss": 0.3219,
+      "step": 16695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5218470846770744e-06,
+      "loss": 0.3452,
+      "step": 16700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.520420103313451e-06,
+      "loss": 0.1205,
+      "step": 16705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.518993121949828e-06,
+      "loss": 0.1505,
+      "step": 16710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.517566140586204e-06,
+      "loss": 0.1963,
+      "step": 16715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.516139159222581e-06,
+      "loss": 0.1394,
+      "step": 16720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.514712177858958e-06,
+      "loss": 0.1005,
+      "step": 16725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.513285196495334e-06,
+      "loss": 0.0825,
+      "step": 16730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.511858215131711e-06,
+      "loss": 0.1205,
+      "step": 16735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.510431233768088e-06,
+      "loss": 0.2447,
+      "step": 16740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.509004252404464e-06,
+      "loss": 0.152,
+      "step": 16745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.50757727104084e-06,
+      "loss": 0.3954,
+      "step": 16750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.506150289677217e-06,
+      "loss": 0.1702,
+      "step": 16755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.504723308313594e-06,
+      "loss": 0.1558,
+      "step": 16760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.50329632694997e-06,
+      "loss": 0.1411,
+      "step": 16765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.501869345586347e-06,
+      "loss": 0.3309,
+      "step": 16770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.500442364222724e-06,
+      "loss": 0.1496,
+      "step": 16775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.499015382859101e-06,
+      "loss": 0.2113,
+      "step": 16780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.497588401495477e-06,
+      "loss": 0.1642,
+      "step": 16785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.496161420131854e-06,
+      "loss": 0.0975,
+      "step": 16790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.49473443876823e-06,
+      "loss": 0.1134,
+      "step": 16795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.493307457404606e-06,
+      "loss": 0.3543,
+      "step": 16800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4918804760409835e-06,
+      "loss": 0.129,
+      "step": 16805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.49045349467736e-06,
+      "loss": 0.1279,
+      "step": 16810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.489026513313737e-06,
+      "loss": 0.1819,
+      "step": 16815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.487599531950113e-06,
+      "loss": 0.1646,
+      "step": 16820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.48617255058649e-06,
+      "loss": 0.1626,
+      "step": 16825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.484745569222867e-06,
+      "loss": 0.0615,
+      "step": 16830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.483318587859243e-06,
+      "loss": 0.0538,
+      "step": 16835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4818916064956195e-06,
+      "loss": 0.2278,
+      "step": 16840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.480464625131996e-06,
+      "loss": 0.2573,
+      "step": 16845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.479037643768372e-06,
+      "loss": 0.3704,
+      "step": 16850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.477610662404749e-06,
+      "loss": 0.2063,
+      "step": 16855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.476183681041127e-06,
+      "loss": 0.1791,
+      "step": 16860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.474756699677503e-06,
+      "loss": 0.1397,
+      "step": 16865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.473329718313879e-06,
+      "loss": 0.2235,
+      "step": 16870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4719027369502565e-06,
+      "loss": 0.1106,
+      "step": 16875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.470475755586633e-06,
+      "loss": 0.0626,
+      "step": 16880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.469048774223009e-06,
+      "loss": 0.0783,
+      "step": 16885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4676217928593855e-06,
+      "loss": 0.1718,
+      "step": 16890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.466194811495762e-06,
+      "loss": 0.319,
+      "step": 16895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.464767830132139e-06,
+      "loss": 0.4389,
+      "step": 16900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.463340848768516e-06,
+      "loss": 0.1144,
+      "step": 16905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4619138674048925e-06,
+      "loss": 0.1494,
+      "step": 16910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.460486886041269e-06,
+      "loss": 0.1487,
+      "step": 16915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.459059904677645e-06,
+      "loss": 0.1755,
+      "step": 16920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.457632923314022e-06,
+      "loss": 0.1109,
+      "step": 16925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.456205941950399e-06,
+      "loss": 0.0623,
+      "step": 16930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.454778960586775e-06,
+      "loss": 0.059,
+      "step": 16935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.453351979223151e-06,
+      "loss": 0.1484,
+      "step": 16940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.451924997859528e-06,
+      "loss": 0.2676,
+      "step": 16945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.450498016495906e-06,
+      "loss": 0.4759,
+      "step": 16950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.449071035132282e-06,
+      "loss": 0.1805,
+      "step": 16955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4476440537686584e-06,
+      "loss": 0.1408,
+      "step": 16960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.446217072405035e-06,
+      "loss": 0.1618,
+      "step": 16965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.444790091041412e-06,
+      "loss": 0.2413,
+      "step": 16970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.443363109677788e-06,
+      "loss": 0.2339,
+      "step": 16975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.441936128314165e-06,
+      "loss": 0.1131,
+      "step": 16980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.440509146950541e-06,
+      "loss": 0.1233,
+      "step": 16985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.439082165586917e-06,
+      "loss": 0.2213,
+      "step": 16990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.437655184223295e-06,
+      "loss": 0.2989,
+      "step": 16995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.436228202859672e-06,
+      "loss": 0.2412,
+      "step": 17000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.434801221496048e-06,
+      "loss": 0.1614,
+      "step": 17005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.433374240132424e-06,
+      "loss": 0.2254,
+      "step": 17010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.431947258768801e-06,
+      "loss": 0.1363,
+      "step": 17015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.430520277405178e-06,
+      "loss": 0.1185,
+      "step": 17020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.429093296041554e-06,
+      "loss": 0.1609,
+      "step": 17025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4276663146779306e-06,
+      "loss": 0.0382,
+      "step": 17030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.426239333314307e-06,
+      "loss": 0.0805,
+      "step": 17035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.424812351950683e-06,
+      "loss": 0.2417,
+      "step": 17040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.423385370587061e-06,
+      "loss": 0.1484,
+      "step": 17045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.421958389223438e-06,
+      "loss": 0.3142,
+      "step": 17050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.420531407859814e-06,
+      "loss": 0.1211,
+      "step": 17055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.41910442649619e-06,
+      "loss": 0.2209,
+      "step": 17060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4176774451325675e-06,
+      "loss": 0.1553,
+      "step": 17065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.416250463768944e-06,
+      "loss": 0.1036,
+      "step": 17070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.41482348240532e-06,
+      "loss": 0.1512,
+      "step": 17075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4133965010416965e-06,
+      "loss": 0.1245,
+      "step": 17080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.411969519678073e-06,
+      "loss": 0.1974,
+      "step": 17085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.410542538314451e-06,
+      "loss": 0.0876,
+      "step": 17090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.409115556950827e-06,
+      "loss": 0.2816,
+      "step": 17095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4076885755872036e-06,
+      "loss": 0.5595,
+      "step": 17100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.40626159422358e-06,
+      "loss": 0.1111,
+      "step": 17105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.404834612859956e-06,
+      "loss": 0.1536,
+      "step": 17110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.403407631496333e-06,
+      "loss": 0.1523,
+      "step": 17115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.40198065013271e-06,
+      "loss": 0.1873,
+      "step": 17120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.400553668769086e-06,
+      "loss": 0.116,
+      "step": 17125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3991266874054624e-06,
+      "loss": 0.2162,
+      "step": 17130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.397699706041839e-06,
+      "loss": 0.1174,
+      "step": 17135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.396272724678217e-06,
+      "loss": 0.1295,
+      "step": 17140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.394845743314593e-06,
+      "loss": 0.4878,
+      "step": 17145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3934187619509695e-06,
+      "loss": 0.4039,
+      "step": 17150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.391991780587346e-06,
+      "loss": 0.1392,
+      "step": 17155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.390564799223723e-06,
+      "loss": 0.1637,
+      "step": 17160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.389137817860099e-06,
+      "loss": 0.2175,
+      "step": 17165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.387710836496476e-06,
+      "loss": 0.184,
+      "step": 17170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.386283855132852e-06,
+      "loss": 0.1522,
+      "step": 17175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.384856873769228e-06,
+      "loss": 0.0885,
+      "step": 17180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.383429892405606e-06,
+      "loss": 0.1185,
+      "step": 17185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.382002911041983e-06,
+      "loss": 0.1916,
+      "step": 17190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.380575929678359e-06,
+      "loss": 0.2418,
+      "step": 17195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.379148948314735e-06,
+      "loss": 0.4827,
+      "step": 17200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.377721966951112e-06,
+      "loss": 0.1402,
+      "step": 17205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.376294985587489e-06,
+      "loss": 0.2107,
+      "step": 17210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.374868004223865e-06,
+      "loss": 0.1448,
+      "step": 17215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.373441022860242e-06,
+      "loss": 0.2342,
+      "step": 17220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.372014041496618e-06,
+      "loss": 0.0894,
+      "step": 17225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.370587060132994e-06,
+      "loss": 0.187,
+      "step": 17230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.369160078769372e-06,
+      "loss": 0.1241,
+      "step": 17235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.367733097405749e-06,
+      "loss": 0.1505,
+      "step": 17240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.366306116042125e-06,
+      "loss": 0.3462,
+      "step": 17245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.364879134678501e-06,
+      "loss": 0.346,
+      "step": 17250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3634521533148785e-06,
+      "loss": 0.2012,
+      "step": 17255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.362025171951255e-06,
+      "loss": 0.1905,
+      "step": 17260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.360598190587631e-06,
+      "loss": 0.1316,
+      "step": 17265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3591712092240075e-06,
+      "loss": 0.1953,
+      "step": 17270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.357744227860384e-06,
+      "loss": 0.1508,
+      "step": 17275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.356317246496762e-06,
+      "loss": 0.0538,
+      "step": 17280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.354890265133138e-06,
+      "loss": 0.1284,
+      "step": 17285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.353463283769515e-06,
+      "loss": 0.2855,
+      "step": 17290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.352036302405891e-06,
+      "loss": 0.215,
+      "step": 17295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.350609321042267e-06,
+      "loss": 0.2805,
+      "step": 17300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3491823396786444e-06,
+      "loss": 0.2176,
+      "step": 17305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.347755358315021e-06,
+      "loss": 0.1513,
+      "step": 17310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.346328376951397e-06,
+      "loss": 0.1399,
+      "step": 17315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3449013955877735e-06,
+      "loss": 0.1568,
+      "step": 17320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.34347441422415e-06,
+      "loss": 0.1231,
+      "step": 17325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.342047432860528e-06,
+      "loss": 0.0907,
+      "step": 17330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.340620451496904e-06,
+      "loss": 0.0697,
+      "step": 17335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3391934701332805e-06,
+      "loss": 0.1351,
+      "step": 17340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.337766488769657e-06,
+      "loss": 0.3501,
+      "step": 17345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.336339507406034e-06,
+      "loss": 0.3423,
+      "step": 17350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.33491252604241e-06,
+      "loss": 0.1806,
+      "step": 17355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.333485544678787e-06,
+      "loss": 0.2168,
+      "step": 17360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.332058563315163e-06,
+      "loss": 0.2062,
+      "step": 17365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.330631581951539e-06,
+      "loss": 0.1623,
+      "step": 17370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.329204600587917e-06,
+      "loss": 0.1093,
+      "step": 17375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.327777619224294e-06,
+      "loss": 0.0914,
+      "step": 17380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.32635063786067e-06,
+      "loss": 0.0801,
+      "step": 17385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3249236564970464e-06,
+      "loss": 0.18,
+      "step": 17390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.323496675133423e-06,
+      "loss": 0.3545,
+      "step": 17395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3220696937698e-06,
+      "loss": 0.2112,
+      "step": 17400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.320642712406176e-06,
+      "loss": 0.1072,
+      "step": 17405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.319215731042553e-06,
+      "loss": 0.1603,
+      "step": 17410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.317788749678929e-06,
+      "loss": 0.1775,
+      "step": 17415
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.316361768315305e-06,
+      "loss": 0.1867,
+      "step": 17420
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.314934786951683e-06,
+      "loss": 0.1381,
+      "step": 17425
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.31350780558806e-06,
+      "loss": 0.1061,
+      "step": 17430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.312080824224436e-06,
+      "loss": 0.1987,
+      "step": 17435
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.310653842860812e-06,
+      "loss": 0.2463,
+      "step": 17440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3092268614971896e-06,
+      "loss": 0.2434,
+      "step": 17445
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.307799880133566e-06,
+      "loss": 0.406,
+      "step": 17450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.306372898769942e-06,
+      "loss": 0.1069,
+      "step": 17455
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3049459174063186e-06,
+      "loss": 0.1613,
+      "step": 17460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.303518936042695e-06,
+      "loss": 0.2305,
+      "step": 17465
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.302091954679073e-06,
+      "loss": 0.2166,
+      "step": 17470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.300664973315449e-06,
+      "loss": 0.1284,
+      "step": 17475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.299237991951826e-06,
+      "loss": 0.0405,
+      "step": 17480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.297811010588202e-06,
+      "loss": 0.1256,
+      "step": 17485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.296384029224578e-06,
+      "loss": 0.1613,
+      "step": 17490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2949570478609555e-06,
+      "loss": 0.1741,
+      "step": 17495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.293530066497332e-06,
+      "loss": 0.4977,
+      "step": 17500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.292103085133708e-06,
+      "loss": 0.1077,
+      "step": 17505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2906761037700845e-06,
+      "loss": 0.1564,
+      "step": 17510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.289249122406462e-06,
+      "loss": 0.1938,
+      "step": 17515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.287822141042839e-06,
+      "loss": 0.1949,
+      "step": 17520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.286395159679215e-06,
+      "loss": 0.2227,
+      "step": 17525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2849681783155915e-06,
+      "loss": 0.0831,
+      "step": 17530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.283541196951968e-06,
+      "loss": 0.1108,
+      "step": 17535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.282114215588345e-06,
+      "loss": 0.0977,
+      "step": 17540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.280687234224721e-06,
+      "loss": 0.1692,
+      "step": 17545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.279260252861098e-06,
+      "loss": 0.3128,
+      "step": 17550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.277833271497474e-06,
+      "loss": 0.1295,
+      "step": 17555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.276406290133851e-06,
+      "loss": 0.1572,
+      "step": 17560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2749793087702285e-06,
+      "loss": 0.1622,
+      "step": 17565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.273552327406605e-06,
+      "loss": 0.1761,
+      "step": 17570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.272125346042981e-06,
+      "loss": 0.0468,
+      "step": 17575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2706983646793575e-06,
+      "loss": 0.0853,
+      "step": 17580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.269271383315734e-06,
+      "loss": 0.196,
+      "step": 17585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.267844401952111e-06,
+      "loss": 0.2446,
+      "step": 17590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.266417420588487e-06,
+      "loss": 0.2089,
+      "step": 17595
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.264990439224864e-06,
+      "loss": 0.8299,
+      "step": 17600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.263563457861241e-06,
+      "loss": 0.1186,
+      "step": 17605
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.262136476497617e-06,
+      "loss": 0.2162,
+      "step": 17610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.260709495133994e-06,
+      "loss": 0.1515,
+      "step": 17615
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.259282513770371e-06,
+      "loss": 0.1404,
+      "step": 17620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.257855532406747e-06,
+      "loss": 0.0805,
+      "step": 17625
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.256428551043123e-06,
+      "loss": 0.1446,
+      "step": 17630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.255001569679501e-06,
+      "loss": 0.1575,
+      "step": 17635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.253574588315877e-06,
+      "loss": 0.1792,
+      "step": 17640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.252147606952253e-06,
+      "loss": 0.2785,
+      "step": 17645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2507206255886304e-06,
+      "loss": 0.3261,
+      "step": 17650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.249293644225007e-06,
+      "loss": 0.1054,
+      "step": 17655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.247866662861384e-06,
+      "loss": 0.1029,
+      "step": 17660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.24643968149776e-06,
+      "loss": 0.1423,
+      "step": 17665
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.245012700134137e-06,
+      "loss": 0.1382,
+      "step": 17670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.243585718770513e-06,
+      "loss": 0.1093,
+      "step": 17675
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.242158737406889e-06,
+      "loss": 0.2346,
+      "step": 17680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2407317560432665e-06,
+      "loss": 0.1895,
+      "step": 17685
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.239304774679643e-06,
+      "loss": 0.268,
+      "step": 17690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.23787779331602e-06,
+      "loss": 0.2064,
+      "step": 17695
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.236450811952396e-06,
+      "loss": 0.4262,
+      "step": 17700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.235023830588773e-06,
+      "loss": 0.1548,
+      "step": 17705
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.23359684922515e-06,
+      "loss": 0.3673,
+      "step": 17710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.232169867861526e-06,
+      "loss": 0.1372,
+      "step": 17715
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2307428864979026e-06,
+      "loss": 0.2551,
+      "step": 17720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.229315905134279e-06,
+      "loss": 0.2321,
+      "step": 17725
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.227888923770656e-06,
+      "loss": 0.0697,
+      "step": 17730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.226461942407033e-06,
+      "loss": 0.0874,
+      "step": 17735
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.22503496104341e-06,
+      "loss": 0.1746,
+      "step": 17740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.223607979679786e-06,
+      "loss": 0.2093,
+      "step": 17745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.222180998316162e-06,
+      "loss": 0.4258,
+      "step": 17750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2207540169525395e-06,
+      "loss": 0.157,
+      "step": 17755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.219327035588916e-06,
+      "loss": 0.1516,
+      "step": 17760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.217900054225292e-06,
+      "loss": 0.1675,
+      "step": 17765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2164730728616685e-06,
+      "loss": 0.2068,
+      "step": 17770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.215046091498045e-06,
+      "loss": 0.1738,
+      "step": 17775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.213619110134423e-06,
+      "loss": 0.0531,
+      "step": 17780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.212192128770799e-06,
+      "loss": 0.1019,
+      "step": 17785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2107651474071756e-06,
+      "loss": 0.2422,
+      "step": 17790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.209338166043552e-06,
+      "loss": 0.2592,
+      "step": 17795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.207911184679928e-06,
+      "loss": 0.4477,
+      "step": 17800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.206484203316305e-06,
+      "loss": 0.1812,
+      "step": 17805
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.205057221952682e-06,
+      "loss": 0.1483,
+      "step": 17810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.203630240589058e-06,
+      "loss": 0.1255,
+      "step": 17815
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2022032592254344e-06,
+      "loss": 0.1468,
+      "step": 17820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2007762778618125e-06,
+      "loss": 0.0766,
+      "step": 17825
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.199349296498189e-06,
+      "loss": 0.0317,
+      "step": 17830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.197922315134565e-06,
+      "loss": 0.1377,
+      "step": 17835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1964953337709415e-06,
+      "loss": 0.1483,
+      "step": 17840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.195068352407318e-06,
+      "loss": 0.2709,
+      "step": 17845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.193641371043695e-06,
+      "loss": 0.3546,
+      "step": 17850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192214389680071e-06,
+      "loss": 0.1371,
+      "step": 17855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.190787408316448e-06,
+      "loss": 0.1574,
+      "step": 17860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.189360426952824e-06,
+      "loss": 0.1483,
+      "step": 17865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1879334455892e-06,
+      "loss": 0.1485,
+      "step": 17870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.186506464225578e-06,
+      "loss": 0.1632,
+      "step": 17875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.185079482861955e-06,
+      "loss": 0.1516,
+      "step": 17880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.183652501498331e-06,
+      "loss": 0.1936,
+      "step": 17885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.182225520134707e-06,
+      "loss": 0.3391,
+      "step": 17890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.180798538771084e-06,
+      "loss": 0.2453,
+      "step": 17895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.179371557407461e-06,
+      "loss": 0.4235,
+      "step": 17900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.177944576043837e-06,
+      "loss": 0.1901,
+      "step": 17905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.176517594680214e-06,
+      "loss": 0.1267,
+      "step": 17910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.17509061331659e-06,
+      "loss": 0.1892,
+      "step": 17915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.173663631952968e-06,
+      "loss": 0.2079,
+      "step": 17920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.172236650589344e-06,
+      "loss": 0.1089,
+      "step": 17925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.170809669225721e-06,
+      "loss": 0.0372,
+      "step": 17930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169382687862097e-06,
+      "loss": 0.0917,
+      "step": 17935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.167955706498473e-06,
+      "loss": 0.1767,
+      "step": 17940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1665287251348505e-06,
+      "loss": 0.2681,
+      "step": 17945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.165101743771227e-06,
+      "loss": 0.6623,
+      "step": 17950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.163674762407603e-06,
+      "loss": 0.1652,
+      "step": 17955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1622477810439795e-06,
+      "loss": 0.1949,
+      "step": 17960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.160820799680356e-06,
+      "loss": 0.1519,
+      "step": 17965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.159393818316734e-06,
+      "loss": 0.1855,
+      "step": 17970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.15796683695311e-06,
+      "loss": 0.1434,
+      "step": 17975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.156539855589487e-06,
+      "loss": 0.1216,
+      "step": 17980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.155112874225863e-06,
+      "loss": 0.2063,
+      "step": 17985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.15368589286224e-06,
+      "loss": 0.1467,
+      "step": 17990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1522589114986164e-06,
+      "loss": 0.1407,
+      "step": 17995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.150831930134993e-06,
+      "loss": 0.5262,
+      "step": 18000
+    },
+    {
+      "epoch": 0.5,
+      "eval_loss": 0.06189865618944168,
+      "eval_runtime": 1891.1772,
+      "eval_samples_per_second": 8.47,
+      "eval_steps_per_second": 2.118,
+      "eval_wer": 0.17731698521456907,
+      "step": 18000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.149404948771369e-06,
+      "loss": 0.1568,
+      "step": 18005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1479779674077455e-06,
+      "loss": 0.1248,
+      "step": 18010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1465509860441235e-06,
+      "loss": 0.207,
+      "step": 18015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1451240046805e-06,
+      "loss": 0.1563,
+      "step": 18020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.143697023316876e-06,
+      "loss": 0.1114,
+      "step": 18025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1422700419532525e-06,
+      "loss": 0.0905,
+      "step": 18030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.140843060589629e-06,
+      "loss": 0.1048,
+      "step": 18035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.139416079226006e-06,
+      "loss": 0.1992,
+      "step": 18040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.137989097862382e-06,
+      "loss": 0.1553,
+      "step": 18045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.136562116498759e-06,
+      "loss": 0.4436,
+      "step": 18050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.135135135135135e-06,
+      "loss": 0.1276,
+      "step": 18055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.133708153771511e-06,
+      "loss": 0.1311,
+      "step": 18060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.132281172407889e-06,
+      "loss": 0.1655,
+      "step": 18065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.130854191044266e-06,
+      "loss": 0.1517,
+      "step": 18070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.129427209680642e-06,
+      "loss": 0.0479,
+      "step": 18075
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1280002283170184e-06,
+      "loss": 0.0382,
+      "step": 18080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.126573246953396e-06,
+      "loss": 0.1758,
+      "step": 18085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.125146265589772e-06,
+      "loss": 0.2523,
+      "step": 18090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.123719284226148e-06,
+      "loss": 0.223,
+      "step": 18095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.122292302862525e-06,
+      "loss": 0.3482,
+      "step": 18100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.120865321498901e-06,
+      "loss": 0.0802,
+      "step": 18105
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.119438340135279e-06,
+      "loss": 0.1132,
+      "step": 18110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.118011358771655e-06,
+      "loss": 0.1595,
+      "step": 18115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.116584377408032e-06,
+      "loss": 0.1813,
+      "step": 18120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.115157396044408e-06,
+      "loss": 0.0865,
+      "step": 18125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.113730414680784e-06,
+      "loss": 0.0802,
+      "step": 18130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1123034333171616e-06,
+      "loss": 0.0854,
+      "step": 18135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.110876451953538e-06,
+      "loss": 0.1714,
+      "step": 18140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.109449470589914e-06,
+      "loss": 0.2229,
+      "step": 18145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1080224892262906e-06,
+      "loss": 0.1786,
+      "step": 18150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.106595507862667e-06,
+      "loss": 0.153,
+      "step": 18155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.105168526499045e-06,
+      "loss": 0.175,
+      "step": 18160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.103741545135421e-06,
+      "loss": 0.201,
+      "step": 18165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.102314563771798e-06,
+      "loss": 0.1102,
+      "step": 18170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.100887582408174e-06,
+      "loss": 0.195,
+      "step": 18175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.099460601044551e-06,
+      "loss": 0.0287,
+      "step": 18180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0980336196809275e-06,
+      "loss": 0.0427,
+      "step": 18185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.096606638317304e-06,
+      "loss": 0.2079,
+      "step": 18190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.09517965695368e-06,
+      "loss": 0.204,
+      "step": 18195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0937526755900565e-06,
+      "loss": 0.3059,
+      "step": 18200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0923256942264345e-06,
+      "loss": 0.1668,
+      "step": 18205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.090898712862811e-06,
+      "loss": 0.2312,
+      "step": 18210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089471731499187e-06,
+      "loss": 0.1517,
+      "step": 18215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0880447501355635e-06,
+      "loss": 0.1628,
+      "step": 18220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08661776877194e-06,
+      "loss": 0.0852,
+      "step": 18225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.085190787408317e-06,
+      "loss": 0.0657,
+      "step": 18230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.083763806044693e-06,
+      "loss": 0.12,
+      "step": 18235
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08233682468107e-06,
+      "loss": 0.1348,
+      "step": 18240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080909843317446e-06,
+      "loss": 0.219,
+      "step": 18245
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.079482861953822e-06,
+      "loss": 0.5334,
+      "step": 18250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0780558805902005e-06,
+      "loss": 0.1511,
+      "step": 18255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.076628899226577e-06,
+      "loss": 0.1443,
+      "step": 18260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.075201917862953e-06,
+      "loss": 0.1843,
+      "step": 18265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0737749364993295e-06,
+      "loss": 0.1302,
+      "step": 18270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.072347955135707e-06,
+      "loss": 0.0972,
+      "step": 18275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.070920973772083e-06,
+      "loss": 0.2034,
+      "step": 18280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.069493992408459e-06,
+      "loss": 0.125,
+      "step": 18285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068067011044836e-06,
+      "loss": 0.1517,
+      "step": 18290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.066640029681212e-06,
+      "loss": 0.2896,
+      "step": 18295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.06521304831759e-06,
+      "loss": 0.297,
+      "step": 18300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.063786066953966e-06,
+      "loss": 0.1865,
+      "step": 18305
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.062359085590343e-06,
+      "loss": 0.1178,
+      "step": 18310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.060932104226719e-06,
+      "loss": 0.181,
+      "step": 18315
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059505122863095e-06,
+      "loss": 0.1583,
+      "step": 18320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.058078141499473e-06,
+      "loss": 0.2078,
+      "step": 18325
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.056651160135849e-06,
+      "loss": 0.0346,
+      "step": 18330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.055224178772225e-06,
+      "loss": 0.142,
+      "step": 18335
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.053797197408602e-06,
+      "loss": 0.0779,
+      "step": 18340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052370216044978e-06,
+      "loss": 0.2198,
+      "step": 18345
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050943234681356e-06,
+      "loss": 0.4314,
+      "step": 18350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.049516253317732e-06,
+      "loss": 0.1005,
+      "step": 18355
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.048089271954109e-06,
+      "loss": 0.0979,
+      "step": 18360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.046662290590485e-06,
+      "loss": 0.2008,
+      "step": 18365
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.045235309226862e-06,
+      "loss": 0.1816,
+      "step": 18370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0438083278632385e-06,
+      "loss": 0.0629,
+      "step": 18375
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.042381346499615e-06,
+      "loss": 0.1564,
+      "step": 18380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.040954365135991e-06,
+      "loss": 0.1654,
+      "step": 18385
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0395273837723675e-06,
+      "loss": 0.2169,
+      "step": 18390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0381004024087456e-06,
+      "loss": 0.2023,
+      "step": 18395
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.036673421045122e-06,
+      "loss": 0.4018,
+      "step": 18400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.035246439681498e-06,
+      "loss": 0.1353,
+      "step": 18405
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0338194583178746e-06,
+      "loss": 0.1655,
+      "step": 18410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032392476954251e-06,
+      "loss": 0.297,
+      "step": 18415
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.030965495590628e-06,
+      "loss": 0.2271,
+      "step": 18420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0295385142270044e-06,
+      "loss": 0.1084,
+      "step": 18425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.028111532863381e-06,
+      "loss": 0.1346,
+      "step": 18430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.026684551499757e-06,
+      "loss": 0.1331,
+      "step": 18435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.025257570136134e-06,
+      "loss": 0.162,
+      "step": 18440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0238305887725115e-06,
+      "loss": 0.1401,
+      "step": 18445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.022403607408888e-06,
+      "loss": 0.4347,
+      "step": 18450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.020976626045264e-06,
+      "loss": 0.107,
+      "step": 18455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0195496446816405e-06,
+      "loss": 0.1016,
+      "step": 18460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.018122663318018e-06,
+      "loss": 0.1409,
+      "step": 18465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.016695681954394e-06,
+      "loss": 0.1334,
+      "step": 18470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.01526870059077e-06,
+      "loss": 0.1433,
+      "step": 18475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0138417192271476e-06,
+      "loss": 0.0913,
+      "step": 18480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.012414737863524e-06,
+      "loss": 0.0915,
+      "step": 18485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.010987756499901e-06,
+      "loss": 0.2491,
+      "step": 18490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.009560775136277e-06,
+      "loss": 0.1918,
+      "step": 18495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.008133793772654e-06,
+      "loss": 0.4121,
+      "step": 18500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.00670681240903e-06,
+      "loss": 0.1865,
+      "step": 18505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0052798310454064e-06,
+      "loss": 0.1793,
+      "step": 18510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.003852849681784e-06,
+      "loss": 0.1419,
+      "step": 18515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.00242586831816e-06,
+      "loss": 0.1374,
+      "step": 18520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.000998886954537e-06,
+      "loss": 0.0294,
+      "step": 18525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9995719055909135e-06,
+      "loss": 0.0931,
+      "step": 18530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.99814492422729e-06,
+      "loss": 0.2111,
+      "step": 18535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.996717942863666e-06,
+      "loss": 0.2355,
+      "step": 18540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.995290961500043e-06,
+      "loss": 0.2988,
+      "step": 18545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.99386398013642e-06,
+      "loss": 0.4671,
+      "step": 18550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.992436998772797e-06,
+      "loss": 0.0891,
+      "step": 18555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.991010017409173e-06,
+      "loss": 0.1915,
+      "step": 18560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9895830360455495e-06,
+      "loss": 0.177,
+      "step": 18565
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.988156054681927e-06,
+      "loss": 0.1975,
+      "step": 18570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.986729073318303e-06,
+      "loss": 0.2193,
+      "step": 18575
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.985302091954679e-06,
+      "loss": 0.1107,
+      "step": 18580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.983875110591056e-06,
+      "loss": 0.1263,
+      "step": 18585
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.982448129227433e-06,
+      "loss": 0.1766,
+      "step": 18590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.981021147863809e-06,
+      "loss": 0.1956,
+      "step": 18595
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.979594166500186e-06,
+      "loss": 0.3,
+      "step": 18600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.978167185136563e-06,
+      "loss": 0.1515,
+      "step": 18605
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.976740203772939e-06,
+      "loss": 0.1442,
+      "step": 18610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.975313222409316e-06,
+      "loss": 0.1284,
+      "step": 18615
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.973886241045693e-06,
+      "loss": 0.1416,
+      "step": 18620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.972459259682069e-06,
+      "loss": 0.1804,
+      "step": 18625
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.971032278318445e-06,
+      "loss": 0.0842,
+      "step": 18630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969605296954822e-06,
+      "loss": 0.1027,
+      "step": 18635
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.968178315591199e-06,
+      "loss": 0.1315,
+      "step": 18640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.966751334227575e-06,
+      "loss": 0.3734,
+      "step": 18645
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.965324352863952e-06,
+      "loss": 0.4681,
+      "step": 18650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.963897371500329e-06,
+      "loss": 0.1777,
+      "step": 18655
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.962470390136705e-06,
+      "loss": 0.2419,
+      "step": 18660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.961043408773082e-06,
+      "loss": 0.1455,
+      "step": 18665
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.959616427409459e-06,
+      "loss": 0.1937,
+      "step": 18670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.958189446045835e-06,
+      "loss": 0.1785,
+      "step": 18675
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.956762464682211e-06,
+      "loss": 0.0755,
+      "step": 18680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9553354833185884e-06,
+      "loss": 0.1912,
+      "step": 18685
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.953908501954965e-06,
+      "loss": 0.1188,
+      "step": 18690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.952481520591341e-06,
+      "loss": 0.1998,
+      "step": 18695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951054539227718e-06,
+      "loss": 0.2255,
+      "step": 18700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949627557864095e-06,
+      "loss": 0.0936,
+      "step": 18705
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.948200576500472e-06,
+      "loss": 0.0934,
+      "step": 18710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.946773595136848e-06,
+      "loss": 0.1606,
+      "step": 18715
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9453466137732245e-06,
+      "loss": 0.1571,
+      "step": 18720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.943919632409601e-06,
+      "loss": 0.0658,
+      "step": 18725
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.942492651045977e-06,
+      "loss": 0.069,
+      "step": 18730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.941065669682354e-06,
+      "loss": 0.1073,
+      "step": 18735
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.939638688318731e-06,
+      "loss": 0.0662,
+      "step": 18740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.938211706955108e-06,
+      "loss": 0.338,
+      "step": 18745
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.936784725591484e-06,
+      "loss": 0.2188,
+      "step": 18750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.935357744227861e-06,
+      "loss": 0.1863,
+      "step": 18755
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.933930762864238e-06,
+      "loss": 0.2082,
+      "step": 18760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.932503781500614e-06,
+      "loss": 0.1422,
+      "step": 18765
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9310768001369904e-06,
+      "loss": 0.2215,
+      "step": 18770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929649818773367e-06,
+      "loss": 0.0921,
+      "step": 18775
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.928222837409744e-06,
+      "loss": 0.098,
+      "step": 18780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.92679585604612e-06,
+      "loss": 0.0824,
+      "step": 18785
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.925368874682497e-06,
+      "loss": 0.1358,
+      "step": 18790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.923941893318874e-06,
+      "loss": 0.1977,
+      "step": 18795
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.92251491195525e-06,
+      "loss": 0.4446,
+      "step": 18800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.921087930591627e-06,
+      "loss": 0.1165,
+      "step": 18805
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.919660949228004e-06,
+      "loss": 0.1562,
+      "step": 18810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91823396786438e-06,
+      "loss": 0.0988,
+      "step": 18815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.916806986500756e-06,
+      "loss": 0.167,
+      "step": 18820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.915380005137133e-06,
+      "loss": 0.0711,
+      "step": 18825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91395302377351e-06,
+      "loss": 0.1165,
+      "step": 18830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.912526042409886e-06,
+      "loss": 0.0943,
+      "step": 18835
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.911099061046263e-06,
+      "loss": 0.231,
+      "step": 18840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.90967207968264e-06,
+      "loss": 0.2684,
+      "step": 18845
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908245098319017e-06,
+      "loss": 0.3325,
+      "step": 18850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906818116955393e-06,
+      "loss": 0.1257,
+      "step": 18855
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.90539113559177e-06,
+      "loss": 0.1308,
