diff --git "a/checkpoint-16000/trainer_state.json" "b/checkpoint-16000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-16000/trainer_state.json"
@@ -0,0 +1,19360 @@
+{
+  "best_metric": 0.023078329861164093,
+  "best_model_checkpoint": "wav2vec2-base-pemlsb-la/checkpoint-16000",
+  "epoch": 3.5522905741002804,
+  "global_step": 16000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 0.0677,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 0.0551,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999444629567922e-05,
+      "loss": 0.0679,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9988892591358436e-05,
+      "loss": 0.0874,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9983338887037656e-05,
+      "loss": 0.0561,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9977785182716875e-05,
+      "loss": 0.0428,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997223147839609e-05,
+      "loss": 0.0696,
+      "step": 35
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996667777407531e-05,
+      "loss": 0.0754,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996112406975453e-05,
+      "loss": 0.1507,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9955570365433744e-05,
+      "loss": 0.26,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9950016661112963e-05,
+      "loss": 0.0669,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9944462956792183e-05,
+      "loss": 0.0512,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938909252471398e-05,
+      "loss": 0.0708,
+      "step": 65
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.993335554815062e-05,
+      "loss": 0.0813,
+      "step": 70
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9927801843829837e-05,
+      "loss": 0.0551,
+      "step": 75
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9922248139509056e-05,
+      "loss": 0.0514,
+      "step": 80
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991669443518827e-05,
+      "loss": 0.0539,
+      "step": 85
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991114073086749e-05,
+      "loss": 0.1137,
+      "step": 90
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990558702654671e-05,
+      "loss": 0.1722,
+      "step": 95
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9900033322225925e-05,
+      "loss": 0.2677,
+      "step": 100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9894479617905144e-05,
+      "loss": 0.0541,
+      "step": 105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9888925913584364e-05,
+      "loss": 0.0