+      "step": 18860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.903964154228146e-06,
+      "loss": 0.1406,
+      "step": 18865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.902537172864522e-06,
+      "loss": 0.1921,
+      "step": 18870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9011101915008995e-06,
+      "loss": 0.1859,
+      "step": 18875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.899683210137276e-06,
+      "loss": 0.1193,
+      "step": 18880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.898256228773652e-06,
+      "loss": 0.0525,
+      "step": 18885
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.896829247410029e-06,
+      "loss": 0.148,
+      "step": 18890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.895402266046406e-06,
+      "loss": 0.2575,
+      "step": 18895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.893975284682783e-06,
+      "loss": 0.3488,
+      "step": 18900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.892548303319159e-06,
+      "loss": 0.1078,
+      "step": 18905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8911213219555355e-06,
+      "loss": 0.1403,
+      "step": 18910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.889694340591912e-06,
+      "loss": 0.151,
+      "step": 18915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.888267359228289e-06,
+      "loss": 0.1313,
+      "step": 18920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.886840377864665e-06,
+      "loss": 0.1577,
+      "step": 18925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.885413396501042e-06,
+      "loss": 0.0513,
+      "step": 18930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.883986415137419e-06,
+      "loss": 0.0805,
+      "step": 18935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.882559433773795e-06,
+      "loss": 0.1599,
+      "step": 18940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8811324524101725e-06,
+      "loss": 0.1188,
+      "step": 18945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.879705471046549e-06,
+      "loss": 0.2917,
+      "step": 18950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.878278489682925e-06,
+      "loss": 0.1129,
+      "step": 18955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8768515083193015e-06,
+      "loss": 0.1425,
+      "step": 18960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.875424526955679e-06,
+      "loss": 0.1773,
+      "step": 18965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.873997545592055e-06,
+      "loss": 0.1527,
+      "step": 18970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.872570564228431e-06,
+      "loss": 0.2487,
+      "step": 18975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.871143582864808e-06,
+      "loss": 0.1026,
+      "step": 18980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.869716601501185e-06,
+      "loss": 0.1377,
+      "step": 18985
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.868289620137561e-06,
+      "loss": 0.1255,
+      "step": 18990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.866862638773938e-06,
+      "loss": 0.3781,
+      "step": 18995
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.865435657410315e-06,
+      "loss": 0.3524,
+      "step": 19000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.864008676046691e-06,
+      "loss": 0.1394,
+      "step": 19005
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.862581694683068e-06,
+      "loss": 0.1842,
+      "step": 19010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.861154713319445e-06,
+      "loss": 0.1438,
+      "step": 19015
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.859727731955821e-06,
+      "loss": 0.1551,
+      "step": 19020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.858300750592197e-06,
+      "loss": 0.131,
+      "step": 19025
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8568737692285744e-06,
+      "loss": 0.1145,
+      "step": 19030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.855446787864951e-06,
+      "loss": 0.0724,
+      "step": 19035
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.854019806501328e-06,
+      "loss": 0.1353,
+      "step": 19040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.852592825137704e-06,
+      "loss": 0.2609,
+      "step": 19045
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.851165843774081e-06,
+      "loss": 0.3902,
+      "step": 19050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.849738862410458e-06,
+      "loss": 0.137,
+      "step": 19055
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.848311881046834e-06,
+      "loss": 0.141,
+      "step": 19060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8468848996832105e-06,
+      "loss": 0.1448,
+      "step": 19065
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.845457918319587e-06,
+      "loss": 0.1721,
+      "step": 19070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.844030936955963e-06,
+      "loss": 0.1683,
+      "step": 19075
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.84260395559234e-06,
+      "loss": 0.2113,
+      "step": 19080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.841176974228717e-06,
+      "loss": 0.158,
+      "step": 19085
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.839749992865094e-06,
+      "loss": 0.3753,
+      "step": 19090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.83832301150147e-06,
+      "loss": 0.1677,
+      "step": 19095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8368960301378474e-06,
+      "loss": 0.5716,
+      "step": 19100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.835469048774224e-06,
+      "loss": 0.1239,
+      "step": 19105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8340420674106e-06,
+      "loss": 0.1726,
+      "step": 19110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8326150860469764e-06,
+      "loss": 0.1212,
+      "step": 19115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.831188104683353e-06,
+      "loss": 0.2156,
+      "step": 19120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.82976112331973e-06,
+      "loss": 0.1658,
+      "step": 19125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.828334141956106e-06,
+      "loss": 0.0955,
+      "step": 19130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8269071605924835e-06,
+      "loss": 0.1119,
+      "step": 19135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.82548017922886e-06,
+      "loss": 0.1332,
+      "step": 19140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.824053197865236e-06,
+      "loss": 0.3604,
+      "step": 19145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.822626216501613e-06,
+      "loss": 0.3792,
+      "step": 19150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.82119923513799e-06,
+      "loss": 0.1281,
+      "step": 19155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.819772253774366e-06,
+      "loss": 0.176,
+      "step": 19160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.818345272410742e-06,
+      "loss": 0.137,
+      "step": 19165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.816918291047119e-06,
+      "loss": 0.1358,
+      "step": 19170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.815491309683496e-06,
+      "loss": 0.2131,
+      "step": 19175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.814064328319872e-06,
+      "loss": 0.1543,
+      "step": 19180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.812637346956249e-06,
+      "loss": 0.1545,
+      "step": 19185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.811210365592626e-06,
+      "loss": 0.2762,
+      "step": 19190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.809783384229003e-06,
+      "loss": 0.21,
+      "step": 19195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.808356402865379e-06,
+      "loss": 0.2492,
+      "step": 19200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.806929421501756e-06,
+      "loss": 0.1531,
+      "step": 19205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.805502440138132e-06,
+      "loss": 0.0952,
+      "step": 19210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.804075458774508e-06,
+      "loss": 0.1695,
+      "step": 19215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8026484774108855e-06,
+      "loss": 0.2047,
+      "step": 19220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.801221496047262e-06,
+      "loss": 0.1026,
+      "step": 19225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.799794514683639e-06,
+      "loss": 0.0705,
+      "step": 19230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.798367533320015e-06,
+      "loss": 0.1268,
+      "step": 19235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.796940551956392e-06,
+      "loss": 0.2088,
+      "step": 19240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.795513570592769e-06,
+      "loss": 0.2085,
+      "step": 19245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.794086589229145e-06,
+      "loss": 0.5394,
+      "step": 19250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7926596078655215e-06,
+      "loss": 0.1162,
+      "step": 19255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.791232626501898e-06,
+      "loss": 0.1291,
+      "step": 19260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789805645138274e-06,
+      "loss": 0.1823,
+      "step": 19265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.788378663774651e-06,
+      "loss": 0.2043,
+      "step": 19270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.786951682411028e-06,
+      "loss": 0.0934,
+      "step": 19275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.785524701047405e-06,
+      "loss": 0.1623,
+      "step": 19280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.784097719683781e-06,
+      "loss": 0.1179,
+      "step": 19285
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7826707383201585e-06,
+      "loss": 0.1375,
+      "step": 19290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.781243756956535e-06,
+      "loss": 0.3264,
+      "step": 19295
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.779816775592911e-06,
+      "loss": 0.3038,
+      "step": 19300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7783897942292875e-06,
+      "loss": 0.2116,
+      "step": 19305
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.776962812865664e-06,
+      "loss": 0.1304,
+      "step": 19310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.775535831502041e-06,
+      "loss": 0.2129,
+      "step": 19315
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.774108850138417e-06,
+      "loss": 0.1695,
+      "step": 19320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7726818687747945e-06,
+      "loss": 0.1732,
+      "step": 19325
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.771254887411171e-06,
+      "loss": 0.0737,
+      "step": 19330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.769827906047547e-06,
+      "loss": 0.0705,
+      "step": 19335
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.768400924683924e-06,
+      "loss": 0.2558,
+      "step": 19340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.766973943320301e-06,
+      "loss": 0.1931,
+      "step": 19345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.765546961956677e-06,
+      "loss": 0.4469,
+      "step": 19350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.764119980593053e-06,
+      "loss": 0.1145,
+      "step": 19355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.762692999229431e-06,
+      "loss": 0.1274,
+      "step": 19360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.761266017865807e-06,
+      "loss": 0.1393,
+      "step": 19365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.759839036502183e-06,
+      "loss": 0.1389,
+      "step": 19370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7584120551385604e-06,
+      "loss": 0.0456,
+      "step": 19375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.756985073774937e-06,
+      "loss": 0.1148,
+      "step": 19380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.755558092411314e-06,
+      "loss": 0.1288,
+      "step": 19385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.75413111104769e-06,
+      "loss": 0.1497,
+      "step": 19390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.752704129684067e-06,
+      "loss": 0.2034,
+      "step": 19395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.751277148320443e-06,
+      "loss": 0.4643,
+      "step": 19400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.74985016695682e-06,
+      "loss": 0.0889,
+      "step": 19405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7484231855931965e-06,
+      "loss": 0.1055,
+      "step": 19410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.746996204229573e-06,
+      "loss": 0.2563,
+      "step": 19415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.74556922286595e-06,
+      "loss": 0.1439,
+      "step": 19420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.744142241502326e-06,
+      "loss": 0.098,
+      "step": 19425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.742715260138703e-06,
+      "loss": 0.1248,
+      "step": 19430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.74128827877508e-06,
+      "loss": 0.1136,
+      "step": 19435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.739861297411456e-06,
+      "loss": 0.134,
+      "step": 19440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.738434316047833e-06,
+      "loss": 0.2237,
+      "step": 19445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.73700733468421e-06,
+      "loss": 0.3613,
+      "step": 19450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.735580353320586e-06,
+      "loss": 0.1377,
+      "step": 19455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7341533719569624e-06,
+      "loss": 0.1604,
+      "step": 19460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.732726390593339e-06,
+      "loss": 0.1725,
+      "step": 19465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.731299409229716e-06,
+      "loss": 0.1327,
+      "step": 19470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.729872427866092e-06,
+      "loss": 0.2093,
+      "step": 19475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7284454465024695e-06,
+      "loss": 0.0596,
+      "step": 19480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.727018465138846e-06,
+      "loss": 0.1679,
+      "step": 19485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.725591483775222e-06,
+      "loss": 0.1075,
+      "step": 19490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.724164502411599e-06,
+      "loss": 0.0746,
+      "step": 19495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.722737521047976e-06,
+      "loss": 0.5567,
+      "step": 19500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.721310539684352e-06,
+      "loss": 0.1269,
+      "step": 19505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.719883558320728e-06,
+      "loss": 0.0943,
+      "step": 19510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7184565769571056e-06,
+      "loss": 0.1285,
+      "step": 19515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.717029595593482e-06,
+      "loss": 0.1681,
+      "step": 19520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.715602614229858e-06,
+      "loss": 0.108,
+      "step": 19525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.714175632866235e-06,
+      "loss": 0.0724,
+      "step": 19530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.712748651502612e-06,
+      "loss": 0.171,
+      "step": 19535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.711321670138989e-06,
+      "loss": 0.2208,
+      "step": 19540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.709894688775365e-06,
+      "loss": 0.2565,
+      "step": 19545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.708467707411742e-06,
+      "loss": 0.5097,
+      "step": 19550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.707040726048118e-06,
+      "loss": 0.1562,
+      "step": 19555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.705613744684494e-06,
+      "loss": 0.134,
+      "step": 19560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7041867633208715e-06,
+      "loss": 0.1897,
+      "step": 19565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.702759781957248e-06,
+      "loss": 0.1167,
+      "step": 19570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.701332800593625e-06,
+      "loss": 0.1494,
+      "step": 19575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.699905819230001e-06,
+      "loss": 0.08,
+      "step": 19580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.698478837866378e-06,
+      "loss": 0.1657,
+      "step": 19585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.697051856502755e-06,
+      "loss": 0.1345,
+      "step": 19590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.695624875139131e-06,
+      "loss": 0.4074,
+      "step": 19595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6941978937755075e-06,
+      "loss": 0.5657,
+      "step": 19600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.692770912411884e-06,
+      "loss": 0.1583,
+      "step": 19605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.691343931048261e-06,
+      "loss": 0.1803,
+      "step": 19610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.689916949684637e-06,
+      "loss": 0.1114,
+      "step": 19615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.688489968321014e-06,
+      "loss": 0.1527,
+      "step": 19620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.687062986957391e-06,
+      "loss": 0.1877,
+      "step": 19625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.685636005593767e-06,
+      "loss": 0.0716,
+      "step": 19630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6842090242301445e-06,
+      "loss": 0.204,
+      "step": 19635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.682782042866521e-06,
+      "loss": 0.1118,
+      "step": 19640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.681355061502897e-06,
+      "loss": 0.1386,
+      "step": 19645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6799280801392735e-06,
+      "loss": 0.4537,
+      "step": 19650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.67850109877565e-06,
+      "loss": 0.1359,
+      "step": 19655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.677074117412027e-06,
+      "loss": 0.1826,
+      "step": 19660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.675647136048403e-06,
+      "loss": 0.1137,
+      "step": 19665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6742201546847805e-06,
+      "loss": 0.1227,
+      "step": 19670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.672793173321157e-06,
+      "loss": 0.1025,
+      "step": 19675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.671366191957533e-06,
+      "loss": 0.1241,
+      "step": 19680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.66993921059391e-06,
+      "loss": 0.1798,
+      "step": 19685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.668512229230287e-06,
+      "loss": 0.2408,
+      "step": 19690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.667085247866663e-06,
+      "loss": 0.1606,
+      "step": 19695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.665658266503039e-06,
+      "loss": 0.3169,
+      "step": 19700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.664231285139417e-06,
+      "loss": 0.1612,
+      "step": 19705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.662804303775793e-06,
+      "loss": 0.1583,
+      "step": 19710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.661377322412169e-06,
+      "loss": 0.096,
+      "step": 19715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6599503410485464e-06,
+      "loss": 0.153,
+      "step": 19720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.658523359684923e-06,
+      "loss": 0.1391,
+      "step": 19725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6570963783213e-06,
+      "loss": 0.0755,
+      "step": 19730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.655669396957676e-06,
+      "loss": 0.3471,
+      "step": 19735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.654242415594053e-06,
+      "loss": 0.328,
+      "step": 19740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.652815434230429e-06,
+      "loss": 0.2589,
+      "step": 19745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.651388452866805e-06,
+      "loss": 0.5256,
+      "step": 19750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6499614715031825e-06,
+      "loss": 0.1261,
+      "step": 19755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.648534490139559e-06,
+      "loss": 0.1566,
+      "step": 19760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.647107508775936e-06,
+      "loss": 0.1264,
+      "step": 19765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.645680527412312e-06,
+      "loss": 0.1441,
+      "step": 19770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.644253546048689e-06,
+      "loss": 0.0962,
+      "step": 19775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.642826564685066e-06,
+      "loss": 0.1559,
+      "step": 19780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.641399583321442e-06,
+      "loss": 0.3271,
+      "step": 19785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.639972601957819e-06,
+      "loss": 0.2518,
+      "step": 19790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.638545620594195e-06,
+      "loss": 0.2339,
+      "step": 19795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.637118639230572e-06,
+      "loss": 0.3957,
+      "step": 19800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6356916578669484e-06,
+      "loss": 0.1017,
+      "step": 19805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.634264676503325e-06,
+      "loss": 0.1268,
+      "step": 19810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.632837695139702e-06,
+      "loss": 0.1723,
+      "step": 19815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.631410713776078e-06,
+      "loss": 0.2455,
+      "step": 19820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6299837324124555e-06,
+      "loss": 0.1241,
+      "step": 19825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.628556751048832e-06,
+      "loss": 0.0737,
+      "step": 19830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.627129769685208e-06,
+      "loss": 0.0642,
+      "step": 19835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6257027883215845e-06,
+      "loss": 0.0967,
+      "step": 19840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.624275806957962e-06,
+      "loss": 0.2759,
+      "step": 19845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.622848825594338e-06,
+      "loss": 0.3172,
+      "step": 19850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.621421844230714e-06,
+      "loss": 0.1524,
+      "step": 19855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6199948628670916e-06,
+      "loss": 0.1349,
+      "step": 19860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.618567881503468e-06,
+      "loss": 0.1402,
+      "step": 19865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.617140900139844e-06,
+      "loss": 0.1438,
+      "step": 19870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.615713918776221e-06,
+      "loss": 0.0545,
+      "step": 19875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.614286937412598e-06,
+      "loss": 0.1018,
+      "step": 19880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.612859956048975e-06,
+      "loss": 0.0954,
+      "step": 19885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.611432974685351e-06,
+      "loss": 0.1953,
+      "step": 19890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.610005993321728e-06,
+      "loss": 0.3253,
+      "step": 19895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.608579011958104e-06,
+      "loss": 0.3014,
+      "step": 19900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.60715203059448e-06,
+      "loss": 0.1242,
+      "step": 19905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6057250492308575e-06,
+      "loss": 0.1286,
+      "step": 19910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.604298067867234e-06,
+      "loss": 0.1293,
+      "step": 19915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.602871086503611e-06,
+      "loss": 0.1594,
+      "step": 19920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.601444105139987e-06,
+      "loss": 0.1248,
+      "step": 19925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6000171237763645e-06,
+      "loss": 0.0895,
+      "step": 19930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.598590142412741e-06,
+      "loss": 0.1539,
+      "step": 19935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.597163161049117e-06,
+      "loss": 0.166,
+      "step": 19940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5957361796854935e-06,
+      "loss": 0.1508,
+      "step": 19945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.59430919832187e-06,
+      "loss": 0.3545,
+      "step": 19950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.592882216958247e-06,
+      "loss": 0.1807,
+      "step": 19955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.591455235594623e-06,
+      "loss": 0.2103,
+      "step": 19960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.590028254231e-06,
+      "loss": 0.2293,
+      "step": 19965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.588601272867377e-06,
+      "loss": 0.1707,
+      "step": 19970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.587174291503753e-06,
+      "loss": 0.0613,
+      "step": 19975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5857473101401305e-06,
+      "loss": 0.091,
+      "step": 19980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.584320328776507e-06,
+      "loss": 0.1545,
+      "step": 19985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.582893347412883e-06,
+      "loss": 0.1857,
+      "step": 19990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5814663660492595e-06,
+      "loss": 0.1609,
+      "step": 19995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.580039384685636e-06,
+      "loss": 0.4993,
+      "step": 20000
+    },
+    {
+      "epoch": 0.55,
+      "eval_loss": 0.0669386014342308,
+      "eval_runtime": 1938.2183,
+      "eval_samples_per_second": 8.264,
+      "eval_steps_per_second": 2.066,
+      "eval_wer": 0.18878470970068517,
+      "step": 20000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.578612403322013e-06,
+      "loss": 0.1699,
+      "step": 20005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.577185421958389e-06,
+      "loss": 0.1706,
+      "step": 20010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5757584405947665e-06,
+      "loss": 0.1171,
+      "step": 20015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.574331459231143e-06,
+      "loss": 0.2332,
+      "step": 20020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.57290447786752e-06,
+      "loss": 0.0381,
+      "step": 20025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.571477496503896e-06,
+      "loss": 0.1741,
+      "step": 20030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.570050515140273e-06,
+      "loss": 0.0784,
+      "step": 20035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.568623533776649e-06,
+      "loss": 0.1647,
+      "step": 20040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.567196552413025e-06,
+      "loss": 0.1712,
+      "step": 20045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.565769571049403e-06,
+      "loss": 0.3781,
+      "step": 20050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.564342589685779e-06,
+      "loss": 0.128,
+      "step": 20055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.562915608322155e-06,
+      "loss": 0.1474,
+      "step": 20060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5614886269585325e-06,
+      "loss": 0.1729,
+      "step": 20065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.560061645594909e-06,
+      "loss": 0.112,
+      "step": 20070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.558634664231286e-06,
+      "loss": 0.1977,
+      "step": 20075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.557207682867662e-06,
+      "loss": 0.0751,
+      "step": 20080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.555780701504039e-06,
+      "loss": 0.1949,
+      "step": 20085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.554353720140415e-06,
+      "loss": 0.048,
+      "step": 20090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.552926738776791e-06,
+      "loss": 0.2332,
+      "step": 20095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5514997574131685e-06,
+      "loss": 0.2679,
+      "step": 20100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.550072776049545e-06,
+      "loss": 0.1536,
+      "step": 20105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.548645794685922e-06,
+      "loss": 0.1591,
+      "step": 20110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.547218813322298e-06,
+      "loss": 0.154,
+      "step": 20115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5457918319586756e-06,
+      "loss": 0.1012,
+      "step": 20120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.544364850595052e-06,
+      "loss": 0.1548,
+      "step": 20125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.542937869231428e-06,
+      "loss": 0.0512,
+      "step": 20130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.541510887867805e-06,
+      "loss": 0.0681,
+      "step": 20135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.540083906504181e-06,
+      "loss": 0.0874,
+      "step": 20140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.538656925140558e-06,
+      "loss": 0.1211,
+      "step": 20145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5372299437769344e-06,
+      "loss": 0.4391,
+      "step": 20150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.535802962413312e-06,
+      "loss": 0.1699,
+      "step": 20155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.534375981049688e-06,
+      "loss": 0.1564,
+      "step": 20160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.532948999686064e-06,
+      "loss": 0.1292,
+      "step": 20165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5315220183224415e-06,
+      "loss": 0.2075,
+      "step": 20170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.530095036958818e-06,
+      "loss": 0.1313,
+      "step": 20175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.528668055595194e-06,
+      "loss": 0.2009,
+      "step": 20180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5272410742315705e-06,
+      "loss": 0.0512,
+      "step": 20185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.525814092867947e-06,
+      "loss": 0.1592,
+      "step": 20190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.524387111504324e-06,
+      "loss": 0.3646,
+      "step": 20195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5229601301407e-06,
+      "loss": 0.3226,
+      "step": 20200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5215331487770776e-06,
+      "loss": 0.1047,
+      "step": 20205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.520106167413454e-06,
+      "loss": 0.1136,
+      "step": 20210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.518679186049831e-06,
+      "loss": 0.1991,
+      "step": 20215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.517252204686207e-06,
+      "loss": 0.1289,
+      "step": 20220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.515825223322584e-06,
+      "loss": 0.1628,
+      "step": 20225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.51439824195896e-06,
+      "loss": 0.0617,
+      "step": 20230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5129712605953364e-06,
+      "loss": 0.0902,
+      "step": 20235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.511544279231714e-06,
+      "loss": 0.1425,
+      "step": 20240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.51011729786809e-06,
+      "loss": 0.1862,
+      "step": 20245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.508690316504467e-06,
+      "loss": 0.39,
+      "step": 20250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5072633351408435e-06,
+      "loss": 0.1516,
+      "step": 20255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.50583635377722e-06,
+      "loss": 0.1834,
+      "step": 20260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.504409372413597e-06,
+      "loss": 0.1808,
+      "step": 20265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.502982391049973e-06,
+      "loss": 0.1624,
+      "step": 20270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.50155540968635e-06,
+      "loss": 0.0985,
+      "step": 20275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.500128428322727e-06,
+      "loss": 0.102,
+      "step": 20280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.498701446959103e-06,
+      "loss": 0.116,
+      "step": 20285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4972744655954795e-06,
+      "loss": 0.1095,
+      "step": 20290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.495847484231856e-06,
+      "loss": 0.2776,
+      "step": 20295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.494420502868233e-06,
+      "loss": 0.2438,
+      "step": 20300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.492993521504609e-06,
+      "loss": 0.1154,
+      "step": 20305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.491566540140987e-06,
+      "loss": 0.169,
+      "step": 20310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.490139558777363e-06,
+      "loss": 0.1826,
+      "step": 20315
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.488712577413739e-06,
+      "loss": 0.1493,
+      "step": 20320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4872855960501165e-06,
+      "loss": 0.1289,
+      "step": 20325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.485858614686493e-06,
+      "loss": 0.0935,
+      "step": 20330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.484431633322869e-06,
+      "loss": 0.0403,
+      "step": 20335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4830046519592455e-06,
+      "loss": 0.1808,
+      "step": 20340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.481577670595623e-06,
+      "loss": 0.0966,
+      "step": 20345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.480150689231999e-06,
+      "loss": 0.411,
+      "step": 20350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.478723707868375e-06,
+      "loss": 0.0877,
+      "step": 20355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4772967265047525e-06,
+      "loss": 0.1105,
+      "step": 20360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.475869745141129e-06,
+      "loss": 0.1447,
+      "step": 20365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.474442763777506e-06,
+      "loss": 0.179,
+      "step": 20370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.473015782413882e-06,
+      "loss": 0.1692,
+      "step": 20375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.471588801050259e-06,
+      "loss": 0.1855,
+      "step": 20380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.470161819686635e-06,
+      "loss": 0.1729,
+      "step": 20385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.468734838323011e-06,
+      "loss": 0.1352,
+      "step": 20390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.467307856959389e-06,
+      "loss": 0.3104,
+      "step": 20395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.465880875595765e-06,
+      "loss": 0.3898,
+      "step": 20400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.464453894232142e-06,
+      "loss": 0.1416,
+      "step": 20405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4630269128685185e-06,
+      "loss": 0.1073,
+      "step": 20410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.461599931504895e-06,
+      "loss": 0.1166,
+      "step": 20415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.460172950141272e-06,
+      "loss": 0.1951,
+      "step": 20420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.458745968777648e-06,
+      "loss": 0.2497,
+      "step": 20425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.457318987414025e-06,
+      "loss": 0.2005,
+      "step": 20430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.455892006050401e-06,
+      "loss": 0.1289,
+      "step": 20435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.454465024686778e-06,
+      "loss": 0.1432,
+      "step": 20440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4530380433231545e-06,
+      "loss": 0.2041,
+      "step": 20445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.451611061959531e-06,
+      "loss": 0.3785,
+      "step": 20450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.450184080595908e-06,
+      "loss": 0.1748,
+      "step": 20455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.448757099232284e-06,
+      "loss": 0.1515,
+      "step": 20460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4473301178686616e-06,
+      "loss": 0.1289,
+      "step": 20465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.445903136505038e-06,
+      "loss": 0.1449,
+      "step": 20470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.444476155141414e-06,
+      "loss": 0.159,
+      "step": 20475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.443049173777791e-06,
+      "loss": 0.0797,
+      "step": 20480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.441622192414167e-06,
+      "loss": 0.1511,
+      "step": 20485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.440195211050544e-06,
+      "loss": 0.1973,
+      "step": 20490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4387682296869204e-06,
+      "loss": 0.2574,
+      "step": 20495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.437341248323298e-06,
+      "loss": 0.3839,
+      "step": 20500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.435914266959674e-06,
+      "loss": 0.1251,
+      "step": 20505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.43448728559605e-06,
+      "loss": 0.2167,
+      "step": 20510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4330603042324275e-06,
+      "loss": 0.1524,
+      "step": 20515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.431633322868804e-06,
+      "loss": 0.1234,
+      "step": 20520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.43020634150518e-06,
+      "loss": 0.1171,
+      "step": 20525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4287793601415565e-06,
+      "loss": 0.084,
+      "step": 20530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.427352378777934e-06,
+      "loss": 0.1054,
+      "step": 20535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.42592539741431e-06,
+      "loss": 0.1512,
+      "step": 20540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.424498416050686e-06,
+      "loss": 0.2869,
+      "step": 20545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4230714346870636e-06,
+      "loss": 0.4615,
+      "step": 20550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.42164445332344e-06,
+      "loss": 0.1295,
+      "step": 20555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.420217471959817e-06,
+      "loss": 0.1585,
+      "step": 20560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.418790490596193e-06,
+      "loss": 0.1597,
+      "step": 20565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.41736350923257e-06,
+      "loss": 0.1442,
+      "step": 20570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.415936527868946e-06,
+      "loss": 0.1386,
+      "step": 20575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4145095465053224e-06,
+      "loss": 0.1077,
+      "step": 20580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4130825651417e-06,
+      "loss": 0.1364,
+      "step": 20585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.411655583778076e-06,
+      "loss": 0.1187,
+      "step": 20590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.410228602414453e-06,
+      "loss": 0.2429,
+      "step": 20595
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.4088016210508295e-06,
+      "loss": 0.3807,
+      "step": 20600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.407374639687206e-06,
+      "loss": 0.1403,
+      "step": 20605
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.405947658323583e-06,
+      "loss": 0.1602,
+      "step": 20610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.404520676959959e-06,
+      "loss": 0.1603,
+      "step": 20615
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.403093695596336e-06,
+      "loss": 0.1754,
+      "step": 20620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.401666714232712e-06,
+      "loss": 0.2135,
+      "step": 20625
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.400239732869089e-06,
+      "loss": 0.0698,
+      "step": 20630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3988127515054655e-06,
+      "loss": 0.0483,
+      "step": 20635
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.397385770141842e-06,
+      "loss": 0.1518,
+      "step": 20640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.395958788778219e-06,
+      "loss": 0.334,
+      "step": 20645
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.394531807414595e-06,
+      "loss": 0.4927,
+      "step": 20650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.393104826050973e-06,
+      "loss": 0.0987,
+      "step": 20655
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.391677844687349e-06,
+      "loss": 0.1963,
+      "step": 20660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.390250863323725e-06,
+      "loss": 0.1392,
+      "step": 20665
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.388823881960102e-06,
+      "loss": 0.1584,
+      "step": 20670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.387396900596479e-06,
+      "loss": 0.0979,
+      "step": 20675
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.385969919232855e-06,
+      "loss": 0.0659,
+      "step": 20680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3845429378692315e-06,
+      "loss": 0.1128,
+      "step": 20685
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.383115956505609e-06,
+      "loss": 0.1779,
+      "step": 20690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.381688975141985e-06,
+      "loss": 0.2824,
+      "step": 20695
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.380261993778361e-06,
+      "loss": 0.2393,
+      "step": 20700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3788350124147385e-06,
+      "loss": 0.1658,
+      "step": 20705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.377408031051115e-06,
+      "loss": 0.0941,
+      "step": 20710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.375981049687491e-06,
+      "loss": 0.1224,
+      "step": 20715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.374554068323868e-06,
+      "loss": 0.2027,
+      "step": 20720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.373127086960245e-06,
+      "loss": 0.0874,
+      "step": 20725
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.371700105596621e-06,
+      "loss": 0.035,
+      "step": 20730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.370273124232997e-06,
+      "loss": 0.1263,
+      "step": 20735
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.368846142869375e-06,
+      "loss": 0.1757,
+      "step": 20740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.367419161505751e-06,
+      "loss": 0.2759,
+      "step": 20745
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.365992180142128e-06,
+      "loss": 0.5597,
+      "step": 20750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3645651987785045e-06,
+      "loss": 0.1479,
+      "step": 20755
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.363138217414881e-06,
+      "loss": 0.1397,
+      "step": 20760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.361711236051258e-06,
+      "loss": 0.1465,
+      "step": 20765
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.360284254687634e-06,
+      "loss": 0.1723,
+      "step": 20770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.358857273324011e-06,
+      "loss": 0.1158,
+      "step": 20775
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.357430291960387e-06,
+      "loss": 0.0537,
+      "step": 20780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.356003310596764e-06,
+      "loss": 0.0526,
+      "step": 20785
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3545763292331405e-06,
+      "loss": 0.2546,
+      "step": 20790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.353149347869517e-06,
+      "loss": 0.26,
+      "step": 20795
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.351722366505894e-06,
+      "loss": 0.7419,
+      "step": 20800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.35029538514227e-06,
+      "loss": 0.186,
+      "step": 20805
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3488684037786476e-06,
+      "loss": 0.1095,
+      "step": 20810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.347441422415024e-06,
+      "loss": 0.1813,
+      "step": 20815
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3460144410514e-06,
+      "loss": 0.169,
+      "step": 20820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.344587459687777e-06,
+      "loss": 0.2242,
+      "step": 20825
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.343160478324153e-06,
+      "loss": 0.1274,
+      "step": 20830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.34173349696053e-06,
+      "loss": 0.1351,
+      "step": 20835
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3403065155969064e-06,
+      "loss": 0.1285,
+      "step": 20840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.338879534233284e-06,
+      "loss": 0.202,
+      "step": 20845
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.33745255286966e-06,
+      "loss": 0.4036,
+      "step": 20850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.336025571506037e-06,
+      "loss": 0.1037,
+      "step": 20855
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3345985901424135e-06,
+      "loss": 0.1949,
+      "step": 20860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.33317160877879e-06,
+      "loss": 0.1631,
+      "step": 20865
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.331744627415166e-06,
+      "loss": 0.1696,
+      "step": 20870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3303176460515425e-06,
+      "loss": 0.1076,
+      "step": 20875
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.32889066468792e-06,
+      "loss": 0.093,
+      "step": 20880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.327463683324296e-06,
+      "loss": 0.0859,
+      "step": 20885
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.326036701960672e-06,
+      "loss": 0.1325,
+      "step": 20890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3246097205970496e-06,
+      "loss": 0.1415,
+      "step": 20895
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.323182739233426e-06,
+      "loss": 0.2313,
+      "step": 20900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.321755757869803e-06,
+      "loss": 0.068,
+      "step": 20905
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.320328776506179e-06,
+      "loss": 0.2165,
+      "step": 20910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.318901795142556e-06,
+      "loss": 0.1739,
+      "step": 20915
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.317474813778932e-06,
+      "loss": 0.2026,
+      "step": 20920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3160478324153084e-06,
+      "loss": 0.0622,
+      "step": 20925
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.314620851051686e-06,
+      "loss": 0.1017,
+      "step": 20930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.313193869688062e-06,
+      "loss": 0.1367,
+      "step": 20935
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.311766888324439e-06,
+      "loss": 0.2419,
+      "step": 20940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3103399069608155e-06,
+      "loss": 0.2738,
+      "step": 20945
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.308912925597193e-06,
+      "loss": 0.3316,
+      "step": 20950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.307485944233569e-06,
+      "loss": 0.1736,
+      "step": 20955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.306058962869945e-06,
+      "loss": 0.1284,
+      "step": 20960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.304631981506322e-06,
+      "loss": 0.1507,
+      "step": 20965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.303205000142698e-06,
+      "loss": 0.2268,
+      "step": 20970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.301778018779075e-06,
+      "loss": 0.0751,
+      "step": 20975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3003510374154515e-06,
+      "loss": 0.153,
+      "step": 20980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.298924056051828e-06,
+      "loss": 0.1526,
+      "step": 20985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.297497074688205e-06,
+      "loss": 0.3756,
+      "step": 20990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.296070093324581e-06,
+      "loss": 0.1955,
+      "step": 20995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.294928508233683e-06,
+      "loss": 0.2918,
+      "step": 21000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.293501526870059e-06,
+      "loss": 0.1349,
+      "step": 21005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.292074545506436e-06,
+      "loss": 0.1168,
+      "step": 21010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.2906475641428125e-06,
+      "loss": 0.1455,
+      "step": 21015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.28922058277919e-06,
+      "loss": 0.2116,
+      "step": 21020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.287793601415566e-06,
+      "loss": 0.0509,
+      "step": 21025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.286366620051942e-06,
+      "loss": 0.093,
+      "step": 21030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.284939638688319e-06,
+      "loss": 0.1425,
+      "step": 21035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.283512657324695e-06,
+      "loss": 0.2013,
+      "step": 21040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.282085675961072e-06,
+      "loss": 0.4366,
+      "step": 21045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.280658694597449e-06,
+      "loss": 0.2633,
+      "step": 21050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.279231713233826e-06,
+      "loss": 0.1191,
+      "step": 21055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.277804731870202e-06,
+      "loss": 0.1401,
+      "step": 21060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.276377750506579e-06,
+      "loss": 0.1819,
+      "step": 21065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.274950769142956e-06,
+      "loss": 0.1411,
+      "step": 21070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.273523787779332e-06,
+      "loss": 0.1296,
+      "step": 21075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.272096806415708e-06,
+      "loss": 0.1174,
+      "step": 21080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.270669825052085e-06,
+      "loss": 0.2018,
+      "step": 21085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.269242843688462e-06,
+      "loss": 0.1544,
+      "step": 21090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.267815862324838e-06,
+      "loss": 0.4319,
+      "step": 21095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2663888809612145e-06,
+      "loss": 0.5075,
+      "step": 21100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.264961899597592e-06,
+      "loss": 0.0764,
+      "step": 21105
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.263534918233968e-06,
+      "loss": 0.1584,
+      "step": 21110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.262107936870345e-06,
+      "loss": 0.1203,
+      "step": 21115
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2606809555067216e-06,
+      "loss": 0.1413,
+      "step": 21120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.259253974143098e-06,
+      "loss": 0.0478,
+      "step": 21125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.257826992779474e-06,
+      "loss": 0.0766,
+      "step": 21130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.256400011415851e-06,
+      "loss": 0.084,
+      "step": 21135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.254973030052228e-06,
+      "loss": 0.1812,
+      "step": 21140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.253546048688604e-06,
+      "loss": 0.1909,
+      "step": 21145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.252119067324981e-06,
+      "loss": 0.4103,
+      "step": 21150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.250692085961358e-06,
+      "loss": 0.1331,
+      "step": 21155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.249265104597735e-06,
+      "loss": 0.1468,
+      "step": 21160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.247838123234111e-06,
+      "loss": 0.1293,
+      "step": 21165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2464111418704875e-06,
+      "loss": 0.1586,
+      "step": 21170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.244984160506864e-06,
+      "loss": 0.1686,
+      "step": 21175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.243557179143241e-06,
+      "loss": 0.0373,
+      "step": 21180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.242130197779617e-06,
+      "loss": 0.1762,
+      "step": 21185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.240703216415994e-06,
+      "loss": 0.1562,
+      "step": 21190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.23927623505237e-06,
+      "loss": 0.1351,
+      "step": 21195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.237849253688747e-06,
+      "loss": 0.2875,
+      "step": 21200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2364222723251235e-06,
+      "loss": 0.1819,
+      "step": 21205
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.234995290961501e-06,
+      "loss": 0.1727,
+      "step": 21210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.233568309597877e-06,
+      "loss": 0.18,
+      "step": 21215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.232141328234253e-06,
+      "loss": 0.2093,
+      "step": 21220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.230714346870631e-06,
+      "loss": 0.0974,
+      "step": 21225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.229287365507007e-06,
+      "loss": 0.1422,
+      "step": 21230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.227860384143383e-06,
+      "loss": 0.0592,
+      "step": 21235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.22643340277976e-06,
+      "loss": 0.0948,
+      "step": 21240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.225006421416137e-06,
+      "loss": 0.1492,
+      "step": 21245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.223579440052513e-06,
+      "loss": 0.4271,
+      "step": 21250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.22215245868889e-06,
+      "loss": 0.1395,
+      "step": 21255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.220725477325267e-06,
+      "loss": 0.1376,
+      "step": 21260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.219298495961643e-06,
+      "loss": 0.1993,
+      "step": 21265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.21787151459802e-06,
+      "loss": 0.159,
+      "step": 21270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2164445332343965e-06,
+      "loss": 0.1568,
+      "step": 21275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.215017551870773e-06,
+      "loss": 0.0631,
+      "step": 21280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.213590570507149e-06,
+      "loss": 0.0991,
+      "step": 21285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2121635891435255e-06,
+      "loss": 0.093,
+      "step": 21290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.210736607779903e-06,
+      "loss": 0.1594,
+      "step": 21295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.209309626416279e-06,
+      "loss": 0.4936,
+      "step": 21300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.207882645052656e-06,
+      "loss": 0.2851,
+      "step": 21305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.206455663689033e-06,
+      "loss": 0.1734,
+      "step": 21310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.20502868232541e-06,
+      "loss": 0.1556,
+      "step": 21315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.203601700961786e-06,
+      "loss": 0.1751,
+      "step": 21320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2021747195981625e-06,
+      "loss": 0.1749,
+      "step": 21325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.200747738234539e-06,
+      "loss": 0.0951,
+      "step": 21330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.199320756870915e-06,
+      "loss": 0.0707,
+      "step": 21335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.197893775507292e-06,
+      "loss": 0.2316,
+      "step": 21340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.196466794143669e-06,
+      "loss": 0.1699,
+      "step": 21345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.195039812780046e-06,
+      "loss": 0.5433,
+      "step": 21350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.193612831416422e-06,
+      "loss": 0.1202,
+      "step": 21355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.1921858500527985e-06,
+      "loss": 0.1348,
+      "step": 21360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.190758868689176e-06,
+      "loss": 0.1644,
+      "step": 21365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.189331887325552e-06,
+      "loss": 0.1987,
+      "step": 21370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.187904905961928e-06,
+      "loss": 0.0504,
+      "step": 21375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.186477924598305e-06,
+      "loss": 0.1044,
+      "step": 21380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.185050943234682e-06,
+      "loss": 0.1106,
+      "step": 21385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.183623961871058e-06,
+      "loss": 0.2073,
+      "step": 21390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.182196980507435e-06,
+      "loss": 0.1753,
+      "step": 21395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.180769999143812e-06,
+      "loss": 0.4103,
+      "step": 21400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.179343017780188e-06,
+      "loss": 0.1441,
+      "step": 21405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.177916036416565e-06,
+      "loss": 0.1364,
+      "step": 21410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.176489055052942e-06,
+      "loss": 0.1536,
+      "step": 21415
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.175062073689318e-06,
+      "loss": 0.1332,
+      "step": 21420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.173635092325694e-06,
+      "loss": 0.1466,
+      "step": 21425
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.172208110962071e-06,
+      "loss": 0.1344,
+      "step": 21430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.170781129598448e-06,
+      "loss": 0.1798,
+      "step": 21435
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.169354148234824e-06,
+      "loss": 0.1823,
+      "step": 21440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.167927166871201e-06,
+      "loss": 0.1012,
+      "step": 21445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.166500185507578e-06,
+      "loss": 0.3861,
+      "step": 21450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.165073204143954e-06,
+      "loss": 0.1132,
+      "step": 21455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.163646222780331e-06,
+      "loss": 0.1297,
+      "step": 21460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1622192414167076e-06,
+      "loss": 0.1917,
+      "step": 21465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.160792260053084e-06,
+      "loss": 0.1914,
+      "step": 21470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.15936527868946e-06,
+      "loss": 0.0978,
+      "step": 21475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.157938297325837e-06,
+      "loss": 0.1377,
+      "step": 21480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.156511315962214e-06,
+      "loss": 0.1585,
+      "step": 21485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.15508433459859e-06,
+      "loss": 0.1966,
+      "step": 21490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.153657353234967e-06,
+      "loss": 0.2469,
+      "step": 21495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.152230371871344e-06,
+      "loss": 0.2976,
+      "step": 21500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.150803390507721e-06,
+      "loss": 0.1127,
+      "step": 21505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.149376409144097e-06,
+      "loss": 0.1079,
+      "step": 21510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1479494277804735e-06,
+      "loss": 0.1243,
+      "step": 21515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.14652244641685e-06,
+      "loss": 0.1525,
+      "step": 21520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.145095465053226e-06,
+      "loss": 0.1036,
+      "step": 21525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.143668483689603e-06,
+      "loss": 0.1125,
+      "step": 21530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.14224150232598e-06,
+      "loss": 0.0137,
+      "step": 21535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.140814520962357e-06,
+      "loss": 0.168,
+      "step": 21540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.139387539598733e-06,
+      "loss": 0.3707,
+      "step": 21545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1379605582351095e-06,
+      "loss": 0.5945,
+      "step": 21550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.136533576871487e-06,
+      "loss": 0.1056,
+      "step": 21555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.135106595507863e-06,
+      "loss": 0.148,
+      "step": 21560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.133679614144239e-06,
+      "loss": 0.1136,
+      "step": 21565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.132252632780616e-06,
+      "loss": 0.1784,
+      "step": 21570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.130825651416993e-06,
+      "loss": 0.0645,
+      "step": 21575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.129398670053369e-06,
+      "loss": 0.0799,
+      "step": 21580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.127971688689746e-06,
+      "loss": 0.0947,
+      "step": 21585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.126544707326123e-06,
+      "loss": 0.0839,
+      "step": 21590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.125117725962499e-06,
+      "loss": 0.1737,
+      "step": 21595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.123690744598876e-06,
+      "loss": 0.2954,
+      "step": 21600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.122263763235253e-06,
+      "loss": 0.1737,
+      "step": 21605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.120836781871629e-06,
+      "loss": 0.1259,
+      "step": 21610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.119409800508006e-06,
+      "loss": 0.1646,
+      "step": 21615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1179828191443825e-06,
+      "loss": 0.1515,
+      "step": 21620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.116555837780759e-06,
+      "loss": 0.085,
+      "step": 21625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.115128856417135e-06,
+      "loss": 0.0497,
+      "step": 21630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.113701875053512e-06,
+      "loss": 0.1347,
+      "step": 21635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.112274893689889e-06,
+      "loss": 0.1324,
+      "step": 21640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.110847912326265e-06,
+      "loss": 0.1714,
+      "step": 21645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.109420930962642e-06,
+      "loss": 0.21,
+      "step": 21650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.107993949599019e-06,
+      "loss": 0.1229,
+      "step": 21655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.106566968235396e-06,
+      "loss": 0.1586,
+      "step": 21660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.105139986871772e-06,
+      "loss": 0.1512,
+      "step": 21665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1037130055081485e-06,
+      "loss": 0.1363,
+      "step": 21670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.102286024144525e-06,
+      "loss": 0.1104,
+      "step": 21675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.100859042780901e-06,
+      "loss": 0.0787,
+      "step": 21680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.099432061417278e-06,
+      "loss": 0.2136,
+      "step": 21685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.098005080053655e-06,
+      "loss": 0.0721,
+      "step": 21690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.096578098690032e-06,
+      "loss": 0.2246,
+      "step": 21695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.095151117326408e-06,
+      "loss": 0.2887,
+      "step": 21700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.0937241359627845e-06,
+      "loss": 0.098,
+      "step": 21705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.092297154599162e-06,
+      "loss": 0.1688,
+      "step": 21710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.090870173235538e-06,
+      "loss": 0.1673,
+      "step": 21715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.089443191871914e-06,
+      "loss": 0.1548,
+      "step": 21720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.088016210508291e-06,
+      "loss": 0.1483,
+      "step": 21725
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.086589229144668e-06,
+      "loss": 0.1353,
+      "step": 21730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.085162247781044e-06,
+      "loss": 0.1639,
+      "step": 21735
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.083735266417421e-06,
+      "loss": 0.1349,
+      "step": 21740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.082308285053798e-06,
+      "loss": 0.1502,
+      "step": 21745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.080881303690174e-06,
+      "loss": 0.2666,
+      "step": 21750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.079454322326551e-06,
+      "loss": 0.0917,
+      "step": 21755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.078027340962928e-06,
+      "loss": 0.1191,
+      "step": 21760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.076600359599304e-06,
+      "loss": 0.1464,
+      "step": 21765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.07517337823568e-06,
+      "loss": 0.118,
+      "step": 21770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.073746396872057e-06,
+      "loss": 0.1638,
+      "step": 21775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.072319415508434e-06,
+      "loss": 0.0496,
+      "step": 21780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.07089243414481e-06,
+      "loss": 0.1049,
+      "step": 21785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.069465452781187e-06,
+      "loss": 0.164,
+      "step": 21790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.068038471417564e-06,
+      "loss": 0.1524,
+      "step": 21795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.06661149005394e-06,
+      "loss": 0.2877,
+      "step": 21800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.065184508690317e-06,
+      "loss": 0.1462,
+      "step": 21805
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.0637575273266936e-06,
+      "loss": 0.1615,
+      "step": 21810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.06233054596307e-06,
+      "loss": 0.1057,
+      "step": 21815
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.060903564599446e-06,
+      "loss": 0.146,
+      "step": 21820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.059476583235823e-06,
+      "loss": 0.1112,
+      "step": 21825
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.0580496018722e-06,
+      "loss": 0.0587,
+      "step": 21830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.056622620508576e-06,
+      "loss": 0.0971,
+      "step": 21835
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.055195639144953e-06,
+      "loss": 0.2333,
+      "step": 21840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.05376865778133e-06,
+      "loss": 0.1488,
+      "step": 21845
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.052341676417707e-06,
+      "loss": 0.7283,
+      "step": 21850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.050914695054083e-06,
+      "loss": 0.0987,
+      "step": 21855
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.0494877136904595e-06,
+      "loss": 0.1239,
+      "step": 21860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.048060732326836e-06,
+      "loss": 0.0952,
+      "step": 21865
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.046633750963212e-06,
+      "loss": 0.136,
+      "step": 21870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.045206769599589e-06,
+      "loss": 0.1609,
+      "step": 21875
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.043779788235966e-06,
+      "loss": 0.063,
+      "step": 21880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.042352806872343e-06,
+      "loss": 0.1227,
+      "step": 21885
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.040925825508719e-06,
+      "loss": 0.1381,
+      "step": 21890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.0394988441450955e-06,
+      "loss": 0.2777,
+      "step": 21895
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.038071862781473e-06,
+      "loss": 0.5384,
+      "step": 21900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.036644881417849e-06,
+      "loss": 0.1427,
+      "step": 21905
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.035217900054225e-06,
+      "loss": 0.1222,
+      "step": 21910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.033790918690602e-06,
+      "loss": 0.1591,
+      "step": 21915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.032363937326979e-06,
+      "loss": 0.0948,
+      "step": 21920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.030936955963355e-06,
+      "loss": 0.1851,
+      "step": 21925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.029509974599732e-06,
+      "loss": 0.0329,
+      "step": 21930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.028082993236109e-06,
+      "loss": 0.1093,
+      "step": 21935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.026656011872485e-06,
+      "loss": 0.3272,
+      "step": 21940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.025229030508862e-06,
+      "loss": 0.306,
+      "step": 21945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.023802049145239e-06,
+      "loss": 0.5171,
+      "step": 21950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.022375067781615e-06,
+      "loss": 0.1122,
+      "step": 21955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.020948086417991e-06,
+      "loss": 0.0891,
+      "step": 21960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.019521105054368e-06,
+      "loss": 0.1444,
+      "step": 21965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.018094123690745e-06,
+      "loss": 0.1775,
+      "step": 21970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.016667142327121e-06,
+      "loss": 0.1363,
+      "step": 21975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.015240160963498e-06,
+      "loss": 0.0984,
+      "step": 21980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.013813179599875e-06,
+      "loss": 0.1114,
+      "step": 21985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.012386198236252e-06,
+      "loss": 0.1603,
+      "step": 21990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.010959216872628e-06,
+      "loss": 0.1489,
+      "step": 21995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.009532235509005e-06,
+      "loss": 0.3808,
+      "step": 22000
+    },
+    {
+      "epoch": 0.61,
+      "eval_loss": 0.059859082102775574,
+      "eval_runtime": 1843.3789,
+      "eval_samples_per_second": 8.689,
+      "eval_steps_per_second": 2.173,
+      "eval_wer": 0.16849981968986658,
+      "step": 22000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.008105254145381e-06,
+      "loss": 0.0803,
+      "step": 22005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.006678272781758e-06,
+      "loss": 0.1432,
+      "step": 22010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.0052512914181345e-06,
+      "loss": 0.1594,
+      "step": 22015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.003824310054511e-06,
+      "loss": 0.2076,
+      "step": 22020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.002397328690887e-06,
+      "loss": 0.0678,
+      "step": 22025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.000970347327264e-06,
+      "loss": 0.0397,
+      "step": 22030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.999543365963641e-06,
+      "loss": 0.1157,
+      "step": 22035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.998116384600018e-06,
+      "loss": 0.1655,
+      "step": 22040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.996689403236394e-06,
+      "loss": 0.0887,
+      "step": 22045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.9952624218727705e-06,
+      "loss": 0.3724,
+      "step": 22050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.993835440509148e-06,
+      "loss": 0.0935,
+      "step": 22055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.992408459145524e-06,
+      "loss": 0.1776,
+      "step": 22060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.9909814777819e-06,
+      "loss": 0.1447,
+      "step": 22065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.989554496418277e-06,
+      "loss": 0.1518,
+      "step": 22070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.988127515054654e-06,
+      "loss": 0.1164,
+      "step": 22075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.98670053369103e-06,
+      "loss": 0.1649,
+      "step": 22080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.9852735523274074e-06,
+      "loss": 0.0446,
+      "step": 22085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.983846570963784e-06,
+      "loss": 0.2027,
+      "step": 22090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.98241958960016e-06,
+      "loss": 0.185,
+      "step": 22095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.980992608236537e-06,
+      "loss": 0.2873,
+      "step": 22100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.979565626872914e-06,
+      "loss": 0.1553,
+      "step": 22105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.97813864550929e-06,
+      "loss": 0.156,
+      "step": 22110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.976711664145666e-06,
+      "loss": 0.1238,
+      "step": 22115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.975284682782043e-06,
+      "loss": 0.1911,
+      "step": 22120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.97385770141842e-06,
+      "loss": 0.0836,
+      "step": 22125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.972430720054796e-06,
+      "loss": 0.1757,
+      "step": 22130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.971003738691173e-06,
+      "loss": 0.093,
+      "step": 22135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.96957675732755e-06,
+      "loss": 0.1251,
+      "step": 22140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.968149775963927e-06,
+      "loss": 0.1767,
+      "step": 22145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.966722794600303e-06,
+      "loss": 0.3181,
+      "step": 22150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.9652958132366796e-06,
+      "loss": 0.1479,
+      "step": 22155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.963868831873056e-06,
+      "loss": 0.1128,
+      "step": 22160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.962441850509432e-06,
+      "loss": 0.1561,
+      "step": 22165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.961014869145809e-06,
+      "loss": 0.2144,
+      "step": 22170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.959587887782186e-06,
+      "loss": 0.0875,
+      "step": 22175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.958160906418563e-06,
+      "loss": 0.1531,
+      "step": 22180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.956733925054939e-06,
+      "loss": 0.1169,
+      "step": 22185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.955306943691316e-06,
+      "loss": 0.1518,
+      "step": 22190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.953879962327693e-06,
+      "loss": 0.1385,
+      "step": 22195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.952452980964069e-06,
+      "loss": 0.3973,
+      "step": 22200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9510259996004455e-06,
+      "loss": 0.1574,
+      "step": 22205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.949599018236822e-06,
+      "loss": 0.2186,
+      "step": 22210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.948172036873198e-06,
+      "loss": 0.1752,
+      "step": 22215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.946745055509575e-06,
+      "loss": 0.1642,
+      "step": 22220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.945318074145952e-06,
+      "loss": 0.1781,
+      "step": 22225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.943891092782329e-06,
+      "loss": 0.0601,
+      "step": 22230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.942464111418705e-06,
+      "loss": 0.0634,
+      "step": 22235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.941037130055082e-06,
+      "loss": 0.1057,
+      "step": 22240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.939610148691459e-06,
+      "loss": 0.2088,
+      "step": 22245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.938183167327835e-06,
+      "loss": 0.2691,
+      "step": 22250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.936756185964211e-06,
+      "loss": 0.1449,
+      "step": 22255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.935329204600588e-06,
+      "loss": 0.105,
+      "step": 22260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.933902223236965e-06,
+      "loss": 0.1411,
+      "step": 22265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.932475241873341e-06,
+      "loss": 0.1517,
+      "step": 22270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9310482605097185e-06,
+      "loss": 0.1313,
+      "step": 22275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.929621279146095e-06,
+      "loss": 0.1082,
+      "step": 22280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.928194297782471e-06,
+      "loss": 0.1042,
+      "step": 22285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.926767316418848e-06,
+      "loss": 0.0742,
+      "step": 22290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.925340335055225e-06,
+      "loss": 0.1411,
+      "step": 22295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.924198749964326e-06,
+      "loss": 0.3382,
+      "step": 22300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.922771768600702e-06,
+      "loss": 0.1718,
+      "step": 22305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9213447872370794e-06,
+      "loss": 0.1203,
+      "step": 22310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.919917805873456e-06,
+      "loss": 0.1173,
+      "step": 22315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.918490824509832e-06,
+      "loss": 0.1667,
+      "step": 22320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9170638431462084e-06,
+      "loss": 0.147,
+      "step": 22325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.915636861782585e-06,
+      "loss": 0.0457,
+      "step": 22330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.914209880418962e-06,
+      "loss": 0.0543,
+      "step": 22335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.912782899055338e-06,
+      "loss": 0.1849,
+      "step": 22340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9113559176917155e-06,
+      "loss": 0.1187,
+      "step": 22345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.909928936328092e-06,
+      "loss": 0.5127,
+      "step": 22350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.908501954964469e-06,
+      "loss": 0.1292,
+      "step": 22355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.907074973600845e-06,
+      "loss": 0.0939,
+      "step": 22360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.905647992237222e-06,
+      "loss": 0.1346,
+      "step": 22365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.904221010873598e-06,
+      "loss": 0.1726,
+      "step": 22370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.902794029509974e-06,
+      "loss": 0.1197,
+      "step": 22375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9013670481463516e-06,
+      "loss": 0.0341,
+      "step": 22380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.899940066782728e-06,
+      "loss": 0.1054,
+      "step": 22385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.898513085419105e-06,
+      "loss": 0.2072,
+      "step": 22390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.897086104055481e-06,
+      "loss": 0.2113,
+      "step": 22395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.895659122691858e-06,
+      "loss": 0.3128,
+      "step": 22400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.894232141328235e-06,
+      "loss": 0.1308,
+      "step": 22405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.892805159964611e-06,
+      "loss": 0.186,
+      "step": 22410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.891378178600988e-06,
+      "loss": 0.1807,
+      "step": 22415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.889951197237364e-06,
+      "loss": 0.1598,
+      "step": 22420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.888524215873741e-06,
+      "loss": 0.0927,
+      "step": 22425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.8870972345101175e-06,
+      "loss": 0.0842,
+      "step": 22430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.885670253146494e-06,
+      "loss": 0.2275,
+      "step": 22435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.884243271782871e-06,
+      "loss": 0.1117,
+      "step": 22440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.882816290419247e-06,
+      "loss": 0.3245,
+      "step": 22445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.8813893090556245e-06,
+      "loss": 0.5859,
+      "step": 22450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.879962327692001e-06,
+      "loss": 0.1725,
+      "step": 22455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.878535346328377e-06,
+      "loss": 0.1432,
+      "step": 22460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.8771083649647535e-06,
+      "loss": 0.2215,
+      "step": 22465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.875681383601131e-06,
+      "loss": 0.1653,
+      "step": 22470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.874254402237507e-06,
+      "loss": 0.1194,
+      "step": 22475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.872827420873883e-06,
+      "loss": 0.1344,
+      "step": 22480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.871400439510261e-06,
+      "loss": 0.256,
+      "step": 22485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.869973458146637e-06,
+      "loss": 0.0864,
+      "step": 22490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.868546476783013e-06,
+      "loss": 0.4852,
+      "step": 22495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.8671194954193905e-06,
+      "loss": 0.4037,
+      "step": 22500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.865692514055767e-06,
+      "loss": 0.1035,
+      "step": 22505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.864265532692143e-06,
+      "loss": 0.1562,
+      "step": 22510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.86283855132852e-06,
+      "loss": 0.1622,
+      "step": 22515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.861411569964897e-06,
+      "loss": 0.1827,
+      "step": 22520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.859984588601273e-06,
+      "loss": 0.1404,
+      "step": 22525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.858557607237649e-06,
+      "loss": 0.1047,
+      "step": 22530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8571306258740265e-06,
+      "loss": 0.1108,
+      "step": 22535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.855703644510403e-06,
+      "loss": 0.1106,
+      "step": 22540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.85427666314678e-06,
+      "loss": 0.1454,
+      "step": 22545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.852849681783156e-06,
+      "loss": 0.4207,
+      "step": 22550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.851422700419533e-06,
+      "loss": 0.1075,
+      "step": 22555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.84999571905591e-06,
+      "loss": 0.1428,
+      "step": 22560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.848568737692286e-06,
+      "loss": 0.1541,
+      "step": 22565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.847141756328663e-06,
+      "loss": 0.2013,
+      "step": 22570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.845714774965039e-06,
+      "loss": 0.136,
+      "step": 22575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.844287793601416e-06,
+      "loss": 0.1132,
+      "step": 22580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8428608122377925e-06,
+      "loss": 0.1936,
+      "step": 22585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.841433830874169e-06,
+      "loss": 0.2513,
+      "step": 22590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.840006849510546e-06,
+      "loss": 0.367,
+      "step": 22595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.838579868146922e-06,
+      "loss": 0.2205,
+      "step": 22600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8371528867832995e-06,
+      "loss": 0.1057,
+      "step": 22605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.835725905419676e-06,
+      "loss": 0.1495,
+      "step": 22610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.834298924056052e-06,
+      "loss": 0.1303,
+      "step": 22615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8328719426924285e-06,
+      "loss": 0.1811,
+      "step": 22620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.831444961328805e-06,
+      "loss": 0.0864,
+      "step": 22625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.830017979965182e-06,
+      "loss": 0.0427,
+      "step": 22630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.828590998601558e-06,
+      "loss": 0.1305,
+      "step": 22635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8271640172379356e-06,
+      "loss": 0.1432,
+      "step": 22640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.825737035874312e-06,
+      "loss": 0.1697,
+      "step": 22645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.824310054510688e-06,
+      "loss": 0.386,
+      "step": 22650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8228830731470654e-06,
+      "loss": 0.1587,
+      "step": 22655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.821456091783442e-06,
+      "loss": 0.1269,
+      "step": 22660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.820029110419818e-06,
+      "loss": 0.1869,
+      "step": 22665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8186021290561944e-06,
+      "loss": 0.1689,
+      "step": 22670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.817175147692572e-06,
+      "loss": 0.157,
+      "step": 22675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.815748166328948e-06,
+      "loss": 0.0621,
+      "step": 22680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8143211849653243e-06,
+      "loss": 0.1575,
+      "step": 22685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8128942036017015e-06,
+      "loss": 0.1115,
+      "step": 22690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.811467222238078e-06,
+      "loss": 0.1717,
+      "step": 22695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8100402408744546e-06,
+      "loss": 0.3862,
+      "step": 22700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.808613259510831e-06,
+      "loss": 0.141,
+      "step": 22705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.807186278147208e-06,
+      "loss": 0.0854,
+      "step": 22710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8057592967835845e-06,
+      "loss": 0.1455,
+      "step": 22715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.804332315419961e-06,
+      "loss": 0.1458,
+      "step": 22720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8029053340563376e-06,
+      "loss": 0.0959,
+      "step": 22725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.801478352692714e-06,
+      "loss": 0.1955,
+      "step": 22730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.800051371329091e-06,
+      "loss": 0.0662,
+      "step": 22735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7986243899654674e-06,
+      "loss": 0.169,
+      "step": 22740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7971974086018438e-06,
+      "loss": 0.1332,
+      "step": 22745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7957704272382205e-06,
+      "loss": 0.4515,
+      "step": 22750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.794343445874597e-06,
+      "loss": 0.0705,
+      "step": 22755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.792916464510974e-06,
+      "loss": 0.2404,
+      "step": 22760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7914894831473504e-06,
+      "loss": 0.0873,
+      "step": 22765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.790062501783727e-06,
+      "loss": 0.1521,
+      "step": 22770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7886355204201035e-06,
+      "loss": 0.0806,
+      "step": 22775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.78720853905648e-06,
+      "loss": 0.1339,
+      "step": 22780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.785781557692857e-06,
+      "loss": 0.1963,
+      "step": 22785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7843545763292333e-06,
+      "loss": 0.1837,
+      "step": 22790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.78292759496561e-06,
+      "loss": 0.1914,
+      "step": 22795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7815006136019864e-06,
+      "loss": 0.3424,
+      "step": 22800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7800736322383636e-06,
+      "loss": 0.1903,
+      "step": 22805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.77864665087474e-06,
+      "loss": 0.1242,
+      "step": 22810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7772196695111163e-06,
+      "loss": 0.1694,
+      "step": 22815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.775792688147493e-06,
+      "loss": 0.1367,
+      "step": 22820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7743657067838694e-06,
+      "loss": 0.0764,
+      "step": 22825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7729387254202466e-06,
+      "loss": 0.0922,
+      "step": 22830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.771511744056623e-06,
+      "loss": 0.0857,
+      "step": 22835
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7700847626929993e-06,
+      "loss": 0.1124,
+      "step": 22840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.768657781329376e-06,
+      "loss": 0.2392,
+      "step": 22845
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7672307999657524e-06,
+      "loss": 0.6366,
+      "step": 22850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7658038186021296e-06,
+      "loss": 0.1192,
+      "step": 22855
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.764376837238506e-06,
+      "loss": 0.1585,
+      "step": 22860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7629498558748827e-06,
+      "loss": 0.1129,
+      "step": 22865
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.761522874511259e-06,
+      "loss": 0.1819,
+      "step": 22870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7600958931476358e-06,
+      "loss": 0.1054,
+      "step": 22875
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7586689117840125e-06,
+      "loss": 0.1105,
+      "step": 22880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.757241930420389e-06,
+      "loss": 0.0801,
+      "step": 22885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7558149490567656e-06,
+      "loss": 0.0801,
+      "step": 22890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7543879676931424e-06,
+      "loss": 0.1595,
+      "step": 22895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.752960986329519e-06,
+      "loss": 0.604,
+      "step": 22900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7515340049658955e-06,
+      "loss": 0.093,
+      "step": 22905
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.750107023602272e-06,
+      "loss": 0.187,
+      "step": 22910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7486800422386486e-06,
+      "loss": 0.1357,
+      "step": 22915
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7472530608750254e-06,
+      "loss": 0.1621,
+      "step": 22920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.745826079511402e-06,
+      "loss": 0.0294,
+      "step": 22925
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7443990981477785e-06,
+      "loss": 0.0212,
+      "step": 22930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.742972116784155e-06,
+      "loss": 0.1594,
+      "step": 22935
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.741545135420532e-06,
+      "loss": 0.1677,
+      "step": 22940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7401181540569083e-06,
+      "loss": 0.1924,
+      "step": 22945
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.738691172693285e-06,
+      "loss": 0.4647,
+      "step": 22950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7372641913296614e-06,
+      "loss": 0.0777,
+      "step": 22955
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.735837209966038e-06,
+      "loss": 0.1238,
+      "step": 22960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.734410228602415e-06,
+      "loss": 0.1013,
+      "step": 22965
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7329832472387913e-06,
+      "loss": 0.1405,
+      "step": 22970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.731556265875168e-06,
+      "loss": 0.1518,
+      "step": 22975
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7301292845115444e-06,
+      "loss": 0.0388,
+      "step": 22980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7287023031479216e-06,
+      "loss": 0.0532,
+      "step": 22985
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.727275321784298e-06,
+      "loss": 0.1454,
+      "step": 22990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7258483404206747e-06,
+      "loss": 0.3034,
+      "step": 22995
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.724421359057051e-06,
+      "loss": 0.447,
+      "step": 23000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7229943776934273e-06,
+      "loss": 0.1081,
+      "step": 23005
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7215673963298045e-06,
+      "loss": 0.1586,
+      "step": 23010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.720140414966181e-06,
+      "loss": 0.1271,
+      "step": 23015
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7187134336025576e-06,
+      "loss": 0.128,
+      "step": 23020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.717286452238934e-06,
+      "loss": 0.081,
+      "step": 23025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7158594708753103e-06,
+      "loss": 0.1812,
+      "step": 23030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7144324895116875e-06,
+      "loss": 0.1127,
+      "step": 23035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.713005508148064e-06,
+      "loss": 0.2226,
+      "step": 23040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7115785267844406e-06,
+      "loss": 0.3506,
+      "step": 23045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.710151545420817e-06,
+      "loss": 0.221,
+      "step": 23050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.708724564057194e-06,
+      "loss": 0.1317,
+      "step": 23055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7072975826935705e-06,
+      "loss": 0.0892,
+      "step": 23060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.705870601329947e-06,
+      "loss": 0.149,
+      "step": 23065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7044436199663236e-06,
+      "loss": 0.1818,
+      "step": 23070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7030166386027e-06,
+      "loss": 0.0497,
+      "step": 23075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.701589657239077e-06,
+      "loss": 0.0308,
+      "step": 23080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7001626758754534e-06,
+      "loss": 0.2085,
+      "step": 23085
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.69873569451183e-06,
+      "loss": 0.2694,
+      "step": 23090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6973087131482065e-06,
+      "loss": 0.232,
+      "step": 23095
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.695881731784583e-06,
+      "loss": 0.3407,
+      "step": 23100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.69445475042096e-06,
+      "loss": 0.1512,
+      "step": 23105
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6930277690573364e-06,
+      "loss": 0.0969,
+      "step": 23110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.691600787693713e-06,
+      "loss": 0.1553,
+      "step": 23115
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6901738063300895e-06,
+      "loss": 0.2289,
+      "step": 23120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.688746824966466e-06,
+      "loss": 0.096,
+      "step": 23125
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.687319843602843e-06,
+      "loss": 0.0663,
+      "step": 23130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6858928622392193e-06,
+      "loss": 0.1077,
+      "step": 23135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.684465880875596e-06,
+      "loss": 0.0969,
+      "step": 23140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6830388995119724e-06,
+      "loss": 0.3644,
+      "step": 23145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6816119181483496e-06,
+      "loss": 0.568,
+      "step": 23150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.680184936784726e-06,
+      "loss": 0.121,
+      "step": 23155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6787579554211023e-06,
+      "loss": 0.1523,
+      "step": 23160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.677330974057479e-06,
+      "loss": 0.1514,
+      "step": 23165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6759039926938554e-06,
+      "loss": 0.1478,
+      "step": 23170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6744770113302326e-06,
+      "loss": 0.0865,
+      "step": 23175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.673050029966609e-06,
+      "loss": 0.0337,
+      "step": 23180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6716230486029857e-06,
+      "loss": 0.0987,
+      "step": 23185
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.670196067239362e-06,
+      "loss": 0.1177,
+      "step": 23190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6687690858757384e-06,
+      "loss": 0.2375,
+      "step": 23195
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6673421045121156e-06,
+      "loss": 0.3224,
+      "step": 23200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.665915123148492e-06,
+      "loss": 0.1336,
+      "step": 23205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6644881417848687e-06,
+      "loss": 0.1603,
+      "step": 23210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.663061160421245e-06,
+      "loss": 0.1432,
+      "step": 23215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6616341790576213e-06,
+      "loss": 0.1694,
+      "step": 23220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6602071976939985e-06,
+      "loss": 0.1676,
+      "step": 23225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.658780216330375e-06,
+      "loss": 0.0683,
+      "step": 23230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.6573532349667516e-06,
+      "loss": 0.1797,
+      "step": 23235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.655926253603128e-06,
+      "loss": 0.1551,
+      "step": 23240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.654499272239505e-06,
+      "loss": 0.3179,
+      "step": 23245
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6530722908758815e-06,
+      "loss": 0.4787,
+      "step": 23250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.651645309512258e-06,
+      "loss": 0.1608,
+      "step": 23255
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6502183281486346e-06,
+      "loss": 0.1087,
+      "step": 23260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.648791346785011e-06,
+      "loss": 0.1297,
+      "step": 23265
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.647364365421388e-06,
+      "loss": 0.1842,
+      "step": 23270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6459373840577645e-06,
+      "loss": 0.0747,
+      "step": 23275
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6445104026941412e-06,
+      "loss": 0.2023,
+      "step": 23280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6430834213305176e-06,
+      "loss": 0.1042,
+      "step": 23285
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6416564399668943e-06,
+      "loss": 0.1426,
+      "step": 23290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.640229458603271e-06,
+      "loss": 0.3445,
+      "step": 23295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6388024772396474e-06,
+      "loss": 0.6512,
+      "step": 23300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.637375495876024e-06,
+      "loss": 0.1894,
+      "step": 23305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6359485145124005e-06,
+      "loss": 0.1087,
+      "step": 23310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6345215331487777e-06,
+      "loss": 0.144,
+      "step": 23315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.633094551785154e-06,
+      "loss": 0.1743,
+      "step": 23320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6316675704215304e-06,
+      "loss": 0.1719,
+      "step": 23325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.630240589057907e-06,
+      "loss": 0.0125,
+      "step": 23330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.628813607694284e-06,
+      "loss": 0.1268,
+      "step": 23335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6273866263306607e-06,
+      "loss": 0.1219,
+      "step": 23340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.625959644967037e-06,
+      "loss": 0.1726,
+      "step": 23345
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6245326636034133e-06,
+      "loss": 0.2476,
+      "step": 23350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.62310568223979e-06,
+      "loss": 0.1386,
+      "step": 23355
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.621678700876167e-06,
+      "loss": 0.1267,
+      "step": 23360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6202517195125436e-06,
+      "loss": 0.1364,
+      "step": 23365
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.61882473814892e-06,
+      "loss": 0.1671,
+      "step": 23370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6173977567852967e-06,
+      "loss": 0.1315,
+      "step": 23375
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6159707754216735e-06,
+      "loss": 0.1379,
+      "step": 23380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.61454379405805e-06,
+      "loss": 0.0916,
+      "step": 23385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6131168126944266e-06,
+      "loss": 0.1413,
+      "step": 23390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.611689831330803e-06,
+      "loss": 0.1736,
+      "step": 23395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6102628499671797e-06,
+      "loss": 0.373,
+      "step": 23400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6088358686035565e-06,
+      "loss": 0.1233,
+      "step": 23405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6074088872399332e-06,
+      "loss": 0.1222,
+      "step": 23410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6059819058763096e-06,
+      "loss": 0.1344,
+      "step": 23415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.604554924512686e-06,
+      "loss": 0.1504,
+      "step": 23420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.603127943149063e-06,
+      "loss": 0.2086,
+      "step": 23425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6017009617854394e-06,
+      "loss": 0.0836,
+      "step": 23430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.600273980421816e-06,
+      "loss": 0.1385,
+      "step": 23435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5988469990581925e-06,
+      "loss": 0.1634,
+      "step": 23440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.597420017694569e-06,
+      "loss": 0.2463,
+      "step": 23445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.595993036330946e-06,
+      "loss": 0.3586,
+      "step": 23450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5945660549673224e-06,
+      "loss": 0.1436,
+      "step": 23455
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.593139073603699e-06,
+      "loss": 0.1569,
+      "step": 23460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5917120922400755e-06,
+      "loss": 0.1627,
+      "step": 23465
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5902851108764527e-06,
+      "loss": 0.2061,
+      "step": 23470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.588858129512829e-06,
+      "loss": 0.127,
+      "step": 23475
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5874311481492053e-06,
+      "loss": 0.0794,
+      "step": 23480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.586004166785582e-06,
+      "loss": 0.0913,
+      "step": 23485
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5845771854219584e-06,
+      "loss": 0.1909,
+      "step": 23490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5831502040583356e-06,
+      "loss": 0.3552,
+      "step": 23495
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.581723222694712e-06,
+      "loss": 0.2849,
+      "step": 23500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5802962413310887e-06,
+      "loss": 0.1741,
+      "step": 23505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.578869259967465e-06,
+      "loss": 0.1317,
+      "step": 23510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5774422786038414e-06,
+      "loss": 0.1657,
+      "step": 23515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5760152972402186e-06,
+      "loss": 0.1563,
+      "step": 23520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.574588315876595e-06,
+      "loss": 0.1125,
+      "step": 23525
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5731613345129717e-06,
+      "loss": 0.1364,
+      "step": 23530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.571734353149348e-06,
+      "loss": 0.216,
+      "step": 23535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5703073717857244e-06,
+      "loss": 0.172,
+      "step": 23540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5688803904221016e-06,
+      "loss": 0.2617,
+      "step": 23545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.567453409058478e-06,
+      "loss": 0.372,
+      "step": 23550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5660264276948547e-06,
+      "loss": 0.1688,
+      "step": 23555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.564599446331231e-06,
+      "loss": 0.1386,
+      "step": 23560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.563172464967608e-06,
+      "loss": 0.1018,
+      "step": 23565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5617454836039845e-06,
+      "loss": 0.1901,
+      "step": 23570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.560318502240361e-06,
+      "loss": 0.1043,
+      "step": 23575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5588915208767376e-06,
+      "loss": 0.0829,
+      "step": 23580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.557464539513114e-06,
+      "loss": 0.1622,
+      "step": 23585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.556037558149491e-06,
+      "loss": 0.1634,
+      "step": 23590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5546105767858675e-06,
+      "loss": 0.3626,
+      "step": 23595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5531835954222443e-06,
+      "loss": 0.3029,
+      "step": 23600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.5517566140586206e-06,
+      "loss": 0.1306,
+      "step": 23605
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.550329632694997e-06,
+      "loss": 0.1171,
+      "step": 23610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.548902651331374e-06,
+      "loss": 0.1345,
+      "step": 23615
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5474756699677505e-06,
+      "loss": 0.1034,
+      "step": 23620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5460486886041272e-06,
+      "loss": 0.0366,
+      "step": 23625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5446217072405036e-06,
+      "loss": 0.0737,
+      "step": 23630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.54319472587688e-06,
+      "loss": 0.1304,
+      "step": 23635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.541767744513257e-06,
+      "loss": 0.1246,
+      "step": 23640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5403407631496334e-06,
+      "loss": 0.306,
+      "step": 23645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.53891378178601e-06,
+      "loss": 0.4532,
+      "step": 23650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5374868004223865e-06,
+      "loss": 0.284,
+      "step": 23655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5360598190587637e-06,
+      "loss": 0.1246,
+      "step": 23660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.53463283769514e-06,
+      "loss": 0.1596,
+      "step": 23665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5332058563315164e-06,
+      "loss": 0.1537,
+      "step": 23670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.531778874967893e-06,
+      "loss": 0.0664,
+      "step": 23675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5303518936042695e-06,
+      "loss": 0.2958,
+      "step": 23680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5289249122406467e-06,
+      "loss": 0.0775,
+      "step": 23685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.527497930877023e-06,
+      "loss": 0.1909,
+      "step": 23690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5260709495133998e-06,
+      "loss": 0.5231,
+      "step": 23695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.524643968149776e-06,
+      "loss": 0.3175,
+      "step": 23700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5232169867861524e-06,
+      "loss": 0.1515,
+      "step": 23705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5217900054225296e-06,
+      "loss": 0.1724,
+      "step": 23710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.520363024058906e-06,
+      "loss": 0.133,
+      "step": 23715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5189360426952827e-06,
+      "loss": 0.1605,
+      "step": 23720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.517509061331659e-06,
+      "loss": 0.1722,
+      "step": 23725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.516082079968036e-06,
+      "loss": 0.0575,
+      "step": 23730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5146550986044126e-06,
+      "loss": 0.0801,
+      "step": 23735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.513228117240789e-06,
+      "loss": 0.2827,
+      "step": 23740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5118011358771657e-06,
+      "loss": 0.1989,
+      "step": 23745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.510374154513542e-06,
+      "loss": 0.3007,
+      "step": 23750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5089471731499192e-06,
+      "loss": 0.0585,
+      "step": 23755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5075201917862956e-06,
+      "loss": 0.1668,
+      "step": 23760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.506093210422672e-06,
+      "loss": 0.1849,
+      "step": 23765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5046662290590487e-06,
+      "loss": 0.1545,
+      "step": 23770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5032392476954254e-06,
+      "loss": 0.0453,
+      "step": 23775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.501812266331802e-06,
+      "loss": 0.091,
+      "step": 23780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5003852849681785e-06,
+      "loss": 0.1137,
+      "step": 23785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4989583036045553e-06,
+      "loss": 0.1104,
+      "step": 23790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.497531322240932e-06,
+      "loss": 0.1074,
+      "step": 23795
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4961043408773084e-06,
+      "loss": 0.5004,
+      "step": 23800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.494677359513685e-06,
+      "loss": 0.0985,
+      "step": 23805
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4932503781500615e-06,
+      "loss": 0.1606,
+      "step": 23810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4918233967864382e-06,
+      "loss": 0.1426,
+      "step": 23815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.490396415422815e-06,
+      "loss": 0.1338,
+      "step": 23820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4889694340591913e-06,
+      "loss": 0.1647,
+      "step": 23825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.487542452695568e-06,
+      "loss": 0.0531,
+      "step": 23830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4861154713319444e-06,
+      "loss": 0.061,
+      "step": 23835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4846884899683216e-06,
+      "loss": 0.1268,
+      "step": 23840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.483261508604698e-06,
+      "loss": 0.166,
+      "step": 23845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4818345272410747e-06,
+      "loss": 0.5046,
+      "step": 23850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.480407545877451e-06,
+      "loss": 0.1384,
+      "step": 23855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4789805645138274e-06,
+      "loss": 0.1635,
+      "step": 23860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4775535831502046e-06,
+      "loss": 0.1346,
+      "step": 23865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.476126601786581e-06,
+      "loss": 0.1902,
+      "step": 23870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4746996204229577e-06,
+      "loss": 0.0762,
+      "step": 23875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.473272639059334e-06,
+      "loss": 0.027,
+      "step": 23880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4718456576957112e-06,
+      "loss": 0.3273,
+      "step": 23885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4704186763320876e-06,
+      "loss": 0.1416,
+      "step": 23890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.468991694968464e-06,
+      "loss": 0.4064,
+      "step": 23895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4675647136048407e-06,
+      "loss": 0.4853,
+      "step": 23900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.466137732241217e-06,
+      "loss": 0.1373,
+      "step": 23905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.464710750877594e-06,
+      "loss": 0.1968,
+      "step": 23910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4632837695139705e-06,
+      "loss": 0.1565,
+      "step": 23915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4618567881503473e-06,
+      "loss": 0.206,
+      "step": 23920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4604298067867236e-06,
+      "loss": 0.1417,
+      "step": 23925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4590028254231e-06,
+      "loss": 0.1064,
+      "step": 23930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.457575844059477e-06,
+      "loss": 0.0539,
+      "step": 23935
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4561488626958535e-06,
+      "loss": 0.1214,
+      "step": 23940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4547218813322303e-06,
+      "loss": 0.0785,
+      "step": 23945
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4532948999686066e-06,
+      "loss": 0.375,
+      "step": 23950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.451867918604983e-06,
+      "loss": 0.1294,
+      "step": 23955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.45044093724136e-06,
+      "loss": 0.1702,
+      "step": 23960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.4490139558777365e-06,
+      "loss": 0.1358,
+      "step": 23965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4475869745141132e-06,
+      "loss": 0.1221,
+      "step": 23970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4461599931504896e-06,
+      "loss": 0.1358,
+      "step": 23975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4447330117868667e-06,
+      "loss": 0.1159,
+      "step": 23980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.443306030423243e-06,
+      "loss": 0.0946,
+      "step": 23985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4418790490596194e-06,
+      "loss": 0.0759,
+      "step": 23990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.440452067695996e-06,
+      "loss": 0.2981,
+      "step": 23995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4390250863323725e-06,
+      "loss": 0.521,
+      "step": 24000
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 0.061234455555677414,
+      "eval_runtime": 1558.3951,
+      "eval_samples_per_second": 10.279,
+      "eval_steps_per_second": 2.57,
+      "eval_wer": 0.1717273710782546,
+      "step": 24000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4375981049687497e-06,
+      "loss": 0.1503,
+      "step": 24005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.436171123605126e-06,
+      "loss": 0.1167,
+      "step": 24010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.434744142241503e-06,
+      "loss": 0.1155,
+      "step": 24015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.433317160877879e-06,
+      "loss": 0.133,
+      "step": 24020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4318901795142555e-06,
+      "loss": 0.1115,
+      "step": 24025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4304631981506327e-06,
+      "loss": 0.1889,
+      "step": 24030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.429036216787009e-06,
+      "loss": 0.0863,
+      "step": 24035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4276092354233858e-06,
+      "loss": 0.3958,
+      "step": 24040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.426182254059762e-06,
+      "loss": 0.3577,
+      "step": 24045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4247552726961384e-06,
+      "loss": 0.49,
+      "step": 24050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4233282913325156e-06,
+      "loss": 0.1459,
+      "step": 24055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.421901309968892e-06,
+      "loss": 0.1008,
+      "step": 24060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4204743286052687e-06,
+      "loss": 0.1263,
+      "step": 24065
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.419047347241645e-06,
+      "loss": 0.152,
+      "step": 24070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4176203658780223e-06,
+      "loss": 0.1596,
+      "step": 24075
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4161933845143986e-06,
+      "loss": 0.0835,
+      "step": 24080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.414766403150775e-06,
+      "loss": 0.1273,
+      "step": 24085
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4133394217871517e-06,
+      "loss": 0.1407,
+      "step": 24090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.411912440423528e-06,
+      "loss": 0.2252,
+      "step": 24095
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4104854590599052e-06,
+      "loss": 0.2954,
+      "step": 24100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4090584776962816e-06,
+      "loss": 0.2298,
+      "step": 24105
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4076314963326583e-06,
+      "loss": 0.0906,
+      "step": 24110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4062045149690347e-06,
+      "loss": 0.1589,
+      "step": 24115
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.404777533605411e-06,
+      "loss": 0.173,
+      "step": 24120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.403350552241788e-06,
+      "loss": 0.113,
+      "step": 24125
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4019235708781645e-06,
+      "loss": 0.1035,
+      "step": 24130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.4004965895145413e-06,
+      "loss": 0.1338,
+      "step": 24135
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3990696081509176e-06,
+      "loss": 0.17,
+      "step": 24140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.397642626787294e-06,
+      "loss": 0.0888,
+      "step": 24145
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.396215645423671e-06,
+      "loss": 0.2657,
+      "step": 24150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3947886640600475e-06,
+      "loss": 0.1518,
+      "step": 24155
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3933616826964242e-06,
+      "loss": 0.1659,
+      "step": 24160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3919347013328006e-06,
+      "loss": 0.148,
+      "step": 24165
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3905077199691778e-06,
+      "loss": 0.2036,
+      "step": 24170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.389080738605554e-06,
+      "loss": 0.1084,
+      "step": 24175
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3876537572419305e-06,
+      "loss": 0.1029,
+      "step": 24180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3862267758783072e-06,
+      "loss": 0.0879,
+      "step": 24185
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.384799794514684e-06,
+      "loss": 0.2674,
+      "step": 24190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3833728131510607e-06,
+      "loss": 0.2432,
+      "step": 24195
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.381945831787437e-06,
+      "loss": 0.6116,
+      "step": 24200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.380518850423814e-06,
+      "loss": 0.1596,
+      "step": 24205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.37909186906019e-06,
+      "loss": 0.1456,
+      "step": 24210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.377664887696567e-06,
+      "loss": 0.1003,
+      "step": 24215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3762379063329437e-06,
+      "loss": 0.177,
+      "step": 24220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.37481092496932e-06,
+      "loss": 0.1047,
+      "step": 24225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.373383943605697e-06,
+      "loss": 0.0667,
+      "step": 24230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3719569622420736e-06,
+      "loss": 0.1686,
+      "step": 24235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.37052998087845e-06,
+      "loss": 0.0855,
+      "step": 24240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3691029995148267e-06,
+      "loss": 0.1652,
+      "step": 24245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.367676018151203e-06,
+      "loss": 0.3478,
+      "step": 24250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3662490367875798e-06,
+      "loss": 0.0967,
+      "step": 24255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3648220554239565e-06,
+      "loss": 0.1227,
+      "step": 24260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3633950740603333e-06,
+      "loss": 0.1537,
+      "step": 24265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3619680926967096e-06,
+      "loss": 0.1587,
+      "step": 24270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.360541111333086e-06,
+      "loss": 0.0498,
+      "step": 24275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.359114129969463e-06,
+      "loss": 0.0124,
+      "step": 24280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3576871486058395e-06,
+      "loss": 0.0995,
+      "step": 24285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3562601672422163e-06,
+      "loss": 0.17,
+      "step": 24290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3548331858785926e-06,
+      "loss": 0.2525,
+      "step": 24295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3534062045149694e-06,
+      "loss": 0.3695,
+      "step": 24300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.351979223151346e-06,
+      "loss": 0.1427,
+      "step": 24305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3505522417877225e-06,
+      "loss": 0.123,
+      "step": 24310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3491252604240992e-06,
+      "loss": 0.1756,
+      "step": 24315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3476982790604756e-06,
+      "loss": 0.2576,
+      "step": 24320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.3462712976968527e-06,
+      "loss": 0.1732,
+      "step": 24325
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.344844316333229e-06,
+      "loss": 0.082,
+      "step": 24330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3434173349696054e-06,
+      "loss": 0.0868,
+      "step": 24335
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.341990353605982e-06,
+      "loss": 0.2351,
+      "step": 24340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3405633722423585e-06,
+      "loss": 0.1013,
+      "step": 24345
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3391363908787357e-06,
+      "loss": 0.2214,
+      "step": 24350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.337709409515112e-06,
+      "loss": 0.1216,
+      "step": 24355
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.336282428151489e-06,
+      "loss": 0.1518,
+      "step": 24360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.334855446787865e-06,
+      "loss": 0.1282,
+      "step": 24365
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3334284654242415e-06,
+      "loss": 0.2552,
+      "step": 24370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3320014840606187e-06,
+      "loss": 0.1445,
+      "step": 24375
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.330574502696995e-06,
+      "loss": 0.1411,
+      "step": 24380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3291475213333718e-06,
+      "loss": 0.1767,
+      "step": 24385
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.327720539969748e-06,
+      "loss": 0.1734,
+      "step": 24390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3262935586061253e-06,
+      "loss": 0.2567,
+      "step": 24395
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3248665772425016e-06,
+      "loss": 0.3992,
+      "step": 24400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.323439595878878e-06,
+      "loss": 0.1647,
+      "step": 24405
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3220126145152547e-06,
+      "loss": 0.1232,
+      "step": 24410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.320585633151631e-06,
+      "loss": 0.1844,
+      "step": 24415
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3191586517880083e-06,
+      "loss": 0.1531,
+      "step": 24420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3177316704243846e-06,
+      "loss": 0.0488,
+      "step": 24425
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.316304689060761e-06,
+      "loss": 0.1118,
+      "step": 24430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3148777076971377e-06,
+      "loss": 0.2331,
+      "step": 24435
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.313450726333514e-06,
+      "loss": 0.2966,
+      "step": 24440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3120237449698912e-06,
+      "loss": 0.2844,
+      "step": 24445
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3105967636062676e-06,
+      "loss": 0.6628,
+      "step": 24450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3091697822426443e-06,
+      "loss": 0.1269,
+      "step": 24455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3077428008790207e-06,
+      "loss": 0.1566,
+      "step": 24460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.306315819515397e-06,
+      "loss": 0.2109,
+      "step": 24465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.304888838151774e-06,
+      "loss": 0.2047,
+      "step": 24470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3034618567881505e-06,
+      "loss": 0.1145,
+      "step": 24475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3020348754245273e-06,
+      "loss": 0.0569,
+      "step": 24480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.3006078940609036e-06,
+      "loss": 0.1161,
+      "step": 24485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.299180912697281e-06,
+      "loss": 0.2084,
+      "step": 24490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.297753931333657e-06,
+      "loss": 0.2207,
+      "step": 24495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2963269499700335e-06,
+      "loss": 0.6217,
+      "step": 24500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2948999686064102e-06,
+      "loss": 0.093,
+      "step": 24505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2934729872427866e-06,
+      "loss": 0.0841,
+      "step": 24510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2920460058791638e-06,
+      "loss": 0.2002,
+      "step": 24515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.29061902451554e-06,
+      "loss": 0.1339,
+      "step": 24520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.289192043151917e-06,
+      "loss": 0.1352,
+      "step": 24525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2877650617882932e-06,
+      "loss": 0.1269,
+      "step": 24530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2863380804246696e-06,
+      "loss": 0.0433,
+      "step": 24535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2849110990610467e-06,
+      "loss": 0.2801,
+      "step": 24540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.283484117697423e-06,
+      "loss": 0.2385,
+      "step": 24545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2820571363338e-06,
+      "loss": 0.2845,
+      "step": 24550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.280630154970176e-06,
+      "loss": 0.0912,
+      "step": 24555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2792031736065525e-06,
+      "loss": 0.1496,
+      "step": 24560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2777761922429297e-06,
+      "loss": 0.2442,
+      "step": 24565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.276349210879306e-06,
+      "loss": 0.1874,
+      "step": 24570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.274922229515683e-06,
+      "loss": 0.1175,
+      "step": 24575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.273495248152059e-06,
+      "loss": 0.0182,
+      "step": 24580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2720682667884363e-06,
+      "loss": 0.2497,
+      "step": 24585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2706412854248127e-06,
+      "loss": 0.1674,
+      "step": 24590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.269214304061189e-06,
+      "loss": 0.1815,
+      "step": 24595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2677873226975658e-06,
+      "loss": 0.37,
+      "step": 24600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.266360341333942e-06,
+      "loss": 0.1292,
+      "step": 24605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2649333599703193e-06,
+      "loss": 0.1061,
+      "step": 24610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2635063786066956e-06,
+      "loss": 0.1565,
+      "step": 24615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2620793972430724e-06,
+      "loss": 0.1779,
+      "step": 24620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2606524158794487e-06,
+      "loss": 0.1077,
+      "step": 24625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2592254345158255e-06,
+      "loss": 0.0938,
+      "step": 24630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2577984531522023e-06,
+      "loss": 0.1082,
+      "step": 24635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2563714717885786e-06,
+      "loss": 0.0819,
+      "step": 24640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2549444904249554e-06,
+      "loss": 0.4498,
+      "step": 24645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2535175090613317e-06,
+      "loss": 0.6766,
+      "step": 24650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2520905276977085e-06,
+      "loss": 0.1111,
+      "step": 24655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2506635463340852e-06,
+      "loss": 0.1768,
+      "step": 24660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2492365649704616e-06,
+      "loss": 0.1568,
+      "step": 24665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.2478095836068383e-06,
+      "loss": 0.1591,
+      "step": 24670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.246382602243215e-06,
+      "loss": 0.1138,
+      "step": 24675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.244955620879592e-06,
+      "loss": 0.1041,
+      "step": 24680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.243528639515968e-06,
+      "loss": 0.0842,
+      "step": 24685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2421016581523445e-06,
+      "loss": 0.181,
+      "step": 24690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2406746767887217e-06,
+      "loss": 0.1652,
+      "step": 24695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.239247695425098e-06,
+      "loss": 0.4201,
+      "step": 24700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.237820714061475e-06,
+      "loss": 0.1204,
+      "step": 24705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.236393732697851e-06,
+      "loss": 0.1886,
+      "step": 24710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.234966751334228e-06,
+      "loss": 0.1273,
+      "step": 24715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2335397699706047e-06,
+      "loss": 0.168,
+      "step": 24720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.232112788606981e-06,
+      "loss": 0.2194,
+      "step": 24725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2306858072433578e-06,
+      "loss": 0.1397,
+      "step": 24730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.229258825879734e-06,
+      "loss": 0.1028,
+      "step": 24735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2278318445161113e-06,
+      "loss": 0.2445,
+      "step": 24740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2264048631524876e-06,
+      "loss": 0.1483,
+      "step": 24745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.224977881788864e-06,
+      "loss": 0.4142,
+      "step": 24750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2235509004252407e-06,
+      "loss": 0.1048,
+      "step": 24755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.222123919061617e-06,
+      "loss": 0.1253,
+      "step": 24760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2206969376979943e-06,
+      "loss": 0.1067,
+      "step": 24765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2192699563343706e-06,
+      "loss": 0.1786,
+      "step": 24770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2178429749707474e-06,
+      "loss": 0.1778,
+      "step": 24775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2164159936071237e-06,
+      "loss": 0.1355,
+      "step": 24780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2149890122435e-06,
+      "loss": 0.1129,
+      "step": 24785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2135620308798772e-06,
+      "loss": 0.1761,
+      "step": 24790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2121350495162536e-06,
+      "loss": 0.2162,
+      "step": 24795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2107080681526303e-06,
+      "loss": 0.4048,
+      "step": 24800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2092810867890067e-06,
+      "loss": 0.1602,
+      "step": 24805
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.207854105425384e-06,
+      "loss": 0.1191,
+      "step": 24810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.20642712406176e-06,
+      "loss": 0.1368,
+      "step": 24815
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2050001426981365e-06,
+      "loss": 0.1553,
+      "step": 24820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2035731613345133e-06,
+      "loss": 0.0768,
+      "step": 24825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2021461799708896e-06,
+      "loss": 0.0305,
+      "step": 24830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.200719198607267e-06,
+      "loss": 0.2172,
+      "step": 24835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.199292217243643e-06,
+      "loss": 0.2116,
+      "step": 24840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1978652358800195e-06,
+      "loss": 0.1883,
+      "step": 24845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1964382545163963e-06,
+      "loss": 0.3431,
+      "step": 24850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1950112731527726e-06,
+      "loss": 0.1449,
+      "step": 24855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1935842917891498e-06,
+      "loss": 0.1767,
+      "step": 24860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.192157310425526e-06,
+      "loss": 0.1833,
+      "step": 24865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.190730329061903e-06,
+      "loss": 0.163,
+      "step": 24870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1893033476982792e-06,
+      "loss": 0.1874,
+      "step": 24875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1878763663346556e-06,
+      "loss": 0.1119,
+      "step": 24880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1864493849710327e-06,
+      "loss": 0.1364,
+      "step": 24885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.185022403607409e-06,
+      "loss": 0.1162,
+      "step": 24890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.183595422243786e-06,
+      "loss": 0.2754,
+      "step": 24895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.182168440880162e-06,
+      "loss": 0.4152,
+      "step": 24900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1807414595165394e-06,
+      "loss": 0.0828,
+      "step": 24905
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1793144781529157e-06,
+      "loss": 0.1073,
+      "step": 24910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.177887496789292e-06,
+      "loss": 0.0854,
+      "step": 24915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.176460515425669e-06,
+      "loss": 0.1355,
+      "step": 24920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.175033534062045e-06,
+      "loss": 0.1573,
+      "step": 24925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1736065526984223e-06,
+      "loss": 0.1358,
+      "step": 24930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1721795713347987e-06,
+      "loss": 0.0669,
+      "step": 24935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.170752589971175e-06,
+      "loss": 0.1749,
+      "step": 24940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1693256086075518e-06,
+      "loss": 0.1238,
+      "step": 24945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1681840235166534e-06,
+      "loss": 0.5303,
+      "step": 24950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1667570421530298e-06,
+      "loss": 0.1184,
+      "step": 24955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.165330060789406e-06,
+      "loss": 0.2232,
+      "step": 24960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.163903079425783e-06,
+      "loss": 0.1473,
+      "step": 24965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.162476098062159e-06,
+      "loss": 0.2284,
+      "step": 24970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1610491166985364e-06,
+      "loss": 0.1185,
+      "step": 24975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1596221353349127e-06,
+      "loss": 0.111,
+      "step": 24980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1581951539712895e-06,
+      "loss": 0.1202,
+      "step": 24985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.156768172607666e-06,
+      "loss": 0.125,
+      "step": 24990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.155341191244042e-06,
+      "loss": 0.2482,
+      "step": 24995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1539142098804194e-06,
+      "loss": 0.3502,
+      "step": 25000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1524872285167957e-06,
+      "loss": 0.1993,
+      "step": 25005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1510602471531725e-06,
+      "loss": 0.0801,
+      "step": 25010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.149633265789549e-06,
+      "loss": 0.166,
+      "step": 25015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.148206284425926e-06,
+      "loss": 0.2255,
+      "step": 25020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1467793030623023e-06,
+      "loss": 0.131,
+      "step": 25025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1453523216986787e-06,
+      "loss": 0.0902,
+      "step": 25030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1439253403350554e-06,
+      "loss": 0.0461,
+      "step": 25035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.1424983589714318e-06,
+      "loss": 0.2501,
+      "step": 25040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.141071377607809e-06,
+      "loss": 0.184,
+      "step": 25045
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1396443962441853e-06,
+      "loss": 0.4096,
+      "step": 25050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1382174148805616e-06,
+      "loss": 0.1553,
+      "step": 25055
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1367904335169384e-06,
+      "loss": 0.1454,
+      "step": 25060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.135363452153315e-06,
+      "loss": 0.114,
+      "step": 25065
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.133936470789692e-06,
+      "loss": 0.1724,
+      "step": 25070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1325094894260682e-06,
+      "loss": 0.028,
+      "step": 25075
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.131082508062445e-06,
+      "loss": 0.1044,
+      "step": 25080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1296555266988213e-06,
+      "loss": 0.0878,
+      "step": 25085
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.128228545335198e-06,
+      "loss": 0.274,
+      "step": 25090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.126801563971575e-06,
+      "loss": 0.2244,
+      "step": 25095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1253745826079512e-06,
+      "loss": 0.6099,
+      "step": 25100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.123947601244328e-06,
+      "loss": 0.1223,
+      "step": 25105
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1225206198807047e-06,
+      "loss": 0.1702,
+      "step": 25110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1210936385170815e-06,
+      "loss": 0.1585,
+      "step": 25115
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.119666657153458e-06,
+      "loss": 0.2234,
+      "step": 25120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.118239675789834e-06,
+      "loss": 0.1095,
+      "step": 25125
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.116812694426211e-06,
+      "loss": 0.108,
+      "step": 25130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1153857130625877e-06,
+      "loss": 0.1036,
+      "step": 25135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1139587316989645e-06,
+      "loss": 0.3185,
+      "step": 25140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.112531750335341e-06,
+      "loss": 0.2863,
+      "step": 25145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1111047689717176e-06,
+      "loss": 0.3655,
+      "step": 25150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1096777876080943e-06,
+      "loss": 0.0881,
+      "step": 25155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1082508062444707e-06,
+      "loss": 0.1322,
+      "step": 25160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1068238248808474e-06,
+      "loss": 0.1561,
+      "step": 25165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1053968435172238e-06,
+      "loss": 0.1948,
+      "step": 25170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.103969862153601e-06,
+      "loss": 0.1177,
+      "step": 25175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1025428807899773e-06,
+      "loss": 0.1238,
+      "step": 25180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1011158994263536e-06,
+      "loss": 0.1788,
+      "step": 25185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0996889180627304e-06,
+      "loss": 0.0869,
+      "step": 25190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0982619366991067e-06,
+      "loss": 0.1081,
+      "step": 25195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.096834955335484e-06,
+      "loss": 0.4041,
+      "step": 25200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0954079739718603e-06,
+      "loss": 0.1634,
+      "step": 25205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.093980992608237e-06,
+      "loss": 0.1709,
+      "step": 25210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0925540112446134e-06,
+      "loss": 0.1343,
+      "step": 25215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0911270298809897e-06,
+      "loss": 0.153,
+      "step": 25220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.089700048517367e-06,
+      "loss": 0.0494,
+      "step": 25225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0882730671537432e-06,
+      "loss": 0.1105,
+      "step": 25230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.08684608579012e-06,
+      "loss": 0.041,
+      "step": 25235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0854191044264963e-06,
+      "loss": 0.1833,
+      "step": 25240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0839921230628735e-06,
+      "loss": 0.1466,
+      "step": 25245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.08256514169925e-06,
+      "loss": 0.44,
+      "step": 25250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.081138160335626e-06,
+      "loss": 0.1542,
+      "step": 25255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.079711178972003e-06,
+      "loss": 0.1108,
+      "step": 25260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0782841976083793e-06,
+      "loss": 0.1412,
+      "step": 25265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0768572162447565e-06,
+      "loss": 0.1025,
+      "step": 25270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.075430234881133e-06,
+      "loss": 0.0943,
+      "step": 25275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.074003253517509e-06,
+      "loss": 0.0713,
+      "step": 25280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.072576272153886e-06,
+      "loss": 0.0824,
+      "step": 25285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0711492907902622e-06,
+      "loss": 0.0725,
+      "step": 25290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0697223094266394e-06,
+      "loss": 0.3779,
+      "step": 25295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0682953280630158e-06,
+      "loss": 0.3996,
+      "step": 25300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0668683466993925e-06,
+      "loss": 0.1536,
+      "step": 25305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.065441365335769e-06,
+      "loss": 0.1816,
+      "step": 25310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.064014383972145e-06,
+      "loss": 0.1062,
+      "step": 25315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0625874026085224e-06,
+      "loss": 0.2004,
+      "step": 25320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0611604212448987e-06,
+      "loss": 0.1371,
+      "step": 25325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0597334398812755e-06,
+      "loss": 0.0809,
+      "step": 25330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.058306458517652e-06,
+      "loss": 0.1249,
+      "step": 25335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.056879477154029e-06,
+      "loss": 0.2874,
+      "step": 25340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0554524957904054e-06,
+      "loss": 0.2563,
+      "step": 25345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0540255144267817e-06,
+      "loss": 0.3192,
+      "step": 25350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0525985330631585e-06,
+      "loss": 0.2538,
+      "step": 25355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.051171551699535e-06,
+      "loss": 0.157,
+      "step": 25360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.049744570335912e-06,
+      "loss": 0.1524,
+      "step": 25365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0483175889722883e-06,
+      "loss": 0.1248,
+      "step": 25370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0468906076086647e-06,
+      "loss": 0.1673,
+      "step": 25375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0454636262450414e-06,
+      "loss": 0.0247,
+      "step": 25380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0440366448814178e-06,
+      "loss": 0.2293,
+      "step": 25385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.042609663517795e-06,
+      "loss": 0.184,
+      "step": 25390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0411826821541713e-06,
+      "loss": 0.1514,
+      "step": 25395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.039755700790548e-06,
+      "loss": 0.383,
+      "step": 25400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0383287194269244e-06,
+      "loss": 0.0815,
+      "step": 25405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0369017380633007e-06,
+      "loss": 0.1618,
+      "step": 25410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.035474756699678e-06,
+      "loss": 0.1624,
+      "step": 25415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0340477753360542e-06,
+      "loss": 0.1379,
+      "step": 25420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.032620793972431e-06,
+      "loss": 0.0769,
+      "step": 25425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0311938126088074e-06,
+      "loss": 0.0432,
+      "step": 25430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0297668312451845e-06,
+      "loss": 0.1571,
+      "step": 25435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.028339849881561e-06,
+      "loss": 0.1198,
+      "step": 25440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0269128685179372e-06,
+      "loss": 0.1957,
+      "step": 25445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.025485887154314e-06,
+      "loss": 0.292,
+      "step": 25450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0240589057906903e-06,
+      "loss": 0.108,
+      "step": 25455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0226319244270675e-06,
+      "loss": 0.1,
+      "step": 25460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.021204943063444e-06,
+      "loss": 0.1045,
+      "step": 25465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.01977796169982e-06,
+      "loss": 0.1527,
+      "step": 25470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.018350980336197e-06,
+      "loss": 0.1147,
+      "step": 25475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0169239989725733e-06,
+      "loss": 0.043,
+      "step": 25480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0154970176089505e-06,
+      "loss": 0.0107,
+      "step": 25485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.014070036245327e-06,
+      "loss": 0.1393,
+      "step": 25490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0126430548817036e-06,
+      "loss": 0.2449,
+      "step": 25495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.01121607351808e-06,
+      "loss": 0.3843,
+      "step": 25500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0097890921544567e-06,
+      "loss": 0.0952,
+      "step": 25505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0083621107908334e-06,
+      "loss": 0.1409,
+      "step": 25510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0069351294272098e-06,
+      "loss": 0.1487,
+      "step": 25515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0055081480635865e-06,
+      "loss": 0.151,
+      "step": 25520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0040811666999633e-06,
+      "loss": 0.1024,
+      "step": 25525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.00265418533634e-06,
+      "loss": 0.1171,
+      "step": 25530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0012272039727164e-06,
+      "loss": 0.1517,
+      "step": 25535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9998002226090927e-06,
+      "loss": 0.3867,
+      "step": 25540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9983732412454695e-06,
+      "loss": 0.2584,
+      "step": 25545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9969462598818463e-06,
+      "loss": 0.4842,
+      "step": 25550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.995519278518223e-06,
+      "loss": 0.1562,
+      "step": 25555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9940922971545994e-06,
+      "loss": 0.1135,
+      "step": 25560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9926653157909757e-06,
+      "loss": 0.1413,
+      "step": 25565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.991238334427353e-06,
+      "loss": 0.1179,
+      "step": 25570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9898113530637292e-06,
+      "loss": 0.1406,
+      "step": 25575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.988384371700106e-06,
+      "loss": 0.1572,
+      "step": 25580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9869573903364823e-06,
+      "loss": 0.1412,
+      "step": 25585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.985530408972859e-06,
+      "loss": 0.266,
+      "step": 25590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.984103427609236e-06,
+      "loss": 0.3005,
+      "step": 25595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.982676446245612e-06,
+      "loss": 0.549,
+      "step": 25600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.981249464881989e-06,
+      "loss": 0.0852,
+      "step": 25605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9798224835183653e-06,
+      "loss": 0.1306,
+      "step": 25610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9783955021547425e-06,
+      "loss": 0.0573,
+      "step": 25615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.976968520791119e-06,
+      "loss": 0.1642,
+      "step": 25620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9755415394274956e-06,
+      "loss": 0.1541,
+      "step": 25625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.974114558063872e-06,
+      "loss": 0.0305,
+      "step": 25630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9726875767002482e-06,
+      "loss": 0.2669,
+      "step": 25635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9712605953366254e-06,
+      "loss": 0.2604,
+      "step": 25640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9698336139730018e-06,
+      "loss": 0.1894,
+      "step": 25645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9684066326093785e-06,
+      "loss": 0.5792,
+      "step": 25650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.966979651245755e-06,
+      "loss": 0.1189,
+      "step": 25655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.965552669882131e-06,
+      "loss": 0.1772,
+      "step": 25660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9641256885185084e-06,
+      "loss": 0.1495,
+      "step": 25665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9626987071548847e-06,
+      "loss": 0.1614,
+      "step": 25670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9612717257912615e-06,
+      "loss": 0.0493,
+      "step": 25675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.959844744427638e-06,
+      "loss": 0.0841,
+      "step": 25680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.958417763064015e-06,
+      "loss": 0.1262,
+      "step": 25685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9569907817003914e-06,
+      "loss": 0.2289,
+      "step": 25690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9555638003367677e-06,
+      "loss": 0.1331,
+      "step": 25695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9541368189731445e-06,
+      "loss": 0.2359,
+      "step": 25700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.952709837609521e-06,
+      "loss": 0.0779,
+      "step": 25705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.951282856245898e-06,
+      "loss": 0.115,
+      "step": 25710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9498558748822743e-06,
+      "loss": 0.1122,
+      "step": 25715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.948428893518651e-06,
+      "loss": 0.1536,
+      "step": 25720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9470019121550274e-06,
+      "loss": 0.1202,
+      "step": 25725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9455749307914038e-06,
+      "loss": 0.0517,
+      "step": 25730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.944147949427781e-06,
+      "loss": 0.1488,
+      "step": 25735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9427209680641573e-06,
+      "loss": 0.2269,
+      "step": 25740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.941293986700534e-06,
+      "loss": 0.353,
+      "step": 25745
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9398670053369104e-06,
+      "loss": 0.2729,
+      "step": 25750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9384400239732876e-06,
+      "loss": 0.1162,
+      "step": 25755
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.937013042609664e-06,
+      "loss": 0.09,
+      "step": 25760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9355860612460403e-06,
+      "loss": 0.125,
+      "step": 25765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.934159079882417e-06,
+      "loss": 0.1344,
+      "step": 25770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9327320985187934e-06,
+      "loss": 0.1441,
+      "step": 25775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9313051171551705e-06,
+      "loss": 0.075,
+      "step": 25780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.929878135791547e-06,
+      "loss": 0.1001,
+      "step": 25785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9284511544279232e-06,
+      "loss": 0.2468,
+      "step": 25790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9270241730643e-06,
+      "loss": 0.2485,
+      "step": 25795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9255971917006763e-06,
+      "loss": 0.4041,
+      "step": 25800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9241702103370535e-06,
+      "loss": 0.1131,
+      "step": 25805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.92274322897343e-06,
+      "loss": 0.1758,
+      "step": 25810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9213162476098066e-06,
+      "loss": 0.1028,
+      "step": 25815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.919889266246183e-06,
+      "loss": 0.1707,
+      "step": 25820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9184622848825593e-06,
+      "loss": 0.0508,
+      "step": 25825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9170353035189365e-06,
+      "loss": 0.1092,
+      "step": 25830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.915608322155313e-06,
+      "loss": 0.1646,
+      "step": 25835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9141813407916896e-06,
+      "loss": 0.2093,
+      "step": 25840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.912754359428066e-06,
+      "loss": 0.256,
+      "step": 25845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.911327378064443e-06,
+      "loss": 0.1588,
+      "step": 25850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9099003967008194e-06,
+      "loss": 0.1443,
+      "step": 25855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9084734153371958e-06,
+      "loss": 0.1734,
+      "step": 25860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9070464339735725e-06,
+      "loss": 0.1045,
+      "step": 25865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.905619452609949e-06,
+      "loss": 0.2762,
+      "step": 25870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.904192471246326e-06,
+      "loss": 0.0354,
+      "step": 25875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9027654898827024e-06,
+      "loss": 0.1832,
+      "step": 25880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9013385085190787e-06,
+      "loss": 0.0922,
+      "step": 25885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8999115271554555e-06,
+      "loss": 0.111,
+      "step": 25890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.898484545791832e-06,
+      "loss": 0.1338,
+      "step": 25895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.897057564428209e-06,
+      "loss": 0.4749,
+      "step": 25900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8956305830645854e-06,
+      "loss": 0.1815,
+      "step": 25905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.894203601700962e-06,
+      "loss": 0.1365,
+      "step": 25910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8927766203373385e-06,
+      "loss": 0.1343,
+      "step": 25915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8913496389737152e-06,
+      "loss": 0.1013,
+      "step": 25920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.889922657610092e-06,
+      "loss": 0.12,
+      "step": 25925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8884956762464683e-06,
+      "loss": 0.1476,
+      "step": 25930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.887068694882845e-06,
+      "loss": 0.1328,
+      "step": 25935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8856417135192214e-06,
+      "loss": 0.1709,
+      "step": 25940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8842147321555986e-06,
+      "loss": 0.123,
+      "step": 25945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.882787750791975e-06,
+      "loss": 0.5435,
+      "step": 25950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8813607694283513e-06,
+      "loss": 0.117,
+      "step": 25955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.879933788064728e-06,
+      "loss": 0.1596,
+      "step": 25960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.878506806701105e-06,
+      "loss": 0.1398,
+      "step": 25965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8770798253374816e-06,
+      "loss": 0.2197,
+      "step": 25970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.875652843973858e-06,
+      "loss": 0.1027,
+      "step": 25975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8742258626102342e-06,
+      "loss": 0.043,
+      "step": 25980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.872798881246611e-06,
+      "loss": 0.0459,
+      "step": 25985
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8713718998829878e-06,
+      "loss": 0.0957,
+      "step": 25990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8699449185193645e-06,
+      "loss": 0.2356,
+      "step": 25995
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.868517937155741e-06,
+      "loss": 0.284,
+      "step": 26000
+    },
+    {
+      "epoch": 0.72,
+      "eval_loss": 0.05732240900397301,
+      "eval_runtime": 1599.5138,
+      "eval_samples_per_second": 10.014,
+      "eval_steps_per_second": 2.504,
+      "eval_wer": 0.16871619184998196,
+      "step": 26000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8670909557921176e-06,
+      "loss": 0.1305,
+      "step": 26005
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8656639744284944e-06,
+      "loss": 0.2248,
+      "step": 26010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8642369930648707e-06,
+      "loss": 0.1827,
+      "step": 26015
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8628100117012475e-06,
+      "loss": 0.1403,
+      "step": 26020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.861383030337624e-06,
+      "loss": 0.2165,
+      "step": 26025
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.859956048974001e-06,
+      "loss": 0.0734,
+      "step": 26030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8585290676103774e-06,
+      "loss": 0.1511,
+      "step": 26035
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.857102086246754e-06,
+      "loss": 0.2121,
+      "step": 26040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8556751048831305e-06,
+      "loss": 0.1674,
+      "step": 26045
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.854248123519507e-06,
+      "loss": 0.3052,
+      "step": 26050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.852821142155884e-06,
+      "loss": 0.1277,
+      "step": 26055
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8513941607922603e-06,
+      "loss": 0.1229,
+      "step": 26060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.849967179428637e-06,
+      "loss": 0.1208,
+      "step": 26065
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8485401980650134e-06,
+      "loss": 0.17,
+      "step": 26070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8471132167013898e-06,
+      "loss": 0.1901,
+      "step": 26075
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.845686235337767e-06,
+      "loss": 0.1311,
+      "step": 26080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8442592539741433e-06,
+      "loss": 0.0815,
+      "step": 26085
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.84283227261052e-06,
+      "loss": 0.1416,
+      "step": 26090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8414052912468964e-06,
+      "loss": 0.1607,
+      "step": 26095
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8399783098832736e-06,
+      "loss": 0.5613,
+      "step": 26100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.83855132851965e-06,
+      "loss": 0.1265,
+      "step": 26105
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8371243471560263e-06,
+      "loss": 0.1373,
+      "step": 26110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.835697365792403e-06,
+      "loss": 0.158,
+      "step": 26115
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8342703844287794e-06,
+      "loss": 0.1546,
+      "step": 26120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8328434030651565e-06,
+      "loss": 0.1494,
+      "step": 26125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.831416421701533e-06,
+      "loss": 0.0512,
+      "step": 26130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8299894403379096e-06,
+      "loss": 0.0766,
+      "step": 26135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.828562458974286e-06,
+      "loss": 0.2311,
+      "step": 26140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8271354776106623e-06,
+      "loss": 0.1519,
+      "step": 26145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8257084962470395e-06,
+      "loss": 0.5217,
+      "step": 26150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.824281514883416e-06,
+      "loss": 0.1174,
+      "step": 26155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8228545335197926e-06,
+      "loss": 0.1366,
+      "step": 26160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.821427552156169e-06,
+      "loss": 0.249,
+      "step": 26165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8200005707925453e-06,
+      "loss": 0.1411,
+      "step": 26170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8185735894289225e-06,
+      "loss": 0.0867,
+      "step": 26175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.817146608065299e-06,
+      "loss": 0.0346,
+      "step": 26180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8157196267016756e-06,
+      "loss": 0.0907,
+      "step": 26185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.814292645338052e-06,
+      "loss": 0.2065,
+      "step": 26190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.812865663974429e-06,
+      "loss": 0.1469,
+      "step": 26195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8114386826108054e-06,
+      "loss": 0.4471,
+      "step": 26200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8100117012471818e-06,
+      "loss": 0.1199,
+      "step": 26205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8085847198835585e-06,
+      "loss": 0.1215,
+      "step": 26210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.807157738519935e-06,
+      "loss": 0.1368,
+      "step": 26215
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.805730757156312e-06,
+      "loss": 0.1158,
+      "step": 26220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8043037757926884e-06,
+      "loss": 0.1989,
+      "step": 26225
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.802876794429065e-06,
+      "loss": 0.2644,
+      "step": 26230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8014498130654415e-06,
+      "loss": 0.1495,
+      "step": 26235
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.800022831701818e-06,
+      "loss": 0.0975,
+      "step": 26240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.798595850338195e-06,
+      "loss": 0.1118,
+      "step": 26245
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7971688689745714e-06,
+      "loss": 0.4682,
+      "step": 26250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.795741887610948e-06,
+      "loss": 0.17,
+      "step": 26255
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7943149062473245e-06,
+      "loss": 0.1316,
+      "step": 26260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.792887924883701e-06,
+      "loss": 0.1506,
+      "step": 26265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.791460943520078e-06,
+      "loss": 0.211,
+      "step": 26270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7900339621564543e-06,
+      "loss": 0.1409,
+      "step": 26275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.788606980792831e-06,
+      "loss": 0.034,
+      "step": 26280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7871799994292074e-06,
+      "loss": 0.1001,
+      "step": 26285
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7857530180655846e-06,
+      "loss": 0.0589,
+      "step": 26290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784326036701961e-06,
+      "loss": 0.2897,
+      "step": 26295
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7828990553383373e-06,
+      "loss": 0.2913,
+      "step": 26300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.781472073974714e-06,
+      "loss": 0.1237,
+      "step": 26305
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7800450926110904e-06,
+      "loss": 0.1621,
+      "step": 26310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7786181112474676e-06,
+      "loss": 0.1255,
+      "step": 26315
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.777191129883844e-06,
+      "loss": 0.1752,
+      "step": 26320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7757641485202207e-06,
+      "loss": 0.1037,
+      "step": 26325
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.774337167156597e-06,
+      "loss": 0.101,
+      "step": 26330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7729101857929733e-06,
+      "loss": 0.2034,
+      "step": 26335
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7714832044293505e-06,
+      "loss": 0.1355,
+      "step": 26340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.770056223065727e-06,
+      "loss": 0.2262,
+      "step": 26345
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7686292417021036e-06,
+      "loss": 0.2849,
+      "step": 26350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.76720226033848e-06,
+      "loss": 0.0806,
+      "step": 26355
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.765775278974857e-06,
+      "loss": 0.1381,
+      "step": 26360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7643482976112335e-06,
+      "loss": 0.1695,
+      "step": 26365
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.76292131624761e-06,
+      "loss": 0.1678,
+      "step": 26370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7614943348839866e-06,
+      "loss": 0.1797,
+      "step": 26375
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.760067353520363e-06,
+      "loss": 0.1025,
+      "step": 26380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.75864037215674e-06,
+      "loss": 0.2294,
+      "step": 26385
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7572133907931165e-06,
+      "loss": 0.2152,
+      "step": 26390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.755786409429493e-06,
+      "loss": 0.1731,
+      "step": 26395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7543594280658696e-06,
+      "loss": 0.3767,
+      "step": 26400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7529324467022463e-06,
+      "loss": 0.0844,
+      "step": 26405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.751505465338623e-06,
+      "loss": 0.1569,
+      "step": 26410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7500784839749994e-06,
+      "loss": 0.2803,
+      "step": 26415
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.748651502611376e-06,
+      "loss": 0.1608,
+      "step": 26420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.747224521247753e-06,
+      "loss": 0.1184,
+      "step": 26425
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7457975398841293e-06,
+      "loss": 0.0333,
+      "step": 26430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.744370558520506e-06,
+      "loss": 0.0804,
+      "step": 26435
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7429435771568824e-06,
+      "loss": 0.2207,
+      "step": 26440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.741516595793259e-06,
+      "loss": 0.231,
+      "step": 26445
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.740089614429636e-06,
+      "loss": 0.4283,
+      "step": 26450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7386626330660127e-06,
+      "loss": 0.1311,
+      "step": 26455
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.737235651702389e-06,
+      "loss": 0.1237,
+      "step": 26460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7358086703387654e-06,
+      "loss": 0.1065,
+      "step": 26465
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7343816889751425e-06,
+      "loss": 0.2122,
+      "step": 26470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.732954707611519e-06,
+      "loss": 0.0672,
+      "step": 26475
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7315277262478956e-06,
+      "loss": 0.18,
+      "step": 26480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.730100744884272e-06,
+      "loss": 0.276,
+      "step": 26485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7286737635206483e-06,
+      "loss": 0.2462,
+      "step": 26490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7272467821570255e-06,
+      "loss": 0.0714,
+      "step": 26495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.725819800793402e-06,
+      "loss": 0.2961,
+      "step": 26500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7243928194297786e-06,
+      "loss": 0.1474,
+      "step": 26505
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.722965838066155e-06,
+      "loss": 0.118,
+      "step": 26510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.721538856702532e-06,
+      "loss": 0.1449,
+      "step": 26515
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7201118753389085e-06,
+      "loss": 0.1919,
+      "step": 26520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.718684893975285e-06,
+      "loss": 0.1371,
+      "step": 26525
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7172579126116616e-06,
+      "loss": 0.1795,
+      "step": 26530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.715830931248038e-06,
+      "loss": 0.0662,
+      "step": 26535
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.714403949884415e-06,
+      "loss": 0.0727,
+      "step": 26540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7129769685207914e-06,
+      "loss": 0.218,
+      "step": 26545
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.711549987157168e-06,
+      "loss": 0.2782,
+      "step": 26550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7101230057935445e-06,
+      "loss": 0.0953,
+      "step": 26555
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.708696024429921e-06,
+      "loss": 0.1086,
+      "step": 26560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.707269043066298e-06,
+      "loss": 0.129,
+      "step": 26565
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7058420617026744e-06,
+      "loss": 0.1382,
+      "step": 26570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.704415080339051e-06,
+      "loss": 0.1681,
+      "step": 26575
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7029880989754275e-06,
+      "loss": 0.0586,
+      "step": 26580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.701561117611804e-06,
+      "loss": 0.0819,
+      "step": 26585
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.700134136248181e-06,
+      "loss": 0.0729,
+      "step": 26590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6987071548845574e-06,
+      "loss": 0.1718,
+      "step": 26595
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.697280173520934e-06,
+      "loss": 0.7693,
+      "step": 26600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6958531921573105e-06,
+      "loss": 0.1223,
+      "step": 26605
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6944262107936876e-06,
+      "loss": 0.1018,
+      "step": 26610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.692999229430064e-06,
+      "loss": 0.1375,
+      "step": 26615
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6915722480664403e-06,
+      "loss": 0.1235,
+      "step": 26620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.690145266702817e-06,
+      "loss": 0.0849,
+      "step": 26625
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6887182853391934e-06,
+      "loss": 0.185,
+      "step": 26630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6872913039755706e-06,
+      "loss": 0.0632,
+      "step": 26635
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.685864322611947e-06,
+      "loss": 0.1054,
+      "step": 26640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6844373412483237e-06,
+      "loss": 0.1819,
+      "step": 26645
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6830103598847e-06,
+      "loss": 0.1812,
+      "step": 26650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6815833785210764e-06,
+      "loss": 0.1602,
+      "step": 26655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6801563971574536e-06,
+      "loss": 0.147,
+      "step": 26660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.67872941579383e-06,
+      "loss": 0.1197,
+      "step": 26665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6773024344302067e-06,
+      "loss": 0.1357,
+      "step": 26670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.675875453066583e-06,
+      "loss": 0.1372,
+      "step": 26675
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6744484717029593e-06,
+      "loss": 0.0134,
+      "step": 26680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6730214903393365e-06,
+      "loss": 0.0678,
+      "step": 26685
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.671594508975713e-06,
+      "loss": 0.0896,
+      "step": 26690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6701675276120896e-06,
+      "loss": 0.2949,
+      "step": 26695
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.668740546248466e-06,
+      "loss": 0.4508,
+      "step": 26700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.667313564884843e-06,
+      "loss": 0.1432,
+      "step": 26705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6658865835212195e-06,
+      "loss": 0.156,
+      "step": 26710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.664459602157596e-06,
+      "loss": 0.1392,
+      "step": 26715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6630326207939726e-06,
+      "loss": 0.1732,
+      "step": 26720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.661605639430349e-06,
+      "loss": 0.0862,
+      "step": 26725
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.660178658066726e-06,
+      "loss": 0.0326,
+      "step": 26730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6587516767031025e-06,
+      "loss": 0.0985,
+      "step": 26735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6573246953394792e-06,
+      "loss": 0.1581,
+      "step": 26740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6558977139758556e-06,
+      "loss": 0.1858,
+      "step": 26745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.654470732612232e-06,
+      "loss": 0.4903,
+      "step": 26750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.653043751248609e-06,
+      "loss": 0.1636,
+      "step": 26755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6516167698849854e-06,
+      "loss": 0.1551,
+      "step": 26760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.650189788521362e-06,
+      "loss": 0.1687,
+      "step": 26765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6487628071577385e-06,
+      "loss": 0.1462,
+      "step": 26770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6473358257941153e-06,
+      "loss": 0.1929,
+      "step": 26775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.645908844430492e-06,
+      "loss": 0.1037,
+      "step": 26780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6444818630668684e-06,
+      "loss": 0.2719,
+      "step": 26785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.643054881703245e-06,
+      "loss": 0.1249,
+      "step": 26790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6416279003396215e-06,
+      "loss": 0.1853,
+      "step": 26795
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6402009189759987e-06,
+      "loss": 0.4481,
+      "step": 26800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.638773937612375e-06,
+      "loss": 0.1416,
+      "step": 26805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6373469562487514e-06,
+      "loss": 0.1283,
+      "step": 26810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.635919974885128e-06,
+      "loss": 0.1811,
+      "step": 26815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.634492993521505e-06,
+      "loss": 0.1281,
+      "step": 26820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6330660121578816e-06,
+      "loss": 0.1362,
+      "step": 26825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.631639030794258e-06,
+      "loss": 0.0721,
+      "step": 26830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6302120494306347e-06,
+      "loss": 0.0757,
+      "step": 26835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.628785068067011e-06,
+      "loss": 0.2048,
+      "step": 26840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.627358086703388e-06,
+      "loss": 0.0996,
+      "step": 26845
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6259311053397646e-06,
+      "loss": 0.3021,
+      "step": 26850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.624504123976141e-06,
+      "loss": 0.1129,
+      "step": 26855
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6230771426125177e-06,
+      "loss": 0.1188,
+      "step": 26860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6216501612488945e-06,
+      "loss": 0.1518,
+      "step": 26865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.620223179885271e-06,
+      "loss": 0.2,
+      "step": 26870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6187961985216476e-06,
+      "loss": 0.0576,
+      "step": 26875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.617369217158024e-06,
+      "loss": 0.0494,
+      "step": 26880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6159422357944007e-06,
+      "loss": 0.2218,
+      "step": 26885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6145152544307774e-06,
+      "loss": 0.1026,
+      "step": 26890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.613088273067154e-06,
+      "loss": 0.144,
+      "step": 26895
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6116612917035305e-06,
+      "loss": 0.445,
+      "step": 26900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.610234310339907e-06,
+      "loss": 0.1412,
+      "step": 26905
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.608807328976284e-06,
+      "loss": 0.1615,
+      "step": 26910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6073803476126604e-06,
+      "loss": 0.096,
+      "step": 26915
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.605953366249037e-06,
+      "loss": 0.1041,
+      "step": 26920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6045263848854135e-06,
+      "loss": 0.1171,
+      "step": 26925
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6030994035217907e-06,
+      "loss": 0.0616,
+      "step": 26930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.601672422158167e-06,
+      "loss": 0.1357,
+      "step": 26935
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6002454407945434e-06,
+      "loss": 0.121,
+      "step": 26940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.59881845943092e-06,
+      "loss": 0.3521,
+      "step": 26945
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5973914780672965e-06,
+      "loss": 0.2587,
+      "step": 26950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5959644967036736e-06,
+      "loss": 0.0782,
+      "step": 26955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.59453751534005e-06,
+      "loss": 0.1999,
+      "step": 26960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5931105339764263e-06,
+      "loss": 0.1359,
+      "step": 26965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.591683552612803e-06,
+      "loss": 0.1388,
+      "step": 26970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5902565712491794e-06,
+      "loss": 0.1563,
+      "step": 26975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5888295898855566e-06,
+      "loss": 0.092,
+      "step": 26980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.587402608521933e-06,
+      "loss": 0.0767,
+      "step": 26985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5859756271583097e-06,
+      "loss": 0.1331,
+      "step": 26990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.584548645794686e-06,
+      "loss": 0.1713,
+      "step": 26995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5831216644310624e-06,
+      "loss": 0.2481,
+      "step": 27000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5816946830674396e-06,
+      "loss": 0.151,
+      "step": 27005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.580267701703816e-06,
+      "loss": 0.1115,
+      "step": 27010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5788407203401927e-06,
+      "loss": 0.1995,
+      "step": 27015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.577413738976569e-06,
+      "loss": 0.0765,
+      "step": 27020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.575986757612946e-06,
+      "loss": 0.0888,
+      "step": 27025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5745597762493225e-06,
+      "loss": 0.0971,
+      "step": 27030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.573132794885699e-06,
+      "loss": 0.103,
+      "step": 27035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5717058135220756e-06,
+      "loss": 0.1728,
+      "step": 27040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.570278832158452e-06,
+      "loss": 0.1494,
+      "step": 27045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.568851850794829e-06,
+      "loss": 0.2001,
+      "step": 27050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5674248694312055e-06,
+      "loss": 0.138,
+      "step": 27055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5659978880675823e-06,
+      "loss": 0.1154,
+      "step": 27060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5645709067039586e-06,
+      "loss": 0.1979,
+      "step": 27065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.563143925340335e-06,
+      "loss": 0.1417,
+      "step": 27070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.561716943976712e-06,
+      "loss": 0.061,
+      "step": 27075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5602899626130885e-06,
+      "loss": 0.061,
+      "step": 27080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5588629812494652e-06,
+      "loss": 0.1333,
+      "step": 27085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5574359998858416e-06,
+      "loss": 0.4135,
+      "step": 27090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.556009018522218e-06,
+      "loss": 0.3602,
+      "step": 27095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.554582037158595e-06,
+      "loss": 0.6737,
+      "step": 27100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5531550557949714e-06,
+      "loss": 0.136,
+      "step": 27105
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.551728074431348e-06,
+      "loss": 0.1231,
+      "step": 27110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5503010930677245e-06,
+      "loss": 0.1484,
+      "step": 27115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5488741117041017e-06,
+      "loss": 0.1621,
+      "step": 27120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.547447130340478e-06,
+      "loss": 0.1258,
+      "step": 27125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5460201489768544e-06,
+      "loss": 0.0405,
+      "step": 27130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.544593167613231e-06,
+      "loss": 0.1641,
+      "step": 27135
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5431661862496075e-06,
+      "loss": 0.1708,
+      "step": 27140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5417392048859847e-06,
+      "loss": 0.1728,
+      "step": 27145
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.540312223522361e-06,
+      "loss": 0.5788,
+      "step": 27150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5388852421587378e-06,
+      "loss": 0.1277,
+      "step": 27155
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.537458260795114e-06,
+      "loss": 0.1229,
+      "step": 27160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5360312794314905e-06,
+      "loss": 0.1463,
+      "step": 27165
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5346042980678676e-06,
+      "loss": 0.2008,
+      "step": 27170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.533177316704244e-06,
+      "loss": 0.115,
+      "step": 27175
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5317503353406207e-06,
+      "loss": 0.1009,
+      "step": 27180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.530323353976997e-06,
+      "loss": 0.1141,
+      "step": 27185
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5288963726133734e-06,
+      "loss": 0.1155,
+      "step": 27190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5274693912497506e-06,
+      "loss": 0.3133,
+      "step": 27195
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.526042409886127e-06,
+      "loss": 0.4243,
+      "step": 27200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5246154285225037e-06,
+      "loss": 0.1038,
+      "step": 27205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.52318844715888e-06,
+      "loss": 0.1554,
+      "step": 27210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5217614657952572e-06,
+      "loss": 0.1083,
+      "step": 27215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5203344844316336e-06,
+      "loss": 0.1736,
+      "step": 27220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.51890750306801e-06,
+      "loss": 0.0624,
+      "step": 27225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5174805217043867e-06,
+      "loss": 0.0619,
+      "step": 27230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.516053540340763e-06,
+      "loss": 0.1438,
+      "step": 27235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.51462655897714e-06,
+      "loss": 0.1014,
+      "step": 27240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5131995776135165e-06,
+      "loss": 0.2204,
+      "step": 27245
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5117725962498933e-06,
+      "loss": 0.3898,
+      "step": 27250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5103456148862696e-06,
+      "loss": 0.1691,
+      "step": 27255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5089186335226464e-06,
+      "loss": 0.1972,
+      "step": 27260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.507491652159023e-06,
+      "loss": 0.1593,
+      "step": 27265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5060646707953995e-06,
+      "loss": 0.1196,
+      "step": 27270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5046376894317763e-06,
+      "loss": 0.1242,
+      "step": 27275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5032107080681526e-06,
+      "loss": 0.0395,
+      "step": 27280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5017837267045294e-06,
+      "loss": 0.2878,
+      "step": 27285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.500356745340906e-06,
+      "loss": 0.1655,
+      "step": 27290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.498929763977283e-06,
+      "loss": 0.3432,
+      "step": 27295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4975027826136592e-06,
+      "loss": 0.2969,
+      "step": 27300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.496075801250036e-06,
+      "loss": 0.158,
+      "step": 27305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4946488198864123e-06,
+      "loss": 0.1344,
+      "step": 27310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.493221838522789e-06,
+      "loss": 0.1479,
+      "step": 27315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.491794857159166e-06,
+      "loss": 0.1574,
+      "step": 27320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4903678757955426e-06,
+      "loss": 0.1102,
+      "step": 27325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.488940894431919e-06,
+      "loss": 0.0472,
+      "step": 27330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4875139130682953e-06,
+      "loss": 0.1569,
+      "step": 27335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.486086931704672e-06,
+      "loss": 0.1781,
+      "step": 27340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.484659950341049e-06,
+      "loss": 0.2265,
+      "step": 27345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4832329689774256e-06,
+      "loss": 0.2439,
+      "step": 27350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.481805987613802e-06,
+      "loss": 0.068,
+      "step": 27355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4803790062501787e-06,
+      "loss": 0.2212,
+      "step": 27360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.478952024886555e-06,
+      "loss": 0.1199,
+      "step": 27365
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4775250435229318e-06,
+      "loss": 0.1074,
+      "step": 27370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4760980621593085e-06,
+      "loss": 0.1394,
+      "step": 27375
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.474671080795685e-06,
+      "loss": 0.0687,
+      "step": 27380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4732440994320616e-06,
+      "loss": 0.0813,
+      "step": 27385
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4718171180684384e-06,
+      "loss": 0.1012,
+      "step": 27390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4703901367048147e-06,
+      "loss": 0.1528,
+      "step": 27395
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4689631553411915e-06,
+      "loss": 0.475,
+      "step": 27400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.467536173977568e-06,
+      "loss": 0.1211,
+      "step": 27405
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4661091926139446e-06,
+      "loss": 0.1699,
+      "step": 27410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4646822112503214e-06,
+      "loss": 0.1281,
+      "step": 27415
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.463255229886698e-06,
+      "loss": 0.1253,
+      "step": 27420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4618282485230745e-06,
+      "loss": 0.2102,
+      "step": 27425
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.460401267159451e-06,
+      "loss": 0.0762,
+      "step": 27430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4589742857958276e-06,
+      "loss": 0.1024,
+      "step": 27435
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4575473044322043e-06,
+      "loss": 0.1654,
+      "step": 27440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.456120323068581e-06,
+      "loss": 0.2602,
+      "step": 27445
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4546933417049574e-06,
+      "loss": 0.525,
+      "step": 27450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.453266360341334e-06,
+      "loss": 0.1613,
+      "step": 27455
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4518393789777105e-06,
+      "loss": 0.1533,
+      "step": 27460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4504123976140873e-06,
+      "loss": 0.1351,
+      "step": 27465
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.448985416250464e-06,
+      "loss": 0.1482,
+      "step": 27470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4475584348868404e-06,
+      "loss": 0.1441,
+      "step": 27475
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.446131453523217e-06,
+      "loss": 0.0479,
+      "step": 27480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.444704472159594e-06,
+      "loss": 0.0756,
+      "step": 27485
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4432774907959707e-06,
+      "loss": 0.1747,
+      "step": 27490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.441850509432347e-06,
+      "loss": 0.186,
+      "step": 27495
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4404235280687238e-06,
+      "loss": 0.2346,
+      "step": 27500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4389965467051e-06,
+      "loss": 0.1429,
+      "step": 27505
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.437569565341477e-06,
+      "loss": 0.1096,
+      "step": 27510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4361425839778536e-06,
+      "loss": 0.1442,
+      "step": 27515
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.43471560261423e-06,
+      "loss": 0.106,
+      "step": 27520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4332886212506067e-06,
+      "loss": 0.1568,
+      "step": 27525
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.431861639886983e-06,
+      "loss": 0.096,
+      "step": 27530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.43043465852336e-06,
+      "loss": 0.0707,
+      "step": 27535
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4290076771597366e-06,
+      "loss": 0.2006,
+      "step": 27540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4275806957961134e-06,
+      "loss": 0.2437,
+      "step": 27545
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4261537144324897e-06,
+      "loss": 0.2953,
+      "step": 27550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.424726733068866e-06,
+      "loss": 0.1089,
+      "step": 27555
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.423299751705243e-06,
+      "loss": 0.1225,
+      "step": 27560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4218727703416196e-06,
+      "loss": 0.1321,
+      "step": 27565
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4204457889779963e-06,
+      "loss": 0.1517,
+      "step": 27570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4190188076143727e-06,
+      "loss": 0.1099,
+      "step": 27575
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4175918262507494e-06,
+      "loss": 0.1013,
+      "step": 27580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.416164844887126e-06,
+      "loss": 0.0818,
+      "step": 27585
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4147378635235025e-06,
+      "loss": 0.1875,
+      "step": 27590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4133108821598793e-06,
+      "loss": 0.1188,
+      "step": 27595
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4118839007962556e-06,
+      "loss": 0.3558,
+      "step": 27600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4104569194326324e-06,
+      "loss": 0.1315,
+      "step": 27605
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.409029938069009e-06,
+      "loss": 0.1606,
+      "step": 27610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.407602956705386e-06,
+      "loss": 0.1286,
+      "step": 27615
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4061759753417623e-06,
+      "loss": 0.12,
+      "step": 27620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4047489939781386e-06,
+      "loss": 0.027,
+      "step": 27625
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4033220126145154e-06,
+      "loss": 0.1582,
+      "step": 27630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.401895031250892e-06,
+      "loss": 0.0792,
+      "step": 27635
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.400468049887269e-06,
+      "loss": 0.1332,
+      "step": 27640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3990410685236452e-06,
+      "loss": 0.1546,
+      "step": 27645
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3976140871600216e-06,
+      "loss": 0.5325,
+      "step": 27650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3961871057963983e-06,
+      "loss": 0.1191,
+      "step": 27655
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.394760124432775e-06,
+      "loss": 0.1728,
+      "step": 27660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.393333143069152e-06,
+      "loss": 0.1316,
+      "step": 27665
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.391906161705528e-06,
+      "loss": 0.1639,
+      "step": 27670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.390479180341905e-06,
+      "loss": 0.0454,
+      "step": 27675
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3890521989782817e-06,
+      "loss": 0.1234,
+      "step": 27680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.387625217614658e-06,
+      "loss": 0.1786,
+      "step": 27685
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.386198236251035e-06,
+      "loss": 0.1624,
+      "step": 27690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.384771254887411e-06,
+      "loss": 0.193,
+      "step": 27695
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.383344273523788e-06,
+      "loss": 0.3262,
+      "step": 27700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3819172921601647e-06,
+      "loss": 0.1204,
+      "step": 27705
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3804903107965414e-06,
+      "loss": 0.063,
+      "step": 27710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3790633294329178e-06,
+      "loss": 0.144,
+      "step": 27715
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3776363480692945e-06,
+      "loss": 0.1995,
+      "step": 27720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.376209366705671e-06,
+      "loss": 0.211,
+      "step": 27725
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3747823853420476e-06,
+      "loss": 0.09,
+      "step": 27730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3733554039784244e-06,
+      "loss": 0.0303,
+      "step": 27735
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3719284226148007e-06,
+      "loss": 0.1759,
+      "step": 27740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3705014412511775e-06,
+      "loss": 0.1973,
+      "step": 27745
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.369074459887554e-06,
+      "loss": 0.422,
+      "step": 27750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3676474785239306e-06,
+      "loss": 0.1344,
+      "step": 27755
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3662204971603074e-06,
+      "loss": 0.1056,
+      "step": 27760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.364793515796684e-06,
+      "loss": 0.1137,
+      "step": 27765
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3633665344330605e-06,
+      "loss": 0.0922,
+      "step": 27770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3619395530694372e-06,
+      "loss": 0.1029,
+      "step": 27775
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3605125717058136e-06,
+      "loss": 0.1035,
+      "step": 27780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3590855903421903e-06,
+      "loss": 0.1228,
+      "step": 27785
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.357658608978567e-06,
+      "loss": 0.1267,
+      "step": 27790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3562316276149434e-06,
+      "loss": 0.3028,
+      "step": 27795
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.35480464625132e-06,
+      "loss": 0.3352,
+      "step": 27800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.353377664887697e-06,
+      "loss": 0.113,
+      "step": 27805
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3519506835240733e-06,
+      "loss": 0.1714,
+      "step": 27810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.35052370216045e-06,
+      "loss": 0.1211,
+      "step": 27815
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3490967207968264e-06,
+      "loss": 0.1798,
+      "step": 27820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.347669739433203e-06,
+      "loss": 0.1656,
+      "step": 27825
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.34624275806958e-06,
+      "loss": 0.101,
+      "step": 27830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3448157767059567e-06,
+      "loss": 0.0754,
+      "step": 27835
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.343388795342333e-06,
+      "loss": 0.1488,
+      "step": 27840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3419618139787094e-06,
+      "loss": 0.2053,
+      "step": 27845
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.340534832615086e-06,
+      "loss": 0.3365,
+      "step": 27850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.339107851251463e-06,
+      "loss": 0.1392,
+      "step": 27855
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3376808698878396e-06,
+      "loss": 0.1091,
+      "step": 27860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.336253888524216e-06,
+      "loss": 0.1421,
+      "step": 27865
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3348269071605927e-06,
+      "loss": 0.1185,
+      "step": 27870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.333399925796969e-06,
+      "loss": 0.0654,
+      "step": 27875
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.331972944433346e-06,
+      "loss": 0.1035,
+      "step": 27880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3305459630697226e-06,
+      "loss": 0.1693,
+      "step": 27885
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.329118981706099e-06,
+      "loss": 0.0753,
+      "step": 27890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3276920003424757e-06,
+      "loss": 0.1208,
+      "step": 27895
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3262650189788525e-06,
+      "loss": 0.4328,
+      "step": 27900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.324838037615229e-06,
+      "loss": 0.0933,
+      "step": 27905
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3234110562516056e-06,
+      "loss": 0.1772,
+      "step": 27910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.321984074887982e-06,
+      "loss": 0.1042,
+      "step": 27915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3205570935243587e-06,
+      "loss": 0.1693,
+      "step": 27920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3191301121607354e-06,
+      "loss": 0.1167,
+      "step": 27925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.317703130797112e-06,
+      "loss": 0.0532,
+      "step": 27930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3162761494334885e-06,
+      "loss": 0.0951,
+      "step": 27935
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3148491680698653e-06,
+      "loss": 0.086,
+      "step": 27940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3134221867062416e-06,
+      "loss": 0.2343,
+      "step": 27945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3119952053426184e-06,
+      "loss": 0.5699,
+      "step": 27950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.310568223978995e-06,
+      "loss": 0.2167,
+      "step": 27955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3091412426153715e-06,
+      "loss": 0.1388,
+      "step": 27960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3077142612517483e-06,
+      "loss": 0.1511,
+      "step": 27965
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3062872798881246e-06,
+      "loss": 0.1833,
+      "step": 27970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3048602985245014e-06,
+      "loss": 0.1014,
+      "step": 27975
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.303433317160878e-06,
+      "loss": 0.0743,
+      "step": 27980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.302006335797255e-06,
+      "loss": 0.097,
+      "step": 27985
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3005793544336312e-06,
+      "loss": 0.1605,
+      "step": 27990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.299152373070008e-06,
+      "loss": 0.2227,
+      "step": 27995
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2977253917063843e-06,
+      "loss": 0.1816,
+      "step": 28000
+    },
+    {
+      "epoch": 0.78,
+      "eval_loss": 0.05539175495505333,
+      "eval_runtime": 1838.9691,
+      "eval_samples_per_second": 8.71,
+      "eval_steps_per_second": 2.178,
+      "eval_wer": 0.16172015867291742,
+      "step": 28000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.296298410342761e-06,
+      "loss": 0.1441,
+      "step": 28005
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.294871428979138e-06,
+      "loss": 0.1065,
+      "step": 28010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.293444447615514e-06,
+      "loss": 0.1141,
+      "step": 28015
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.292017466251891e-06,
+      "loss": 0.3045,
+      "step": 28020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2905904848882677e-06,
+      "loss": 0.1289,
+      "step": 28025
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2891635035246445e-06,
+      "loss": 0.1346,
+      "step": 28030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.287736522161021e-06,
+      "loss": 0.0897,
+      "step": 28035
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.286309540797397e-06,
+      "loss": 0.1108,
+      "step": 28040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.284882559433774e-06,
+      "loss": 0.1147,
+      "step": 28045
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2834555780701507e-06,
+      "loss": 0.3611,
+      "step": 28050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2820285967065274e-06,
+      "loss": 0.1684,
+      "step": 28055
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2806016153429038e-06,
+      "loss": 0.1289,
+      "step": 28060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.27917463397928e-06,
+      "loss": 0.2035,
+      "step": 28065
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.277747652615657e-06,
+      "loss": 0.1857,
+      "step": 28070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2763206712520336e-06,
+      "loss": 0.0604,
+      "step": 28075
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2748936898884104e-06,
+      "loss": 0.0587,
+      "step": 28080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2734667085247867e-06,
+      "loss": 0.1677,
+      "step": 28085
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2720397271611635e-06,
+      "loss": 0.1768,
+      "step": 28090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2706127457975403e-06,
+      "loss": 0.2347,
+      "step": 28095
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2691857644339166e-06,
+      "loss": 0.5117,
+      "step": 28100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2677587830702934e-06,
+      "loss": 0.1398,
+      "step": 28105
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2663318017066697e-06,
+      "loss": 0.0991,
+      "step": 28110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2649048203430465e-06,
+      "loss": 0.1617,
+      "step": 28115
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2634778389794232e-06,
+      "loss": 0.1641,
+      "step": 28120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2620508576158e-06,
+      "loss": 0.1207,
+      "step": 28125
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2606238762521763e-06,
+      "loss": 0.0788,
+      "step": 28130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2591968948885527e-06,
+      "loss": 0.2249,
+      "step": 28135
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2577699135249294e-06,
+      "loss": 0.2089,
+      "step": 28140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.256342932161306e-06,
+      "loss": 0.1748,
+      "step": 28145
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.254915950797683e-06,
+      "loss": 0.3743,
+      "step": 28150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2534889694340593e-06,
+      "loss": 0.1678,
+      "step": 28155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.252061988070436e-06,
+      "loss": 0.1129,
+      "step": 28160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2506350067068124e-06,
+      "loss": 0.1907,
+      "step": 28165
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.249208025343189e-06,
+      "loss": 0.1486,
+      "step": 28170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.247781043979566e-06,
+      "loss": 0.0594,
+      "step": 28175
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2463540626159423e-06,
+      "loss": 0.0934,
+      "step": 28180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.244927081252319e-06,
+      "loss": 0.0663,
+      "step": 28185
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2435000998886958e-06,
+      "loss": 0.0894,
+      "step": 28190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.242073118525072e-06,
+      "loss": 0.3374,
+      "step": 28195
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.240646137161449e-06,
+      "loss": 0.3275,
+      "step": 28200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2392191557978256e-06,
+      "loss": 0.077,
+      "step": 28205
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.237792174434202e-06,
+      "loss": 0.1487,
+      "step": 28210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2363651930705787e-06,
+      "loss": 0.1179,
+      "step": 28215
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2349382117069555e-06,
+      "loss": 0.1686,
+      "step": 28220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.233511230343332e-06,
+      "loss": 0.2046,
+      "step": 28225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2320842489797086e-06,
+      "loss": 0.1222,
+      "step": 28230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.230657267616085e-06,
+      "loss": 0.0498,
+      "step": 28235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2292302862524617e-06,
+      "loss": 0.1486,
+      "step": 28240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2278033048888385e-06,
+      "loss": 0.2174,
+      "step": 28245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2263763235252152e-06,
+      "loss": 0.5739,
+      "step": 28250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2249493421615916e-06,
+      "loss": 0.1467,
+      "step": 28255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.223522360797968e-06,
+      "loss": 0.168,
+      "step": 28260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2220953794343447e-06,
+      "loss": 0.1441,
+      "step": 28265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2206683980707214e-06,
+      "loss": 0.2266,
+      "step": 28270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.219241416707098e-06,
+      "loss": 0.2335,
+      "step": 28275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2178144353434745e-06,
+      "loss": 0.1233,
+      "step": 28280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2163874539798513e-06,
+      "loss": 0.0237,
+      "step": 28285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2149604726162276e-06,
+      "loss": 0.1866,
+      "step": 28290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2135334912526044e-06,
+      "loss": 0.2368,
+      "step": 28295
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.212106509888981e-06,
+      "loss": 0.466,
+      "step": 28300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2106795285253575e-06,
+      "loss": 0.2206,
+      "step": 28305
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2092525471617343e-06,
+      "loss": 0.1756,
+      "step": 28310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.207825565798111e-06,
+      "loss": 0.1234,
+      "step": 28315
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2063985844344874e-06,
+      "loss": 0.1879,
+      "step": 28320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.204971603070864e-06,
+      "loss": 0.0889,
+      "step": 28325
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2035446217072405e-06,
+      "loss": 0.0262,
+      "step": 28330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2021176403436172e-06,
+      "loss": 0.0909,
+      "step": 28335
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200690658979994e-06,
+      "loss": 0.2125,
+      "step": 28340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1992636776163707e-06,
+      "loss": 0.1656,
+      "step": 28345
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.197836696252747e-06,
+      "loss": 0.3062,
+      "step": 28350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1964097148891234e-06,
+      "loss": 0.1637,
+      "step": 28355
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1949827335255e-06,
+      "loss": 0.1259,
+      "step": 28360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.193555752161877e-06,
+      "loss": 0.162,
+      "step": 28365
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1921287707982537e-06,
+      "loss": 0.1843,
+      "step": 28370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.19070178943463e-06,
+      "loss": 0.084,
+      "step": 28375
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.189274808071007e-06,
+      "loss": 0.048,
+      "step": 28380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.187847826707383e-06,
+      "loss": 0.2166,
+      "step": 28385
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.18642084534376e-06,
+      "loss": 0.1625,
+      "step": 28390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1849938639801367e-06,
+      "loss": 0.1636,
+      "step": 28395
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1835668826165134e-06,
+      "loss": 0.5252,
+      "step": 28400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1821399012528898e-06,
+      "loss": 0.1424,
+      "step": 28405
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1807129198892665e-06,
+      "loss": 0.1969,
+      "step": 28410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.179285938525643e-06,
+      "loss": 0.115,
+      "step": 28415
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1778589571620196e-06,
+      "loss": 0.1032,
+      "step": 28420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1764319757983964e-06,
+      "loss": 0.0824,
+      "step": 28425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1750049944347727e-06,
+      "loss": 0.0057,
+      "step": 28430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1735780130711495e-06,
+      "loss": 0.152,
+      "step": 28435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1721510317075263e-06,
+      "loss": 0.1678,
+      "step": 28440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.170724050343903e-06,
+      "loss": 0.3145,
+      "step": 28445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1692970689802794e-06,
+      "loss": 0.3238,
+      "step": 28450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1678700876166557e-06,
+      "loss": 0.1323,
+      "step": 28455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1664431062530325e-06,
+      "loss": 0.1021,
+      "step": 28460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1650161248894092e-06,
+      "loss": 0.1061,
+      "step": 28465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.163589143525786e-06,
+      "loss": 0.2099,
+      "step": 28470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1621621621621623e-06,
+      "loss": 0.2323,
+      "step": 28475
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1607351807985387e-06,
+      "loss": 0.0497,
+      "step": 28480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1593081994349154e-06,
+      "loss": 0.1263,
+      "step": 28485
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.157881218071292e-06,
+      "loss": 0.1116,
+      "step": 28490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156454236707669e-06,
+      "loss": 0.1851,
+      "step": 28495
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1550272553440453e-06,
+      "loss": 0.337,
+      "step": 28500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.153600273980422e-06,
+      "loss": 0.13,
+      "step": 28505
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1521732926167984e-06,
+      "loss": 0.0929,
+      "step": 28510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150746311253175e-06,
+      "loss": 0.1028,
+      "step": 28515
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.149319329889552e-06,
+      "loss": 0.1104,
+      "step": 28520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1478923485259283e-06,
+      "loss": 0.1081,
+      "step": 28525
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.146465367162305e-06,
+      "loss": 0.1647,
+      "step": 28530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1450383857986818e-06,
+      "loss": 0.0791,
+      "step": 28535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1436114044350585e-06,
+      "loss": 0.2013,
+      "step": 28540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.142184423071435e-06,
+      "loss": 0.0707,
+      "step": 28545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1407574417078112e-06,
+      "loss": 0.4112,
+      "step": 28550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.139330460344188e-06,
+      "loss": 0.1367,
+      "step": 28555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1379034789805647e-06,
+      "loss": 0.1587,
+      "step": 28560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1364764976169415e-06,
+      "loss": 0.1337,
+      "step": 28565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.135049516253318e-06,
+      "loss": 0.1745,
+      "step": 28570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1336225348896946e-06,
+      "loss": 0.0903,
+      "step": 28575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.132195553526071e-06,
+      "loss": 0.0845,
+      "step": 28580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1307685721624477e-06,
+      "loss": 0.0549,
+      "step": 28585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1293415907988245e-06,
+      "loss": 0.1265,
+      "step": 28590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127914609435201e-06,
+      "loss": 0.2066,
+      "step": 28595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1264876280715776e-06,
+      "loss": 0.4262,
+      "step": 28600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.125060646707954e-06,
+      "loss": 0.0848,
+      "step": 28605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1236336653443307e-06,
+      "loss": 0.1373,
+      "step": 28610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1222066839807074e-06,
+      "loss": 0.1562,
+      "step": 28615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.120779702617084e-06,
+      "loss": 0.1411,
+      "step": 28620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1193527212534605e-06,
+      "loss": 0.1259,
+      "step": 28625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1179257398898373e-06,
+      "loss": 0.0769,
+      "step": 28630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.116498758526214e-06,
+      "loss": 0.1427,
+      "step": 28635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1150717771625904e-06,
+      "loss": 0.127,
+      "step": 28640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.113644795798967e-06,
+      "loss": 0.3177,
+      "step": 28645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1122178144353435e-06,
+      "loss": 0.2704,
+      "step": 28650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1107908330717203e-06,
+      "loss": 0.1654,
+      "step": 28655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.109363851708097e-06,
+      "loss": 0.1295,
+      "step": 28660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1079368703444738e-06,
+      "loss": 0.1461,
+      "step": 28665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.10650988898085e-06,
+      "loss": 0.1282,
+      "step": 28670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1050829076172265e-06,
+      "loss": 0.0971,
+      "step": 28675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1036559262536032e-06,
+      "loss": 0.0834,
+      "step": 28680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.10222894488998e-06,
+      "loss": 0.1027,
+      "step": 28685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1008019635263567e-06,
+      "loss": 0.1586,
+      "step": 28690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.099374982162733e-06,
+      "loss": 0.3027,
+      "step": 28695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.09794800079911e-06,
+      "loss": 0.2732,
+      "step": 28700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.096521019435486e-06,
+      "loss": 0.0897,
+      "step": 28705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095094038071863e-06,
+      "loss": 0.1643,
+      "step": 28710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0936670567082397e-06,
+      "loss": 0.1752,
+      "step": 28715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.092240075344616e-06,
+      "loss": 0.1947,
+      "step": 28720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.090813093980993e-06,
+      "loss": 0.206,
+      "step": 28725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0893861126173696e-06,
+      "loss": 0.0614,
+      "step": 28730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.087959131253746e-06,
+      "loss": 0.1007,
+      "step": 28735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0865321498901227e-06,
+      "loss": 0.2373,
+      "step": 28740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.085390564799224e-06,
+      "loss": 0.8765,
+      "step": 28745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0839635834356007e-06,
+      "loss": 0.4749,
+      "step": 28750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.082536602071977e-06,
+      "loss": 0.1613,
+      "step": 28755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0811096207083538e-06,
+      "loss": 0.1467,
+      "step": 28760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.07968263934473e-06,
+      "loss": 0.1365,
+      "step": 28765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.078255657981107e-06,
+      "loss": 0.1313,
+      "step": 28770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0768286766174836e-06,
+      "loss": 0.1381,
+      "step": 28775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0754016952538604e-06,
+      "loss": 0.1045,
+      "step": 28780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0739747138902367e-06,
+      "loss": 0.1228,
+      "step": 28785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072547732526613e-06,
+      "loss": 0.1446,
+      "step": 28790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.07112075116299e-06,
+      "loss": 0.186,
+      "step": 28795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0696937697993666e-06,
+      "loss": 0.3615,
+      "step": 28800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0682667884357434e-06,
+      "loss": 0.1192,
+      "step": 28805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0668398070721197e-06,
+      "loss": 0.1187,
+      "step": 28810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0654128257084965e-06,
+      "loss": 0.15,
+      "step": 28815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.063985844344873e-06,
+      "loss": 0.123,
+      "step": 28820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0625588629812496e-06,
+      "loss": 0.1635,
+      "step": 28825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0611318816176263e-06,
+      "loss": 0.1189,
+      "step": 28830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.059704900254003e-06,
+      "loss": 0.0586,
+      "step": 28835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0582779188903794e-06,
+      "loss": 0.0981,
+      "step": 28840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.056850937526756e-06,
+      "loss": 0.0877,
+      "step": 28845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0554239561631325e-06,
+      "loss": 0.4166,
+      "step": 28850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0539969747995093e-06,
+      "loss": 0.1074,
+      "step": 28855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.052569993435886e-06,
+      "loss": 0.1494,
+      "step": 28860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0511430120722624e-06,
+      "loss": 0.1262,
+      "step": 28865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.049716030708639e-06,
+      "loss": 0.1526,
+      "step": 28870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.048289049345016e-06,
+      "loss": 0.1377,
+      "step": 28875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0468620679813923e-06,
+      "loss": 0.0577,
+      "step": 28880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.045435086617769e-06,
+      "loss": 0.0924,
+      "step": 28885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0440081052541454e-06,
+      "loss": 0.1076,
+      "step": 28890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.042581123890522e-06,
+      "loss": 0.1331,
+      "step": 28895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.041154142526899e-06,
+      "loss": 0.2208,
+      "step": 28900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0397271611632756e-06,
+      "loss": 0.0929,
+      "step": 28905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.038300179799652e-06,
+      "loss": 0.1111,
+      "step": 28910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0368731984360283e-06,
+      "loss": 0.1567,
+      "step": 28915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.035446217072405e-06,
+      "loss": 0.1782,
+      "step": 28920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.034019235708782e-06,
+      "loss": 0.1117,
+      "step": 28925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0325922543451586e-06,
+      "loss": 0.0977,
+      "step": 28930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.031165272981535e-06,
+      "loss": 0.105,
+      "step": 28935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0297382916179117e-06,
+      "loss": 0.328,
+      "step": 28940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.028311310254288e-06,
+      "loss": 0.1402,
+      "step": 28945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.026884328890665e-06,
+      "loss": 0.2992,
+      "step": 28950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0254573475270416e-06,
+      "loss": 0.1506,
+      "step": 28955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.024030366163418e-06,
+      "loss": 0.1239,
+      "step": 28960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0226033847997947e-06,
+      "loss": 0.1656,
+      "step": 28965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0211764034361714e-06,
+      "loss": 0.11,
+      "step": 28970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0197494220725478e-06,
+      "loss": 0.1406,
+      "step": 28975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0183224407089245e-06,
+      "loss": 0.1073,
+      "step": 28980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016895459345301e-06,
+      "loss": 0.1071,
+      "step": 28985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0154684779816776e-06,
+      "loss": 0.2867,
+      "step": 28990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0140414966180544e-06,
+      "loss": 0.2095,
+      "step": 28995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.012614515254431e-06,
+      "loss": 0.412,
+      "step": 29000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0111875338908075e-06,
+      "loss": 0.0761,
+      "step": 29005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.009760552527184e-06,
+      "loss": 0.167,
+      "step": 29010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0083335711635606e-06,
+      "loss": 0.1434,
+      "step": 29015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0069065897999374e-06,
+      "loss": 0.1703,
+      "step": 29020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.005479608436314e-06,
+      "loss": 0.1315,
+      "step": 29025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0040526270726905e-06,
+      "loss": 0.0482,
+      "step": 29030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0026256457090672e-06,
+      "loss": 0.0902,
+      "step": 29035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0011986643454436e-06,
+      "loss": 0.13,
+      "step": 29040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9997716829818203e-06,
+      "loss": 0.1629,
+      "step": 29045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.998344701618197e-06,
+      "loss": 0.306,
+      "step": 29050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.996917720254574e-06,
+      "loss": 0.1238,
+      "step": 29055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.99549073889095e-06,
+      "loss": 0.1279,
+      "step": 29060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.994063757527327e-06,
+      "loss": 0.1375,
+      "step": 29065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9926367761637037e-06,
+      "loss": 0.1839,
+      "step": 29070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.99120979480008e-06,
+      "loss": 0.1504,
+      "step": 29075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.989782813436457e-06,
+      "loss": 0.0427,
+      "step": 29080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.988355832072833e-06,
+      "loss": 0.1892,
+      "step": 29085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.98692885070921e-06,
+      "loss": 0.1265,
+      "step": 29090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9855018693455867e-06,
+      "loss": 0.1475,
+      "step": 29095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9840748879819634e-06,
+      "loss": 0.1996,
+      "step": 29100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9826479066183398e-06,
+      "loss": 0.1312,
+      "step": 29105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.981220925254716e-06,
+      "loss": 0.1338,
+      "step": 29110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.979793943891093e-06,
+      "loss": 0.1202,
+      "step": 29115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9783669625274696e-06,
+      "loss": 0.1662,
+      "step": 29120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9769399811638464e-06,
+      "loss": 0.176,
+      "step": 29125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9755129998002227e-06,
+      "loss": 0.0265,
+      "step": 29130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.974086018436599e-06,
+      "loss": 0.0414,
+      "step": 29135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.972659037072976e-06,
+      "loss": 0.1461,
+      "step": 29140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9712320557093526e-06,
+      "loss": 0.2434,
+      "step": 29145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9698050743457294e-06,
+      "loss": 0.2549,
+      "step": 29150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9683780929821057e-06,
+      "loss": 0.1046,
+      "step": 29155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9669511116184825e-06,
+      "loss": 0.1638,
+      "step": 29160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9655241302548592e-06,
+      "loss": 0.2054,
+      "step": 29165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9640971488912356e-06,
+      "loss": 0.1687,
+      "step": 29170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9626701675276123e-06,
+      "loss": 0.182,
+      "step": 29175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9612431861639887e-06,
+      "loss": 0.0696,
+      "step": 29180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9598162048003654e-06,
+      "loss": 0.0819,
+      "step": 29185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.958389223436742e-06,
+      "loss": 0.0868,
+      "step": 29190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.956962242073119e-06,
+      "loss": 0.1614,
+      "step": 29195
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9555352607094953e-06,
+      "loss": 0.2675,
+      "step": 29200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9541082793458716e-06,
+      "loss": 0.1194,
+      "step": 29205
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9526812979822484e-06,
+      "loss": 0.1698,
+      "step": 29210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.951254316618625e-06,
+      "loss": 0.1408,
+      "step": 29215
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.949827335255002e-06,
+      "loss": 0.1261,
+      "step": 29220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9484003538913783e-06,
+      "loss": 0.0789,
+      "step": 29225
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.946973372527755e-06,
+      "loss": 0.1169,
+      "step": 29230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9455463911641314e-06,
+      "loss": 0.0752,
+      "step": 29235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.944119409800508e-06,
+      "loss": 0.1521,
+      "step": 29240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.942692428436885e-06,
+      "loss": 0.1285,
+      "step": 29245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9412654470732612e-06,
+      "loss": 0.3726,
+      "step": 29250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.939838465709638e-06,
+      "loss": 0.1007,
+      "step": 29255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9384114843460147e-06,
+      "loss": 0.1351,
+      "step": 29260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.936984502982391e-06,
+      "loss": 0.1856,
+      "step": 29265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.935557521618768e-06,
+      "loss": 0.1899,
+      "step": 29270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9341305402551446e-06,
+      "loss": 0.1647,
+      "step": 29275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.932703558891521e-06,
+      "loss": 0.1379,
+      "step": 29280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9312765775278977e-06,
+      "loss": 0.1229,
+      "step": 29285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9298495961642745e-06,
+      "loss": 0.1981,
+      "step": 29290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.928422614800651e-06,
+      "loss": 0.1811,
+      "step": 29295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9269956334370276e-06,
+      "loss": 0.3038,
+      "step": 29300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.925568652073404e-06,
+      "loss": 0.0714,
+      "step": 29305
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9241416707097807e-06,
+      "loss": 0.1724,
+      "step": 29310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9227146893461574e-06,
+      "loss": 0.1529,
+      "step": 29315
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.921287707982534e-06,
+      "loss": 0.1615,
+      "step": 29320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9198607266189105e-06,
+      "loss": 0.1092,
+      "step": 29325
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.918433745255287e-06,
+      "loss": 0.0957,
+      "step": 29330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9170067638916636e-06,
+      "loss": 0.1288,
+      "step": 29335
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9155797825280404e-06,
+      "loss": 0.1648,
+      "step": 29340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.914152801164417e-06,
+      "loss": 0.1914,
+      "step": 29345
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9127258198007935e-06,
+      "loss": 0.3363,
+      "step": 29350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9112988384371703e-06,
+      "loss": 0.1357,
+      "step": 29355
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9098718570735466e-06,
+      "loss": 0.1336,
+      "step": 29360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9084448757099234e-06,
+      "loss": 0.1413,
+      "step": 29365
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9070178943463e-06,
+      "loss": 0.1587,
+      "step": 29370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.9055909129826767e-06,
+      "loss": 0.2068,
+      "step": 29375
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.9041639316190532e-06,
+      "loss": 0.0638,
+      "step": 29380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.90273695025543e-06,
+      "loss": 0.3071,
+      "step": 29385
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.9013099688918063e-06,
+      "loss": 0.0693,
+      "step": 29390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8998829875281829e-06,
+      "loss": 0.1451,
+      "step": 29395
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8984560061645596e-06,
+      "loss": 0.2641,
+      "step": 29400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8970290248009362e-06,
+      "loss": 0.1411,
+      "step": 29405
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.895602043437313e-06,
+      "loss": 0.1326,
+      "step": 29410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8941750620736895e-06,
+      "loss": 0.1322,
+      "step": 29415
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8927480807100663e-06,
+      "loss": 0.1163,
+      "step": 29420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8913210993464426e-06,
+      "loss": 0.1126,
+      "step": 29425
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8898941179828192e-06,
+      "loss": 0.0469,
+      "step": 29430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.888467136619196e-06,
+      "loss": 0.0579,
+      "step": 29435
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8870401552555725e-06,
+      "loss": 0.1405,
+      "step": 29440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8856131738919492e-06,
+      "loss": 0.3488,
+      "step": 29445
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8841861925283258e-06,
+      "loss": 0.4339,
+      "step": 29450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8827592111647023e-06,
+      "loss": 0.2233,
+      "step": 29455
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8813322298010789e-06,
+      "loss": 0.153,
+      "step": 29460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8799052484374554e-06,
+      "loss": 0.1254,
+      "step": 29465
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8784782670738322e-06,
+      "loss": 0.1401,
+      "step": 29470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8770512857102087e-06,
+      "loss": 0.0785,
+      "step": 29475
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8756243043465855e-06,
+      "loss": 0.1434,
+      "step": 29480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8741973229829618e-06,
+      "loss": 0.1638,
+      "step": 29485
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8727703416193386e-06,
+      "loss": 0.254,
+      "step": 29490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8713433602557152e-06,
+      "loss": 0.1793,
+      "step": 29495
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.869916378892092e-06,
+      "loss": 0.4827,
+      "step": 29500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8684893975284685e-06,
+      "loss": 0.1294,
+      "step": 29505
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8670624161648452e-06,
+      "loss": 0.1245,
+      "step": 29510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8656354348012218e-06,
+      "loss": 0.2984,
+      "step": 29515
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8642084534375981e-06,
+      "loss": 0.1426,
+      "step": 29520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8627814720739749e-06,
+      "loss": 0.0468,
+      "step": 29525
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8613544907103514e-06,
+      "loss": 0.0844,
+      "step": 29530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8599275093467282e-06,
+      "loss": 0.0907,
+      "step": 29535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8585005279831047e-06,
+      "loss": 0.1162,
+      "step": 29540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8570735466194815e-06,
+      "loss": 0.2419,
+      "step": 29545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8556465652558578e-06,
+      "loss": 0.3571,
+      "step": 29550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8542195838922344e-06,
+      "loss": 0.1416,
+      "step": 29555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8527926025286112e-06,
+      "loss": 0.1364,
+      "step": 29560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8513656211649877e-06,
+      "loss": 0.1309,
+      "step": 29565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8499386398013645e-06,
+      "loss": 0.1941,
+      "step": 29570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.848511658437741e-06,
+      "loss": 0.162,
+      "step": 29575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8470846770741174e-06,
+      "loss": 0.0987,
+      "step": 29580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8456576957104941e-06,
+      "loss": 0.1344,
+      "step": 29585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8442307143468707e-06,
+      "loss": 0.1251,
+      "step": 29590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8428037329832474e-06,
+      "loss": 0.1892,
+      "step": 29595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.841376751619624e-06,
+      "loss": 0.5304,
+      "step": 29600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8399497702560007e-06,
+      "loss": 0.144,
+      "step": 29605
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8385227888923773e-06,
+      "loss": 0.1337,
+      "step": 29610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8370958075287536e-06,
+      "loss": 0.1263,
+      "step": 29615
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8356688261651304e-06,
+      "loss": 0.1888,
+      "step": 29620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.834241844801507e-06,
+      "loss": 0.1318,
+      "step": 29625
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8328148634378837e-06,
+      "loss": 0.1316,
+      "step": 29630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8313878820742603e-06,
+      "loss": 0.0733,
+      "step": 29635
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.829960900710637e-06,
+      "loss": 0.1689,
+      "step": 29640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8285339193470134e-06,
+      "loss": 0.1285,
+      "step": 29645
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.82710693798339e-06,
+      "loss": 0.4993,
+      "step": 29650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8256799566197667e-06,
+      "loss": 0.1109,
+      "step": 29655
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8242529752561432e-06,
+      "loss": 0.1775,
+      "step": 29660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.82282599389252e-06,
+      "loss": 0.1561,
+      "step": 29665
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8213990125288965e-06,
+      "loss": 0.0936,
+      "step": 29670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8199720311652733e-06,
+      "loss": 0.0657,
+      "step": 29675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8185450498016496e-06,
+      "loss": 0.0613,
+      "step": 29680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8171180684380262e-06,
+      "loss": 0.088,
+      "step": 29685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.815691087074403e-06,
+      "loss": 0.0859,
+      "step": 29690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8142641057107795e-06,
+      "loss": 0.203,
+      "step": 29695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8128371243471563e-06,
+      "loss": 0.3191,
+      "step": 29700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8114101429835328e-06,
+      "loss": 0.1306,
+      "step": 29705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8099831616199094e-06,
+      "loss": 0.1267,
+      "step": 29710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.808556180256286e-06,
+      "loss": 0.1168,
+      "step": 29715
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8071291988926627e-06,
+      "loss": 0.1634,
+      "step": 29720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8057022175290392e-06,
+      "loss": 0.1051,
+      "step": 29725
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.804275236165416e-06,
+      "loss": 0.0556,
+      "step": 29730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8028482548017925e-06,
+      "loss": 0.0653,
+      "step": 29735
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8014212734381689e-06,
+      "loss": 0.2693,
+      "step": 29740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7999942920745456e-06,
+      "loss": 0.1349,
+      "step": 29745
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7985673107109222e-06,
+      "loss": 0.3507,
+      "step": 29750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.797140329347299e-06,
+      "loss": 0.1368,
+      "step": 29755
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7957133479836755e-06,
+      "loss": 0.1215,
+      "step": 29760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7942863666200523e-06,
+      "loss": 0.1801,
+      "step": 29765
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7928593852564288e-06,
+      "loss": 0.1228,
+      "step": 29770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7914324038928052e-06,
+      "loss": 0.0795,
+      "step": 29775
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.790005422529182e-06,
+      "loss": 0.0496,
+      "step": 29780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7885784411655585e-06,
+      "loss": 0.141,
+      "step": 29785
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7871514598019352e-06,
+      "loss": 0.1736,
+      "step": 29790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7857244784383118e-06,
+      "loss": 0.163,
+      "step": 29795
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7842974970746885e-06,
+      "loss": 0.1673,
+      "step": 29800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7828705157110649e-06,
+      "loss": 0.1442,
+      "step": 29805
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7814435343474414e-06,
+      "loss": 0.1148,
+      "step": 29810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7800165529838182e-06,
+      "loss": 0.1531,
+      "step": 29815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7785895716201947e-06,
+      "loss": 0.1766,
+      "step": 29820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7771625902565715e-06,
+      "loss": 0.1422,
+      "step": 29825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.775735608892948e-06,
+      "loss": 0.0921,
+      "step": 29830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7743086275293244e-06,
+      "loss": 0.0602,
+      "step": 29835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7728816461657012e-06,
+      "loss": 0.1568,
+      "step": 29840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7714546648020777e-06,
+      "loss": 0.2646,
+      "step": 29845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7700276834384545e-06,
+      "loss": 0.37,
+      "step": 29850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.768600702074831e-06,
+      "loss": 0.1707,
+      "step": 29855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7671737207112078e-06,
+      "loss": 0.158,
+      "step": 29860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7657467393475843e-06,
+      "loss": 0.1415,
+      "step": 29865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7643197579839607e-06,
+      "loss": 0.195,
+      "step": 29870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7628927766203374e-06,
+      "loss": 0.1228,
+      "step": 29875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.761465795256714e-06,
+      "loss": 0.0604,
+      "step": 29880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7600388138930907e-06,
+      "loss": 0.0835,
+      "step": 29885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7586118325294673e-06,
+      "loss": 0.1226,
+      "step": 29890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.757184851165844e-06,
+      "loss": 0.202,
+      "step": 29895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7557578698022204e-06,
+      "loss": 0.2161,
+      "step": 29900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7543308884385972e-06,
+      "loss": 0.0818,
+      "step": 29905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7529039070749737e-06,
+      "loss": 0.1051,
+      "step": 29910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7514769257113503e-06,
+      "loss": 0.1467,
+      "step": 29915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.750049944347727e-06,
+      "loss": 0.1323,
+      "step": 29920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7486229629841036e-06,
+      "loss": 0.1149,
+      "step": 29925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7471959816204801e-06,
+      "loss": 0.0566,
+      "step": 29930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7457690002568567e-06,
+      "loss": 0.1111,
+      "step": 29935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7443420188932334e-06,
+      "loss": 0.165,
+      "step": 29940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.74291503752961e-06,
+      "loss": 0.2132,
+      "step": 29945
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7414880561659867e-06,
+      "loss": 0.3464,
+      "step": 29950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7400610748023633e-06,
+      "loss": 0.1466,
+      "step": 29955
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.73863409343874e-06,
+      "loss": 0.1953,
+      "step": 29960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7372071120751164e-06,
+      "loss": 0.174,
+      "step": 29965
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.735780130711493e-06,
+      "loss": 0.1493,
+      "step": 29970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7343531493478697e-06,
+      "loss": 0.062,
+      "step": 29975
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7329261679842463e-06,
+      "loss": 0.0588,
+      "step": 29980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.731499186620623e-06,
+      "loss": 0.084,
+      "step": 29985
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7300722052569996e-06,
+      "loss": 0.1432,
+      "step": 29990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.728645223893376e-06,
+      "loss": 0.14,
+      "step": 29995
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7272182425297527e-06,
+      "loss": 0.6414,
+      "step": 30000
+    },
+    {
+      "epoch": 0.83,
+      "eval_loss": 0.05838534235954285,
+      "eval_runtime": 1928.5758,
+      "eval_samples_per_second": 8.306,
+      "eval_steps_per_second": 2.077,
+      "eval_wer": 0.16357735304724125,
+      "step": 30000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7257912611661292e-06,
+      "loss": 0.1612,
+      "step": 30005
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.724364279802506e-06,
+      "loss": 0.1507,
+      "step": 30010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7229372984388825e-06,
+      "loss": 0.1451,
+      "step": 30015
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7215103170752593e-06,
+      "loss": 0.1457,
+      "step": 30020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7200833357116359e-06,
+      "loss": 0.1106,
+      "step": 30025
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7186563543480122e-06,
+      "loss": 0.1255,
+      "step": 30030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.717229372984389e-06,
+      "loss": 0.061,
+      "step": 30035
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7158023916207655e-06,
+      "loss": 0.1902,
+      "step": 30040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7143754102571423e-06,
+      "loss": 0.1521,
+      "step": 30045
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7129484288935188e-06,
+      "loss": 0.3394,
+      "step": 30050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7115214475298956e-06,
+      "loss": 0.1139,
+      "step": 30055
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.710094466166272e-06,
+      "loss": 0.1206,
+      "step": 30060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7086674848026485e-06,
+      "loss": 0.163,
+      "step": 30065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7072405034390252e-06,
+      "loss": 0.1296,
+      "step": 30070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7058135220754018e-06,
+      "loss": 0.0688,
+      "step": 30075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7043865407117785e-06,
+      "loss": 0.2041,
+      "step": 30080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.702959559348155e-06,
+      "loss": 0.1542,
+      "step": 30085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7015325779845314e-06,
+      "loss": 0.0918,
+      "step": 30090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7001055966209082e-06,
+      "loss": 0.1101,
+      "step": 30095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6986786152572847e-06,
+      "loss": 0.5718,
+      "step": 30100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6972516338936615e-06,
+      "loss": 0.141,
+      "step": 30105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.695824652530038e-06,
+      "loss": 0.1428,
+      "step": 30110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6943976711664148e-06,
+      "loss": 0.1814,
+      "step": 30115
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6929706898027914e-06,
+      "loss": 0.1347,
+      "step": 30120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.691543708439168e-06,
+      "loss": 0.0723,
+      "step": 30125
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6901167270755445e-06,
+      "loss": 0.1598,
+      "step": 30130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6886897457119212e-06,
+      "loss": 0.1219,
+      "step": 30135
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6872627643482978e-06,
+      "loss": 0.2055,
+      "step": 30140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6858357829846743e-06,
+      "loss": 0.2999,
+      "step": 30145
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.684408801621051e-06,
+      "loss": 0.5244,
+      "step": 30150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6829818202574274e-06,
+      "loss": 0.2096,
+      "step": 30155
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6815548388938042e-06,
+      "loss": 0.1008,
+      "step": 30160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6801278575301807e-06,
+      "loss": 0.1061,
+      "step": 30165
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6787008761665575e-06,
+      "loss": 0.14,
+      "step": 30170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.677273894802934e-06,
+      "loss": 0.1069,
+      "step": 30175
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6758469134393108e-06,
+      "loss": 0.0708,
+      "step": 30180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6744199320756872e-06,
+      "loss": 0.0701,
+      "step": 30185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6729929507120637e-06,
+      "loss": 0.1315,
+      "step": 30190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6715659693484405e-06,
+      "loss": 0.2271,
+      "step": 30195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.670138987984817e-06,
+      "loss": 0.5044,
+      "step": 30200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6687120066211938e-06,
+      "loss": 0.1734,
+      "step": 30205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6672850252575703e-06,
+      "loss": 0.0858,
+      "step": 30210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.665858043893947e-06,
+      "loss": 0.1001,
+      "step": 30215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6644310625303234e-06,
+      "loss": 0.1547,
+      "step": 30220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6630040811667e-06,
+      "loss": 0.1222,
+      "step": 30225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6615770998030767e-06,
+      "loss": 0.0436,
+      "step": 30230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6601501184394533e-06,
+      "loss": 0.0757,
+      "step": 30235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.65872313707583e-06,
+      "loss": 0.2041,
+      "step": 30240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6572961557122066e-06,
+      "loss": 0.2425,
+      "step": 30245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.655869174348583e-06,
+      "loss": 0.3683,
+      "step": 30250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6544421929849597e-06,
+      "loss": 0.0863,
+      "step": 30255
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6530152116213363e-06,
+      "loss": 0.1347,
+      "step": 30260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.651588230257713e-06,
+      "loss": 0.1001,
+      "step": 30265
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6501612488940896e-06,
+      "loss": 0.098,
+      "step": 30270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6487342675304663e-06,
+      "loss": 0.0426,
+      "step": 30275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6473072861668429e-06,
+      "loss": 0.0671,
+      "step": 30280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6458803048032192e-06,
+      "loss": 0.2194,
+      "step": 30285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.644453323439596e-06,
+      "loss": 0.0606,
+      "step": 30290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6430263420759725e-06,
+      "loss": 0.1529,
+      "step": 30295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6415993607123493e-06,
+      "loss": 0.3677,
+      "step": 30300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6401723793487258e-06,
+      "loss": 0.0818,
+      "step": 30305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6387453979851026e-06,
+      "loss": 0.0956,
+      "step": 30310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637318416621479e-06,
+      "loss": 0.1368,
+      "step": 30315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6358914352578555e-06,
+      "loss": 0.1383,
+      "step": 30320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6344644538942323e-06,
+      "loss": 0.1168,
+      "step": 30325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6330374725306088e-06,
+      "loss": 0.109,
+      "step": 30330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6316104911669856e-06,
+      "loss": 0.3013,
+      "step": 30335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6301835098033621e-06,
+      "loss": 0.3281,
+      "step": 30340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6287565284397387e-06,
+      "loss": 0.2898,
+      "step": 30345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6273295470761152e-06,
+      "loss": 0.383,
+      "step": 30350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.625902565712492e-06,
+      "loss": 0.1322,
+      "step": 30355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6244755843488685e-06,
+      "loss": 0.1253,
+      "step": 30360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.623048602985245e-06,
+      "loss": 0.1888,
+      "step": 30365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6216216216216219e-06,
+      "loss": 0.1797,
+      "step": 30370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6201946402579984e-06,
+      "loss": 0.0675,
+      "step": 30375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.618767658894375e-06,
+      "loss": 0.2116,
+      "step": 30380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6173406775307515e-06,
+      "loss": 0.1694,
+      "step": 30385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6159136961671283e-06,
+      "loss": 0.2422,
+      "step": 30390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6144867148035048e-06,
+      "loss": 0.1664,
+      "step": 30395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6130597334398816e-06,
+      "loss": 0.2869,
+      "step": 30400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6116327520762581e-06,
+      "loss": 0.1527,
+      "step": 30405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6102057707126345e-06,
+      "loss": 0.1121,
+      "step": 30410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6087787893490112e-06,
+      "loss": 0.1529,
+      "step": 30415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6073518079853878e-06,
+      "loss": 0.1104,
+      "step": 30420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6059248266217645e-06,
+      "loss": 0.0833,
+      "step": 30425
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.604497845258141e-06,
+      "loss": 0.1159,
+      "step": 30430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6030708638945179e-06,
+      "loss": 0.042,
+      "step": 30435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6016438825308942e-06,
+      "loss": 0.2316,
+      "step": 30440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6002169011672707e-06,
+      "loss": 0.2693,
+      "step": 30445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5987899198036475e-06,
+      "loss": 0.439,
+      "step": 30450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.597362938440024e-06,
+      "loss": 0.1224,
+      "step": 30455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5959359570764008e-06,
+      "loss": 0.1079,
+      "step": 30460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5945089757127774e-06,
+      "loss": 0.2648,
+      "step": 30465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5930819943491541e-06,
+      "loss": 0.1247,
+      "step": 30470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5916550129855305e-06,
+      "loss": 0.1804,
+      "step": 30475
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.590228031621907e-06,
+      "loss": 0.0743,
+      "step": 30480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5888010502582838e-06,
+      "loss": 0.0926,
+      "step": 30485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5873740688946603e-06,
+      "loss": 0.117,
+      "step": 30490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.585947087531037e-06,
+      "loss": 0.1502,
+      "step": 30495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5845201061674136e-06,
+      "loss": 0.6149,
+      "step": 30500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.58309312480379e-06,
+      "loss": 0.1727,
+      "step": 30505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5816661434401667e-06,
+      "loss": 0.118,
+      "step": 30510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5802391620765433e-06,
+      "loss": 0.1302,
+      "step": 30515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.57881218071292e-06,
+      "loss": 0.2098,
+      "step": 30520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5773851993492966e-06,
+      "loss": 0.115,
+      "step": 30525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5759582179856734e-06,
+      "loss": 0.0751,
+      "step": 30530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5745312366220497e-06,
+      "loss": 0.1084,
+      "step": 30535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5731042552584263e-06,
+      "loss": 0.1145,
+      "step": 30540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.571677273894803e-06,
+      "loss": 0.1338,
+      "step": 30545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5702502925311796e-06,
+      "loss": 0.4143,
+      "step": 30550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5688233111675563e-06,
+      "loss": 0.08,
+      "step": 30555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5673963298039329e-06,
+      "loss": 0.1183,
+      "step": 30560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5659693484403096e-06,
+      "loss": 0.1822,
+      "step": 30565
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.564542367076686e-06,
+      "loss": 0.1639,
+      "step": 30570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5631153857130627e-06,
+      "loss": 0.0777,
+      "step": 30575
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5616884043494393e-06,
+      "loss": 0.094,
+      "step": 30580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.560261422985816e-06,
+      "loss": 0.0695,
+      "step": 30585
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5588344416221926e-06,
+      "loss": 0.1541,
+      "step": 30590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5574074602585692e-06,
+      "loss": 0.1856,
+      "step": 30595
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5559804788949457e-06,
+      "loss": 0.6199,
+      "step": 30600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5545534975313223e-06,
+      "loss": 0.1189,
+      "step": 30605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.553126516167699e-06,
+      "loss": 0.1934,
+      "step": 30610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5516995348040756e-06,
+      "loss": 0.2187,
+      "step": 30615
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5502725534404523e-06,
+      "loss": 0.122,
+      "step": 30620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5488455720768289e-06,
+      "loss": 0.1632,
+      "step": 30625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5474185907132056e-06,
+      "loss": 0.0978,
+      "step": 30630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.545991609349582e-06,
+      "loss": 0.1313,
+      "step": 30635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5445646279859585e-06,
+      "loss": 0.1484,
+      "step": 30640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5431376466223353e-06,
+      "loss": 0.089,
+      "step": 30645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5417106652587119e-06,
+      "loss": 0.2633,
+      "step": 30650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5402836838950886e-06,
+      "loss": 0.1022,
+      "step": 30655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5388567025314652e-06,
+      "loss": 0.1222,
+      "step": 30660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5374297211678415e-06,
+      "loss": 0.1354,
+      "step": 30665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5360027398042183e-06,
+      "loss": 0.1546,
+      "step": 30670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5345757584405948e-06,
+      "loss": 0.0776,
+      "step": 30675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5331487770769716e-06,
+      "loss": 0.1215,
+      "step": 30680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5317217957133481e-06,
+      "loss": 0.0722,
+      "step": 30685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5302948143497249e-06,
+      "loss": 0.2274,
+      "step": 30690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5288678329861012e-06,
+      "loss": 0.4315,
+      "step": 30695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5274408516224778e-06,
+      "loss": 0.4468,
+      "step": 30700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5260138702588545e-06,
+      "loss": 0.1523,
+      "step": 30705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.524586888895231e-06,
+      "loss": 0.172,
+      "step": 30710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5231599075316079e-06,
+      "loss": 0.1838,
+      "step": 30715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5217329261679844e-06,
+      "loss": 0.154,
+      "step": 30720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5203059448043612e-06,
+      "loss": 0.0943,
+      "step": 30725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5188789634407375e-06,
+      "loss": 0.1492,
+      "step": 30730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.517451982077114e-06,
+      "loss": 0.1907,
+      "step": 30735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5160250007134908e-06,
+      "loss": 0.1713,
+      "step": 30740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5145980193498674e-06,
+      "loss": 0.2305,
+      "step": 30745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5131710379862441e-06,
+      "loss": 0.3482,
+      "step": 30750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5117440566226207e-06,
+      "loss": 0.0889,
+      "step": 30755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.510317075258997e-06,
+      "loss": 0.1548,
+      "step": 30760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5088900938953738e-06,
+      "loss": 0.1263,
+      "step": 30765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5074631125317503e-06,
+      "loss": 0.2411,
+      "step": 30770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.506036131168127e-06,
+      "loss": 0.0851,
+      "step": 30775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5046091498045036e-06,
+      "loss": 0.0612,
+      "step": 30780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5031821684408804e-06,
+      "loss": 0.1413,
+      "step": 30785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5017551870772567e-06,
+      "loss": 0.1998,
+      "step": 30790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5003282057136335e-06,
+      "loss": 0.2209,
+      "step": 30795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.49890122435001e-06,
+      "loss": 0.4875,
+      "step": 30800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4974742429863868e-06,
+      "loss": 0.1209,
+      "step": 30805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4960472616227634e-06,
+      "loss": 0.1503,
+      "step": 30810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.49462028025914e-06,
+      "loss": 0.1309,
+      "step": 30815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4931932988955167e-06,
+      "loss": 0.1346,
+      "step": 30820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.491766317531893e-06,
+      "loss": 0.2226,
+      "step": 30825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4903393361682698e-06,
+      "loss": 0.0535,
+      "step": 30830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4889123548046463e-06,
+      "loss": 0.0537,
+      "step": 30835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.487485373441023e-06,
+      "loss": 0.0843,
+      "step": 30840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4860583920773996e-06,
+      "loss": 0.2167,
+      "step": 30845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4846314107137764e-06,
+      "loss": 0.5243,
+      "step": 30850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4832044293501527e-06,
+      "loss": 0.1337,
+      "step": 30855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4817774479865293e-06,
+      "loss": 0.2129,
+      "step": 30860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.480350466622906e-06,
+      "loss": 0.1939,
+      "step": 30865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4789234852592826e-06,
+      "loss": 0.1159,
+      "step": 30870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4774965038956594e-06,
+      "loss": 0.1603,
+      "step": 30875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.476069522532036e-06,
+      "loss": 0.0404,
+      "step": 30880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4746425411684123e-06,
+      "loss": 0.0998,
+      "step": 30885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.473215559804789e-06,
+      "loss": 0.2034,
+      "step": 30890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4717885784411656e-06,
+      "loss": 0.1779,
+      "step": 30895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4703615970775423e-06,
+      "loss": 0.3657,
+      "step": 30900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4689346157139189e-06,
+      "loss": 0.1614,
+      "step": 30905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4675076343502956e-06,
+      "loss": 0.1086,
+      "step": 30910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4660806529866722e-06,
+      "loss": 0.133,
+      "step": 30915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4646536716230485e-06,
+      "loss": 0.1572,
+      "step": 30920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4632266902594253e-06,
+      "loss": 0.1444,
+      "step": 30925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4617997088958018e-06,
+      "loss": 0.0483,
+      "step": 30930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4603727275321786e-06,
+      "loss": 0.1232,
+      "step": 30935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4589457461685552e-06,
+      "loss": 0.0518,
+      "step": 30940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.457518764804932e-06,
+      "loss": 0.1726,
+      "step": 30945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4560917834413083e-06,
+      "loss": 0.2552,
+      "step": 30950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4546648020776848e-06,
+      "loss": 0.0978,
+      "step": 30955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4532378207140616e-06,
+      "loss": 0.1388,
+      "step": 30960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4518108393504381e-06,
+      "loss": 0.1976,
+      "step": 30965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4503838579868149e-06,
+      "loss": 0.1231,
+      "step": 30970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4489568766231914e-06,
+      "loss": 0.1281,
+      "step": 30975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4475298952595682e-06,
+      "loss": 0.0442,
+      "step": 30980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4461029138959445e-06,
+      "loss": 0.1627,
+      "step": 30985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.444675932532321e-06,
+      "loss": 0.2215,
+      "step": 30990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4432489511686979e-06,
+      "loss": 0.1335,
+      "step": 30995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4418219698050744e-06,
+      "loss": 0.5371,
+      "step": 31000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4403949884414512e-06,
+      "loss": 0.1816,
+      "step": 31005
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4389680070778277e-06,
+      "loss": 0.143,
+      "step": 31010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4375410257142043e-06,
+      "loss": 0.153,
+      "step": 31015
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4361140443505808e-06,
+      "loss": 0.16,
+      "step": 31020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4346870629869576e-06,
+      "loss": 0.066,
+      "step": 31025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4332600816233341e-06,
+      "loss": 0.0365,
+      "step": 31030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4318331002597109e-06,
+      "loss": 0.0748,
+      "step": 31035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4304061188960874e-06,
+      "loss": 0.1075,
+      "step": 31040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4289791375324638e-06,
+      "loss": 0.2343,
+      "step": 31045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4275521561688405e-06,
+      "loss": 0.2463,
+      "step": 31050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426125174805217e-06,
+      "loss": 0.1182,
+      "step": 31055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4246981934415939e-06,
+      "loss": 0.1763,
+      "step": 31060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4232712120779704e-06,
+      "loss": 0.1413,
+      "step": 31065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4218442307143472e-06,
+      "loss": 0.1772,
+      "step": 31070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4204172493507237e-06,
+      "loss": 0.067,
+      "step": 31075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4189902679871e-06,
+      "loss": 0.1534,
+      "step": 31080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4175632866234768e-06,
+      "loss": 0.1724,
+      "step": 31085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4161363052598534e-06,
+      "loss": 0.2235,
+      "step": 31090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4147093238962301e-06,
+      "loss": 0.1799,
+      "step": 31095
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4132823425326067e-06,
+      "loss": 0.4721,
+      "step": 31100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4118553611689834e-06,
+      "loss": 0.1463,
+      "step": 31105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4104283798053598e-06,
+      "loss": 0.1202,
+      "step": 31110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4090013984417363e-06,
+      "loss": 0.1882,
+      "step": 31115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.407574417078113e-06,
+      "loss": 0.2152,
+      "step": 31120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4061474357144896e-06,
+      "loss": 0.1173,
+      "step": 31125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4047204543508664e-06,
+      "loss": 0.1154,
+      "step": 31130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.403293472987243e-06,
+      "loss": 0.0957,
+      "step": 31135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4018664916236193e-06,
+      "loss": 0.0875,
+      "step": 31140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.400439510259996e-06,
+      "loss": 0.2762,
+      "step": 31145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3990125288963726e-06,
+      "loss": 0.232,
+      "step": 31150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3975855475327494e-06,
+      "loss": 0.0594,
+      "step": 31155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.396158566169126e-06,
+      "loss": 0.1108,
+      "step": 31160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3947315848055027e-06,
+      "loss": 0.1318,
+      "step": 31165
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3933046034418792e-06,
+      "loss": 0.1815,
+      "step": 31170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3918776220782556e-06,
+      "loss": 0.1016,
+      "step": 31175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3904506407146323e-06,
+      "loss": 0.0792,
+      "step": 31180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3890236593510089e-06,
+      "loss": 0.1084,
+      "step": 31185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3875966779873856e-06,
+      "loss": 0.1061,
+      "step": 31190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3861696966237622e-06,
+      "loss": 0.251,
+      "step": 31195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.384742715260139e-06,
+      "loss": 0.4001,
+      "step": 31200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3833157338965153e-06,
+      "loss": 0.1066,
+      "step": 31205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3818887525328918e-06,
+      "loss": 0.1573,
+      "step": 31210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3804617711692686e-06,
+      "loss": 0.1168,
+      "step": 31215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3790347898056452e-06,
+      "loss": 0.1735,
+      "step": 31220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.377607808442022e-06,
+      "loss": 0.0906,
+      "step": 31225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3761808270783985e-06,
+      "loss": 0.0829,
+      "step": 31230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3747538457147752e-06,
+      "loss": 0.0907,
+      "step": 31235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3733268643511516e-06,
+      "loss": 0.1583,
+      "step": 31240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3718998829875283e-06,
+      "loss": 0.1788,
+      "step": 31245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3704729016239049e-06,
+      "loss": 0.3387,
+      "step": 31250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3690459202602816e-06,
+      "loss": 0.1139,
+      "step": 31255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3676189388966582e-06,
+      "loss": 0.1146,
+      "step": 31260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3661919575330347e-06,
+      "loss": 0.1633,
+      "step": 31265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3647649761694113e-06,
+      "loss": 0.137,
+      "step": 31270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3633379948057878e-06,
+      "loss": 0.1519,
+      "step": 31275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3619110134421646e-06,
+      "loss": 0.1357,
+      "step": 31280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3604840320785412e-06,
+      "loss": 0.1447,
+      "step": 31285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.359057050714918e-06,
+      "loss": 0.1816,
+      "step": 31290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3576300693512945e-06,
+      "loss": 0.3118,
+      "step": 31295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3562030879876708e-06,
+      "loss": 0.3088,
+      "step": 31300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3547761066240476e-06,
+      "loss": 0.1523,
+      "step": 31305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3533491252604241e-06,
+      "loss": 0.1324,
+      "step": 31310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3519221438968009e-06,
+      "loss": 0.185,
+      "step": 31315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3504951625331774e-06,
+      "loss": 0.1934,
+      "step": 31320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3490681811695542e-06,
+      "loss": 0.0723,
+      "step": 31325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3476411998059308e-06,
+      "loss": 0.0841,
+      "step": 31330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.346214218442307e-06,
+      "loss": 0.0896,
+      "step": 31335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3447872370786839e-06,
+      "loss": 0.0524,
+      "step": 31340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3433602557150604e-06,
+      "loss": 0.2578,
+      "step": 31345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3419332743514372e-06,
+      "loss": 0.304,
+      "step": 31350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3405062929878137e-06,
+      "loss": 0.1347,
+      "step": 31355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3390793116241905e-06,
+      "loss": 0.0809,
+      "step": 31360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3376523302605668e-06,
+      "loss": 0.1255,
+      "step": 31365
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3362253488969434e-06,
+      "loss": 0.1806,
+      "step": 31370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3347983675333201e-06,
+      "loss": 0.1287,
+      "step": 31375
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3333713861696967e-06,
+      "loss": 0.0332,
+      "step": 31380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3319444048060734e-06,
+      "loss": 0.0514,
+      "step": 31385
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.33051742344245e-06,
+      "loss": 0.1219,
+      "step": 31390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3290904420788263e-06,
+      "loss": 0.1158,
+      "step": 31395
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.327663460715203e-06,
+      "loss": 0.3232,
+      "step": 31400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3262364793515796e-06,
+      "loss": 0.1389,
+      "step": 31405
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3248094979879564e-06,
+      "loss": 0.1863,
+      "step": 31410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.323382516624333e-06,
+      "loss": 0.1359,
+      "step": 31415
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3219555352607097e-06,
+      "loss": 0.0884,
+      "step": 31420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3205285538970863e-06,
+      "loss": 0.2151,
+      "step": 31425
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3191015725334628e-06,
+      "loss": 0.0598,
+      "step": 31430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3176745911698394e-06,
+      "loss": 0.0646,
+      "step": 31435
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.316247609806216e-06,
+      "loss": 0.166,
+      "step": 31440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3148206284425927e-06,
+      "loss": 0.1687,
+      "step": 31445
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3133936470789692e-06,
+      "loss": 0.2897,
+      "step": 31450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.311966665715346e-06,
+      "loss": 0.1198,
+      "step": 31455
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3105396843517223e-06,
+      "loss": 0.1161,
+      "step": 31460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.309112702988099e-06,
+      "loss": 0.1257,
+      "step": 31465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3076857216244756e-06,
+      "loss": 0.1707,
+      "step": 31470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3062587402608524e-06,
+      "loss": 0.1482,
+      "step": 31475
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.304831758897229e-06,
+      "loss": 0.1203,
+      "step": 31480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3034047775336057e-06,
+      "loss": 0.1097,
+      "step": 31485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.301977796169982e-06,
+      "loss": 0.1457,
+      "step": 31490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3005508148063586e-06,
+      "loss": 0.1297,
+      "step": 31495
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2991238334427354e-06,
+      "loss": 0.5984,
+      "step": 31500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.297696852079112e-06,
+      "loss": 0.1224,
+      "step": 31505
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2962698707154887e-06,
+      "loss": 0.1312,
+      "step": 31510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2948428893518652e-06,
+      "loss": 0.1565,
+      "step": 31515
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.293415907988242e-06,
+      "loss": 0.1429,
+      "step": 31520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2919889266246183e-06,
+      "loss": 0.1364,
+      "step": 31525
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2905619452609949e-06,
+      "loss": 0.0708,
+      "step": 31530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2891349638973716e-06,
+      "loss": 0.1318,
+      "step": 31535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2877079825337482e-06,
+      "loss": 0.1534,
+      "step": 31540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.286281001170125e-06,
+      "loss": 0.218,
+      "step": 31545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2848540198065015e-06,
+      "loss": 0.2157,
+      "step": 31550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2834270384428778e-06,
+      "loss": 0.1351,
+      "step": 31555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2820000570792546e-06,
+      "loss": 0.0787,
+      "step": 31560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2805730757156312e-06,
+      "loss": 0.13,
+      "step": 31565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.279146094352008e-06,
+      "loss": 0.0911,
+      "step": 31570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2777191129883845e-06,
+      "loss": 0.0631,
+      "step": 31575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2762921316247612e-06,
+      "loss": 0.1197,
+      "step": 31580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2748651502611378e-06,
+      "loss": 0.0612,
+      "step": 31585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2734381688975141e-06,
+      "loss": 0.1578,
+      "step": 31590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2720111875338909e-06,
+      "loss": 0.2743,
+      "step": 31595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2705842061702674e-06,
+      "loss": 0.2193,
+      "step": 31600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2691572248066442e-06,
+      "loss": 0.1699,
+      "step": 31605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2677302434430207e-06,
+      "loss": 0.1013,
+      "step": 31610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2663032620793975e-06,
+      "loss": 0.1239,
+      "step": 31615
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2648762807157738e-06,
+      "loss": 0.1823,
+      "step": 31620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2634492993521504e-06,
+      "loss": 0.1302,
+      "step": 31625
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2620223179885272e-06,
+      "loss": 0.0766,
+      "step": 31630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2605953366249037e-06,
+      "loss": 0.0278,
+      "step": 31635
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2591683552612805e-06,
+      "loss": 0.1773,
+      "step": 31640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.257741373897657e-06,
+      "loss": 0.1424,
+      "step": 31645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2563143925340336e-06,
+      "loss": 0.2133,
+      "step": 31650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2548874111704101e-06,
+      "loss": 0.1234,
+      "step": 31655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2534604298067869e-06,
+      "loss": 0.1191,
+      "step": 31660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2520334484431634e-06,
+      "loss": 0.1727,
+      "step": 31665
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.25060646707954e-06,
+      "loss": 0.1757,
+      "step": 31670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2491794857159165e-06,
+      "loss": 0.1101,
+      "step": 31675
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2477525043522933e-06,
+      "loss": 0.0878,
+      "step": 31680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2463255229886699e-06,
+      "loss": 0.217,
+      "step": 31685
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2448985416250464e-06,
+      "loss": 0.2261,
+      "step": 31690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2434715602614232e-06,
+      "loss": 0.1033,
+      "step": 31695
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2420445788977997e-06,
+      "loss": 0.5031,
+      "step": 31700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2406175975341763e-06,
+      "loss": 0.136,
+      "step": 31705
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2391906161705528e-06,
+      "loss": 0.128,
+      "step": 31710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2377636348069296e-06,
+      "loss": 0.122,
+      "step": 31715
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2363366534433061e-06,
+      "loss": 0.2872,
+      "step": 31720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2349096720796827e-06,
+      "loss": 0.0875,
+      "step": 31725
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2334826907160594e-06,
+      "loss": 0.0886,
+      "step": 31730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.232055709352436e-06,
+      "loss": 0.2008,
+      "step": 31735
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2306287279888125e-06,
+      "loss": 0.2025,
+      "step": 31740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.229201746625189e-06,
+      "loss": 0.2424,
+      "step": 31745
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2277747652615659e-06,
+      "loss": 0.423,
+      "step": 31750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2263477838979424e-06,
+      "loss": 0.124,
+      "step": 31755
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.224920802534319e-06,
+      "loss": 0.1409,
+      "step": 31760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2234938211706957e-06,
+      "loss": 0.1322,
+      "step": 31765
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2220668398070723e-06,
+      "loss": 0.1554,
+      "step": 31770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2206398584434488e-06,
+      "loss": 0.1138,
+      "step": 31775
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2192128770798254e-06,
+      "loss": 0.1026,
+      "step": 31780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.217785895716202e-06,
+      "loss": 0.0916,
+      "step": 31785
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2163589143525787e-06,
+      "loss": 0.1633,
+      "step": 31790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2149319329889552e-06,
+      "loss": 0.1434,
+      "step": 31795
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.213504951625332e-06,
+      "loss": 0.2791,
+      "step": 31800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2120779702617085e-06,
+      "loss": 0.1665,
+      "step": 31805
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.210650988898085e-06,
+      "loss": 0.171,
+      "step": 31810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2092240075344619e-06,
+      "loss": 0.1433,
+      "step": 31815
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2077970261708382e-06,
+      "loss": 0.1034,
+      "step": 31820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.206370044807215e-06,
+      "loss": 0.0952,
+      "step": 31825
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2049430634435915e-06,
+      "loss": 0.0653,
+      "step": 31830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.203516082079968e-06,
+      "loss": 0.0148,
+      "step": 31835
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2020891007163448e-06,
+      "loss": 0.1245,
+      "step": 31840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2006621193527214e-06,
+      "loss": 0.1621,
+      "step": 31845
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.199235137989098e-06,
+      "loss": 0.2584,
+      "step": 31850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1978081566254745e-06,
+      "loss": 0.1122,
+      "step": 31855
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1963811752618512e-06,
+      "loss": 0.0627,
+      "step": 31860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1949541938982278e-06,
+      "loss": 0.1209,
+      "step": 31865
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1935272125346043e-06,
+      "loss": 0.095,
+      "step": 31870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.192100231170981e-06,
+      "loss": 0.1409,
+      "step": 31875
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1906732498073576e-06,
+      "loss": 0.1354,
+      "step": 31880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1892462684437342e-06,
+      "loss": 0.1059,
+      "step": 31885
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1878192870801107e-06,
+      "loss": 0.2424,
+      "step": 31890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1863923057164875e-06,
+      "loss": 0.2392,
+      "step": 31895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.184965324352864e-06,
+      "loss": 0.4215,
+      "step": 31900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1835383429892406e-06,
+      "loss": 0.1059,
+      "step": 31905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1821113616256174e-06,
+      "loss": 0.153,
+      "step": 31910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.180684380261994e-06,
+      "loss": 0.1594,
+      "step": 31915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1792573988983705e-06,
+      "loss": 0.197,
+      "step": 31920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1778304175347472e-06,
+      "loss": 0.1146,
+      "step": 31925
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1764034361711236e-06,
+      "loss": 0.0842,
+      "step": 31930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1749764548075003e-06,
+      "loss": 0.0934,
+      "step": 31935
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1735494734438769e-06,
+      "loss": 0.1809,
+      "step": 31940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1721224920802534e-06,
+      "loss": 0.2709,
+      "step": 31945
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1706955107166302e-06,
+      "loss": 0.2268,
+      "step": 31950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1692685293530067e-06,
+      "loss": 0.1638,
+      "step": 31955
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1678415479893833e-06,
+      "loss": 0.2125,
+      "step": 31960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1664145666257599e-06,
+      "loss": 0.1602,
+      "step": 31965
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1649875852621366e-06,
+      "loss": 0.1404,
+      "step": 31970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1635606038985132e-06,
+      "loss": 0.1611,
+      "step": 31975
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1621336225348897e-06,
+      "loss": 0.0797,
+      "step": 31980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1607066411712665e-06,
+      "loss": 0.1523,
+      "step": 31985
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.159279659807643e-06,
+      "loss": 0.0897,
+      "step": 31990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1578526784440196e-06,
+      "loss": 0.272,
+      "step": 31995
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1564256970803961e-06,
+      "loss": 0.2478,
+      "step": 32000
+    },
+    {
+      "epoch": 0.89,
+      "eval_loss": 0.05729741230607033,
+      "eval_runtime": 1893.4584,
+      "eval_samples_per_second": 8.46,
+      "eval_steps_per_second": 2.115,
+      "eval_wer": 0.16078254597908404,
+      "step": 32000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1549987157167729e-06,
+      "loss": 0.117,
+      "step": 32005
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1535717343531494e-06,
+      "loss": 0.1405,
+      "step": 32010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.152144752989526e-06,
+      "loss": 0.1392,
+      "step": 32015
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1507177716259028e-06,
+      "loss": 0.1713,
+      "step": 32020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1492907902622793e-06,
+      "loss": 0.0718,
+      "step": 32025
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1478638088986559e-06,
+      "loss": 0.1474,
+      "step": 32030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1464368275350326e-06,
+      "loss": 0.0845,
+      "step": 32035
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.145009846171409e-06,
+      "loss": 0.1377,
+      "step": 32040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1435828648077857e-06,
+      "loss": 0.1113,
+      "step": 32045
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1421558834441623e-06,
+      "loss": 0.5476,
+      "step": 32050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1407289020805388e-06,
+      "loss": 0.1391,
+      "step": 32055
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1393019207169156e-06,
+      "loss": 0.1375,
+      "step": 32060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1378749393532921e-06,
+      "loss": 0.1473,
+      "step": 32065
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1364479579896689e-06,
+      "loss": 0.1647,
+      "step": 32070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1350209766260452e-06,
+      "loss": 0.1695,
+      "step": 32075
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.133593995262422e-06,
+      "loss": 0.0699,
+      "step": 32080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1321670138987985e-06,
+      "loss": 0.1329,
+      "step": 32085
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.130740032535175e-06,
+      "loss": 0.1167,
+      "step": 32090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1293130511715519e-06,
+      "loss": 0.3794,
+      "step": 32095
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1278860698079284e-06,
+      "loss": 0.2516,
+      "step": 32100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.126459088444305e-06,
+      "loss": 0.161,
+      "step": 32105
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1250321070806817e-06,
+      "loss": 0.1521,
+      "step": 32110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1236051257170583e-06,
+      "loss": 0.1331,
+      "step": 32115
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1221781443534348e-06,
+      "loss": 0.1449,
+      "step": 32120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1207511629898114e-06,
+      "loss": 0.1729,
+      "step": 32125
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1193241816261881e-06,
+      "loss": 0.0596,
+      "step": 32130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1178972002625647e-06,
+      "loss": 0.1664,
+      "step": 32135
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1164702188989412e-06,
+      "loss": 0.1217,
+      "step": 32140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.115043237535318e-06,
+      "loss": 0.238,
+      "step": 32145
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1136162561716945e-06,
+      "loss": 0.3408,
+      "step": 32150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.112189274808071e-06,
+      "loss": 0.135,
+      "step": 32155
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1107622934444476e-06,
+      "loss": 0.1022,
+      "step": 32160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1093353120808244e-06,
+      "loss": 0.1174,
+      "step": 32165
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.107908330717201e-06,
+      "loss": 0.1456,
+      "step": 32170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1064813493535775e-06,
+      "loss": 0.1333,
+      "step": 32175
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1050543679899543e-06,
+      "loss": 0.0945,
+      "step": 32180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1036273866263306e-06,
+      "loss": 0.1796,
+      "step": 32185
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1022004052627074e-06,
+      "loss": 0.1358,
+      "step": 32190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.100773423899084e-06,
+      "loss": 0.1462,
+      "step": 32195
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0993464425354605e-06,
+      "loss": 0.349,
+      "step": 32200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0979194611718372e-06,
+      "loss": 0.0936,
+      "step": 32205
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0964924798082138e-06,
+      "loss": 0.1575,
+      "step": 32210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0950654984445903e-06,
+      "loss": 0.0929,
+      "step": 32215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.093638517080967e-06,
+      "loss": 0.1964,
+      "step": 32220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0922115357173436e-06,
+      "loss": 0.0948,
+      "step": 32225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0907845543537202e-06,
+      "loss": 0.0401,
+      "step": 32230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0893575729900967e-06,
+      "loss": 0.2228,
+      "step": 32235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0879305916264735e-06,
+      "loss": 0.2275,
+      "step": 32240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.08650361026285e-06,
+      "loss": 0.1291,
+      "step": 32245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0850766288992266e-06,
+      "loss": 0.2682,
+      "step": 32250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0836496475356034e-06,
+      "loss": 0.1324,
+      "step": 32255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.08222266617198e-06,
+      "loss": 0.2206,
+      "step": 32260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0807956848083565e-06,
+      "loss": 0.1372,
+      "step": 32265
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.079368703444733e-06,
+      "loss": 0.1072,
+      "step": 32270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0779417220811098e-06,
+      "loss": 0.2071,
+      "step": 32275
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0765147407174863e-06,
+      "loss": 0.0434,
+      "step": 32280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0750877593538629e-06,
+      "loss": 0.0655,
+      "step": 32285
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0736607779902396e-06,
+      "loss": 0.1538,
+      "step": 32290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.072233796626616e-06,
+      "loss": 0.2186,
+      "step": 32295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0708068152629928e-06,
+      "loss": 0.3193,
+      "step": 32300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0693798338993693e-06,
+      "loss": 0.1266,
+      "step": 32305
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0679528525357459e-06,
+      "loss": 0.1263,
+      "step": 32310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0665258711721226e-06,
+      "loss": 0.121,
+      "step": 32315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0650988898084992e-06,
+      "loss": 0.1924,
+      "step": 32320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.063671908444876e-06,
+      "loss": 0.0982,
+      "step": 32325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0622449270812525e-06,
+      "loss": 0.0966,
+      "step": 32330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.060817945717629e-06,
+      "loss": 0.079,
+      "step": 32335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0593909643540056e-06,
+      "loss": 0.2549,
+      "step": 32340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0579639829903821e-06,
+      "loss": 0.2105,
+      "step": 32345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0565370016267589e-06,
+      "loss": 0.2951,
+      "step": 32350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0551100202631354e-06,
+      "loss": 0.0777,
+      "step": 32355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.053683038899512e-06,
+      "loss": 0.108,
+      "step": 32360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0522560575358888e-06,
+      "loss": 0.144,
+      "step": 32365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0508290761722653e-06,
+      "loss": 0.1481,
+      "step": 32370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0494020948086419e-06,
+      "loss": 0.2673,
+      "step": 32375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0479751134450184e-06,
+      "loss": 0.1464,
+      "step": 32380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0465481320813952e-06,
+      "loss": 0.0679,
+      "step": 32385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0451211507177717e-06,
+      "loss": 0.0727,
+      "step": 32390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0436941693541483e-06,
+      "loss": 0.1183,
+      "step": 32395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.042267187990525e-06,
+      "loss": 0.3782,
+      "step": 32400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0408402066269016e-06,
+      "loss": 0.1716,
+      "step": 32405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0394132252632781e-06,
+      "loss": 0.1466,
+      "step": 32410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0379862438996547e-06,
+      "loss": 0.131,
+      "step": 32415
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0365592625360314e-06,
+      "loss": 0.172,
+      "step": 32420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.035132281172408e-06,
+      "loss": 0.2046,
+      "step": 32425
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0337052998087845e-06,
+      "loss": 0.134,
+      "step": 32430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0322783184451613e-06,
+      "loss": 0.1269,
+      "step": 32435
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0308513370815379e-06,
+      "loss": 0.1931,
+      "step": 32440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0294243557179144e-06,
+      "loss": 0.3455,
+      "step": 32445
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.027997374354291e-06,
+      "loss": 0.4596,
+      "step": 32450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0265703929906675e-06,
+      "loss": 0.1474,
+      "step": 32455
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0251434116270443e-06,
+      "loss": 0.1049,
+      "step": 32460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0237164302634208e-06,
+      "loss": 0.1086,
+      "step": 32465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0222894488997974e-06,
+      "loss": 0.1306,
+      "step": 32470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0208624675361741e-06,
+      "loss": 0.1247,
+      "step": 32475
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0194354861725507e-06,
+      "loss": 0.1273,
+      "step": 32480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0180085048089272e-06,
+      "loss": 0.1708,
+      "step": 32485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0165815234453038e-06,
+      "loss": 0.1291,
+      "step": 32490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0151545420816805e-06,
+      "loss": 0.1723,
+      "step": 32495
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.013727560718057e-06,
+      "loss": 0.4103,
+      "step": 32500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0123005793544336e-06,
+      "loss": 0.1288,
+      "step": 32505
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0108735979908104e-06,
+      "loss": 0.117,
+      "step": 32510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.009446616627187e-06,
+      "loss": 0.2106,
+      "step": 32515
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0080196352635635e-06,
+      "loss": 0.1813,
+      "step": 32520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.00659265389994e-06,
+      "loss": 0.2038,
+      "step": 32525
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0051656725363168e-06,
+      "loss": 0.04,
+      "step": 32530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0037386911726934e-06,
+      "loss": 0.0859,
+      "step": 32535
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.00231170980907e-06,
+      "loss": 0.1195,
+      "step": 32540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0008847284454467e-06,
+      "loss": 0.3733,
+      "step": 32545
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.994577470818232e-07,
+      "loss": 0.4709,
+      "step": 32550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.980307657181998e-07,
+      "loss": 0.1486,
+      "step": 32555
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.966037843545765e-07,
+      "loss": 0.2014,
+      "step": 32560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.951768029909529e-07,
+      "loss": 0.129,
+      "step": 32565
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.937498216273296e-07,
+      "loss": 0.1603,
+      "step": 32570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.923228402637062e-07,
+      "loss": 0.1285,
+      "step": 32575
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.90895858900083e-07,
+      "loss": 0.132,
+      "step": 32580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.894688775364595e-07,
+      "loss": 0.0625,
+      "step": 32585
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.88041896172836e-07,
+      "loss": 0.3227,
+      "step": 32590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.866149148092128e-07,
+      "loss": 0.2231,
+      "step": 32595
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.851879334455892e-07,
+      "loss": 0.2693,
+      "step": 32600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.83760952081966e-07,
+      "loss": 0.1329,
+      "step": 32605
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.823339707183425e-07,
+      "loss": 0.0944,
+      "step": 32610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.80906989354719e-07,
+      "loss": 0.1512,
+      "step": 32615
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.794800079910958e-07,
+      "loss": 0.1565,
+      "step": 32620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.780530266274723e-07,
+      "loss": 0.0535,
+      "step": 32625
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.766260452638489e-07,
+      "loss": 0.0325,
+      "step": 32630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.751990639002254e-07,
+      "loss": 0.139,
+      "step": 32635
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.737720825366022e-07,
+      "loss": 0.1512,
+      "step": 32640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.723451011729788e-07,
+      "loss": 0.2666,
+      "step": 32645
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.709181198093553e-07,
+      "loss": 0.3004,
+      "step": 32650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.69491138445732e-07,
+      "loss": 0.1139,
+      "step": 32655
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.680641570821086e-07,
+      "loss": 0.125,
+      "step": 32660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.666371757184852e-07,
+      "loss": 0.1188,
+      "step": 32665
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.65210194354862e-07,
+      "loss": 0.1088,
+      "step": 32670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.637832129912385e-07,
+      "loss": 0.1492,
+      "step": 32675
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.62356231627615e-07,
+      "loss": 0.0954,
+      "step": 32680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.609292502639916e-07,
+      "loss": 0.1257,
+      "step": 32685
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.595022689003683e-07,
+      "loss": 0.1355,
+      "step": 32690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.580752875367449e-07,
+      "loss": 0.3235,
+      "step": 32695
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.566483061731214e-07,
+      "loss": 0.2797,
+      "step": 32700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.552213248094982e-07,
+      "loss": 0.1561,
+      "step": 32705
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.537943434458745e-07,
+      "loss": 0.1145,
+      "step": 32710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.523673620822513e-07,
+      "loss": 0.1294,
+      "step": 32715
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.50940380718628e-07,
+      "loss": 0.2222,
+      "step": 32720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.495133993550044e-07,
+      "loss": 0.061,
+      "step": 32725
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.480864179913811e-07,
+      "loss": 0.0274,
+      "step": 32730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.466594366277577e-07,
+      "loss": 0.0451,
+      "step": 32735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.452324552641343e-07,
+      "loss": 0.2271,
+      "step": 32740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.438054739005109e-07,
+      "loss": 0.197,
+      "step": 32745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.423784925368876e-07,
+      "loss": 0.3383,
+      "step": 32750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.409515111732642e-07,
+      "loss": 0.1195,
+      "step": 32755
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.395245298096408e-07,
+      "loss": 0.1709,
+      "step": 32760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.380975484460173e-07,
+      "loss": 0.108,
+      "step": 32765
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.36670567082394e-07,
+      "loss": 0.1488,
+      "step": 32770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.352435857187705e-07,
+      "loss": 0.2289,
+      "step": 32775
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.338166043551472e-07,
+      "loss": 0.05,
+      "step": 32780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.323896229915239e-07,
+      "loss": 0.0459,
+      "step": 32785
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.309626416279004e-07,
+      "loss": 0.1959,
+      "step": 32790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.295356602642771e-07,
+      "loss": 0.4017,
+      "step": 32795
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.281086789006537e-07,
+      "loss": 0.2562,
+      "step": 32800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.266816975370302e-07,
+      "loss": 0.1345,
+      "step": 32805
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.252547161734068e-07,
+      "loss": 0.1602,
+      "step": 32810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.238277348097835e-07,
+      "loss": 0.1423,
+      "step": 32815
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.2240075344616e-07,
+      "loss": 0.0745,
+      "step": 32820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.209737720825367e-07,
+      "loss": 0.0695,
+      "step": 32825
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.195467907189133e-07,
+      "loss": 0.1367,
+      "step": 32830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.181198093552898e-07,
+      "loss": 0.1857,
+      "step": 32835
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.169782242643911e-07,
+      "loss": 0.2616,
+      "step": 32840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.155512429007678e-07,
+      "loss": 0.2157,
+      "step": 32845
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.141242615371444e-07,
+      "loss": 0.2924,
+      "step": 32850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.126972801735211e-07,
+      "loss": 0.2066,
+      "step": 32855
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.112702988098975e-07,
+      "loss": 0.1573,
+      "step": 32860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.098433174462742e-07,
+      "loss": 0.1145,
+      "step": 32865
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.084163360826509e-07,
+      "loss": 0.1879,
+      "step": 32870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.069893547190274e-07,
+      "loss": 0.17,
+      "step": 32875
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.055623733554041e-07,
+      "loss": 0.0442,
+      "step": 32880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.041353919917807e-07,
+      "loss": 0.0631,
+      "step": 32885
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.027084106281573e-07,
+      "loss": 0.2354,
+      "step": 32890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.012814292645339e-07,
+      "loss": 0.1421,
+      "step": 32895
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.998544479009106e-07,
+      "loss": 0.2422,
+      "step": 32900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.98427466537287e-07,
+      "loss": 0.1385,
+      "step": 32905
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.970004851736637e-07,
+      "loss": 0.1083,
+      "step": 32910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.955735038100403e-07,
+      "loss": 0.1864,
+      "step": 32915
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.941465224464169e-07,
+      "loss": 0.149,
+      "step": 32920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.927195410827935e-07,
+      "loss": 0.2301,
+      "step": 32925
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.912925597191702e-07,
+      "loss": 0.029,
+      "step": 32930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.898655783555466e-07,
+      "loss": 0.1289,
+      "step": 32935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.884385969919233e-07,
+      "loss": 0.1008,
+      "step": 32940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.870116156283e-07,
+      "loss": 0.1501,
+      "step": 32945
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.855846342646766e-07,
+      "loss": 0.4542,
+      "step": 32950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.841576529010532e-07,
+      "loss": 0.1106,
+      "step": 32955
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.827306715374298e-07,
+      "loss": 0.1505,
+      "step": 32960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.813036901738065e-07,
+      "loss": 0.1458,
+      "step": 32965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.798767088101829e-07,
+      "loss": 0.2285,
+      "step": 32970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.784497274465596e-07,
+      "loss": 0.1022,
+      "step": 32975
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.770227460829362e-07,
+      "loss": 0.0422,
+      "step": 32980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.755957647193128e-07,
+      "loss": 0.0593,
+      "step": 32985
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.741687833556894e-07,
+      "loss": 0.2012,
+      "step": 32990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.727418019920661e-07,
+      "loss": 0.1078,
+      "step": 32995
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.713148206284426e-07,
+      "loss": 0.245,
+      "step": 33000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.698878392648193e-07,
+      "loss": 0.1325,
+      "step": 33005
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.68460857901196e-07,
+      "loss": 0.093,
+      "step": 33010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.670338765375724e-07,
+      "loss": 0.1878,
+      "step": 33015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.656068951739491e-07,
+      "loss": 0.119,
+      "step": 33020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.641799138103257e-07,
+      "loss": 0.1321,
+      "step": 33025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.627529324467024e-07,
+      "loss": 0.1079,
+      "step": 33030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.613259510830789e-07,
+      "loss": 0.1126,
+      "step": 33035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.598989697194556e-07,
+      "loss": 0.1239,
+      "step": 33040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.584719883558322e-07,
+      "loss": 0.2793,
+      "step": 33045
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.570450069922087e-07,
+      "loss": 0.5398,
+      "step": 33050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.556180256285853e-07,
+      "loss": 0.1289,
+      "step": 33055
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.54191044264962e-07,
+      "loss": 0.1211,
+      "step": 33060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.527640629013385e-07,
+      "loss": 0.1767,
+      "step": 33065
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.513370815377152e-07,
+      "loss": 0.182,
+      "step": 33070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.499101001740919e-07,
+      "loss": 0.0843,
+      "step": 33075
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.484831188104683e-07,
+      "loss": 0.218,
+      "step": 33080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.47056137446845e-07,
+      "loss": 0.1083,
+      "step": 33085
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.456291560832216e-07,
+      "loss": 0.2618,
+      "step": 33090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.442021747195982e-07,
+      "loss": 0.2359,
+      "step": 33095
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.427751933559748e-07,
+      "loss": 0.2874,
+      "step": 33100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.413482119923515e-07,
+      "loss": 0.1613,
+      "step": 33105
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.39921230628728e-07,
+      "loss": 0.1171,
+      "step": 33110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.384942492651047e-07,
+      "loss": 0.1422,
+      "step": 33115
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.370672679014813e-07,
+      "loss": 0.1607,
+      "step": 33120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.35640286537858e-07,
+      "loss": 0.0315,
+      "step": 33125
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.342133051742344e-07,
+      "loss": 0.0461,
+      "step": 33130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.327863238106111e-07,
+      "loss": 0.0458,
+      "step": 33135
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.313593424469878e-07,
+      "loss": 0.2426,
+      "step": 33140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.299323610833643e-07,
+      "loss": 0.2135,
+      "step": 33145
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.28505379719741e-07,
+      "loss": 0.2423,
+      "step": 33150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.270783983561176e-07,
+      "loss": 0.1246,
+      "step": 33155
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.256514169924941e-07,
+      "loss": 0.1177,
+      "step": 33160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.242244356288707e-07,
+      "loss": 0.1109,
+      "step": 33165
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.227974542652474e-07,
+      "loss": 0.1596,
+      "step": 33170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.213704729016239e-07,
+      "loss": 0.0513,
+      "step": 33175
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.199434915380006e-07,
+      "loss": 0.0563,
+      "step": 33180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.185165101743772e-07,
+      "loss": 0.1136,
+      "step": 33185
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.170895288107538e-07,
+      "loss": 0.2111,
+      "step": 33190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.156625474471303e-07,
+      "loss": 0.3876,
+      "step": 33195
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.14235566083507e-07,
+      "loss": 0.3911,
+      "step": 33200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.128085847198836e-07,
+      "loss": 0.0825,
+      "step": 33205
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.113816033562602e-07,
+      "loss": 0.1268,
+      "step": 33210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.099546219926369e-07,
+      "loss": 0.1564,
+      "step": 33215
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.085276406290135e-07,
+      "loss": 0.203,
+      "step": 33220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.071006592653901e-07,
+      "loss": 0.1381,
+      "step": 33225
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.056736779017667e-07,
+      "loss": 0.0553,
+      "step": 33230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.042466965381434e-07,
+      "loss": 0.1891,
+      "step": 33235
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.028197151745198e-07,
+      "loss": 0.119,
+      "step": 33240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.013927338108965e-07,
+      "loss": 0.2053,
+      "step": 33245
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.999657524472731e-07,
+      "loss": 0.2148,
+      "step": 33250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.985387710836497e-07,
+      "loss": 0.0989,
+      "step": 33255
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.971117897200263e-07,
+      "loss": 0.1083,
+      "step": 33260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.95684808356403e-07,
+      "loss": 0.1214,
+      "step": 33265
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.942578269927794e-07,
+      "loss": 0.1139,
+      "step": 33270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.928308456291561e-07,
+      "loss": 0.1189,
+      "step": 33275
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.914038642655328e-07,
+      "loss": 0.112,
+      "step": 33280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.899768829019093e-07,
+      "loss": 0.2058,
+      "step": 33285
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.88549901538286e-07,
+      "loss": 0.1444,
+      "step": 33290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.871229201746626e-07,
+      "loss": 0.2797,
+      "step": 33295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.856959388110393e-07,
+      "loss": 0.1918,
+      "step": 33300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.842689574474157e-07,
+      "loss": 0.0939,
+      "step": 33305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.828419760837924e-07,
+      "loss": 0.1664,
+      "step": 33310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.81414994720169e-07,
+      "loss": 0.0634,
+      "step": 33315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.799880133565456e-07,
+      "loss": 0.1292,
+      "step": 33320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.785610319929222e-07,
+      "loss": 0.1139,
+      "step": 33325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.771340506292989e-07,
+      "loss": 0.0575,
+      "step": 33330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.757070692656754e-07,
+      "loss": 0.0883,
+      "step": 33335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.742800879020521e-07,
+      "loss": 0.0945,
+      "step": 33340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.728531065384288e-07,
+      "loss": 0.4529,
+      "step": 33345
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.714261251748052e-07,
+      "loss": 0.2622,
+      "step": 33350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.699991438111819e-07,
+      "loss": 0.1283,
+      "step": 33355
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.685721624475585e-07,
+      "loss": 0.1511,
+      "step": 33360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.671451810839351e-07,
+      "loss": 0.1189,
+      "step": 33365
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.657181997203117e-07,
+      "loss": 0.1398,
+      "step": 33370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.642912183566884e-07,
+      "loss": 0.0603,
+      "step": 33375
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.62864236993065e-07,
+      "loss": 0.1477,
+      "step": 33380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.614372556294415e-07,
+      "loss": 0.1919,
+      "step": 33385
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.600102742658181e-07,
+      "loss": 0.0869,
+      "step": 33390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.585832929021948e-07,
+      "loss": 0.3016,
+      "step": 33395
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.571563115385713e-07,
+      "loss": 0.1987,
+      "step": 33400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.55729330174948e-07,
+      "loss": 0.1416,
+      "step": 33405
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.543023488113247e-07,
+      "loss": 0.1149,
+      "step": 33410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.528753674477012e-07,
+      "loss": 0.1641,
+      "step": 33415
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.514483860840778e-07,
+      "loss": 0.0868,
+      "step": 33420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.500214047204544e-07,
+      "loss": 0.0629,
+      "step": 33425
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.48594423356831e-07,
+      "loss": 0.0216,
+      "step": 33430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.471674419932076e-07,
+      "loss": 0.1032,
+      "step": 33435
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.457404606295843e-07,
+      "loss": 0.1709,
+      "step": 33440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.443134792659608e-07,
+      "loss": 0.175,
+      "step": 33445
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.428864979023375e-07,
+      "loss": 0.309,
+      "step": 33450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.414595165387141e-07,
+      "loss": 0.0988,
+      "step": 33455
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.400325351750906e-07,
+      "loss": 0.1871,
+      "step": 33460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.386055538114672e-07,
+      "loss": 0.0989,
+      "step": 33465
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.371785724478439e-07,
+      "loss": 0.1144,
+      "step": 33470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.357515910842205e-07,
+      "loss": 0.1737,
+      "step": 33475
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.343246097205971e-07,
+      "loss": 0.1028,
+      "step": 33480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.328976283569738e-07,
+      "loss": 0.1129,
+      "step": 33485
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.314706469933504e-07,
+      "loss": 0.17,
+      "step": 33490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.300436656297269e-07,
+      "loss": 0.1031,
+      "step": 33495
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.286166842661035e-07,
+      "loss": 0.2834,
+      "step": 33500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.271897029024802e-07,
+      "loss": 0.0981,
+      "step": 33505
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.257627215388567e-07,
+      "loss": 0.0891,
+      "step": 33510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.243357401752334e-07,
+      "loss": 0.1161,
+      "step": 33515
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.2290875881161e-07,
+      "loss": 0.1518,
+      "step": 33520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.214817774479866e-07,
+      "loss": 0.1283,
+      "step": 33525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.200547960843631e-07,
+      "loss": 0.0354,
+      "step": 33530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.186278147207398e-07,
+      "loss": 0.0688,
+      "step": 33535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.172008333571163e-07,
+      "loss": 0.0727,
+      "step": 33540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.15773851993493e-07,
+      "loss": 0.0814,
+      "step": 33545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.143468706298696e-07,
+      "loss": 0.5167,
+      "step": 33550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.129198892662463e-07,
+      "loss": 0.1586,
+      "step": 33555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.114929079026229e-07,
+      "loss": 0.1507,
+      "step": 33560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.100659265389995e-07,
+      "loss": 0.0854,
+      "step": 33565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.086389451753762e-07,
+      "loss": 0.1534,
+      "step": 33570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.072119638117526e-07,
+      "loss": 0.0691,
+      "step": 33575
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.057849824481293e-07,
+      "loss": 0.0863,
+      "step": 33580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.043580010845059e-07,
+      "loss": 0.216,
+      "step": 33585
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.029310197208825e-07,
+      "loss": 0.2592,
+      "step": 33590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.015040383572591e-07,
+      "loss": 0.2099,
+      "step": 33595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.000770569936358e-07,
+      "loss": 0.4863,
+      "step": 33600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.986500756300122e-07,
+      "loss": 0.1039,
+      "step": 33605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.972230942663889e-07,
+      "loss": 0.1078,
+      "step": 33610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.957961129027655e-07,
+      "loss": 0.1677,
+      "step": 33615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.943691315391421e-07,
+      "loss": 0.1357,
+      "step": 33620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.929421501755188e-07,
+      "loss": 0.1395,
+      "step": 33625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.915151688118954e-07,
+      "loss": 0.0897,
+      "step": 33630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.900881874482721e-07,
+      "loss": 0.127,
+      "step": 33635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.886612060846486e-07,
+      "loss": 0.112,
+      "step": 33640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.872342247210252e-07,
+      "loss": 0.2524,
+      "step": 33645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.858072433574018e-07,
+      "loss": 0.3926,
+      "step": 33650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.843802619937784e-07,
+      "loss": 0.1851,
+      "step": 33655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.82953280630155e-07,
+      "loss": 0.1679,
+      "step": 33660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.815262992665317e-07,
+      "loss": 0.118,
+      "step": 33665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.800993179029082e-07,
+      "loss": 0.151,
+      "step": 33670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.786723365392849e-07,
+      "loss": 0.1039,
+      "step": 33675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.772453551756615e-07,
+      "loss": 0.0517,
+      "step": 33680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.75818373812038e-07,
+      "loss": 0.0684,
+      "step": 33685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.743913924484146e-07,
+      "loss": 0.2406,
+      "step": 33690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.729644110847913e-07,
+      "loss": 0.22,
+      "step": 33695
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.715374297211679e-07,
+      "loss": 0.3595,
+      "step": 33700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.701104483575445e-07,
+      "loss": 0.0667,
+      "step": 33705
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.686834669939212e-07,
+      "loss": 0.118,
+      "step": 33710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.672564856302976e-07,
+      "loss": 0.1186,
+      "step": 33715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.658295042666743e-07,
+      "loss": 0.2361,
+      "step": 33720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.644025229030509e-07,
+      "loss": 0.082,
+      "step": 33725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.629755415394276e-07,
+      "loss": 0.0234,
+      "step": 33730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.615485601758041e-07,
+      "loss": 0.2071,
+      "step": 33735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.601215788121808e-07,
+      "loss": 0.1544,
+      "step": 33740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.586945974485574e-07,
+      "loss": 0.2061,
+      "step": 33745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.57267616084934e-07,
+      "loss": 0.4038,
+      "step": 33750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.558406347213105e-07,
+      "loss": 0.1776,
+      "step": 33755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.544136533576872e-07,
+      "loss": 0.0942,
+      "step": 33760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.529866719940638e-07,
+      "loss": 0.1116,
+      "step": 33765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.515596906304404e-07,
+      "loss": 0.1359,
+      "step": 33770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.501327092668171e-07,
+      "loss": 0.2254,
+      "step": 33775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.487057279031936e-07,
+      "loss": 0.0678,
+      "step": 33780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.472787465395703e-07,
+      "loss": 0.1444,
+      "step": 33785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.458517651759469e-07,
+      "loss": 0.1639,
+      "step": 33790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.444247838123234e-07,
+      "loss": 0.2531,
+      "step": 33795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.429978024487e-07,
+      "loss": 0.3698,
+      "step": 33800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.415708210850767e-07,
+      "loss": 0.2081,
+      "step": 33805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.401438397214533e-07,
+      "loss": 0.1877,
+      "step": 33810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.387168583578299e-07,
+      "loss": 0.1325,
+      "step": 33815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.372898769942065e-07,
+      "loss": 0.1665,
+      "step": 33820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.358628956305832e-07,
+      "loss": 0.0933,
+      "step": 33825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.344359142669596e-07,
+      "loss": 0.1967,
+      "step": 33830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.330089329033363e-07,
+      "loss": 0.1324,
+      "step": 33835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.31581951539713e-07,
+      "loss": 0.2487,
+      "step": 33840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.301549701760895e-07,
+      "loss": 0.2832,
+      "step": 33845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.287279888124662e-07,
+      "loss": 0.2227,
+      "step": 33850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.273010074488428e-07,
+      "loss": 0.1731,
+      "step": 33855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.258740260852194e-07,
+      "loss": 0.1842,
+      "step": 33860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.24447044721596e-07,
+      "loss": 0.1187,
+      "step": 33865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.230200633579726e-07,
+      "loss": 0.1944,
+      "step": 33870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.215930819943492e-07,
+      "loss": 0.0547,
+      "step": 33875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.201661006307258e-07,
+      "loss": 0.162,
+      "step": 33880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.187391192671024e-07,
+      "loss": 0.1699,
+      "step": 33885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.17312137903479e-07,
+      "loss": 0.1574,
+      "step": 33890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.158851565398557e-07,
+      "loss": 0.3172,
+      "step": 33895
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.144581751762323e-07,
+      "loss": 0.1541,
+      "step": 33900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.130311938126089e-07,
+      "loss": 0.1386,
+      "step": 33905
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.116042124489854e-07,
+      "loss": 0.1226,
+      "step": 33910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.101772310853621e-07,
+      "loss": 0.1656,
+      "step": 33915
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.087502497217387e-07,
+      "loss": 0.1377,
+      "step": 33920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.073232683581153e-07,
+      "loss": 0.1143,
+      "step": 33925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.058962869944919e-07,
+      "loss": 0.119,
+      "step": 33930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.044693056308685e-07,
+      "loss": 0.1434,
+      "step": 33935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.030423242672451e-07,
+      "loss": 0.199,
+      "step": 33940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.016153429036217e-07,
+      "loss": 0.2497,
+      "step": 33945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.001883615399983e-07,
+      "loss": 0.2519,
+      "step": 33950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.98761380176375e-07,
+      "loss": 0.0983,
+      "step": 33955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.973343988127515e-07,
+      "loss": 0.1828,
+      "step": 33960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.959074174491282e-07,
+      "loss": 0.1122,
+      "step": 33965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.944804360855048e-07,
+      "loss": 0.179,
+      "step": 33970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.930534547218814e-07,
+      "loss": 0.084,
+      "step": 33975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.916264733582581e-07,
+      "loss": 0.0931,
+      "step": 33980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.901994919946346e-07,
+      "loss": 0.035,
+      "step": 33985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.887725106310112e-07,
+      "loss": 0.2044,
+      "step": 33990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.873455292673878e-07,
+      "loss": 0.1952,
+      "step": 33995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.859185479037644e-07,
+      "loss": 0.372,
+      "step": 34000
+    },
+    {
+      "epoch": 0.94,
+      "eval_loss": 0.05467037484049797,
+      "eval_runtime": 1883.2508,
+      "eval_samples_per_second": 8.506,
+      "eval_steps_per_second": 2.127,
+      "eval_wer": 0.1568698160836639,
+      "step": 34000
+    }
+  ],
+  "max_steps": 36039,
+  "num_train_epochs": 1,
+  "total_flos": 4.299772388019842e+18,
+  "trial_name": null,
+  "trial_params": null
+}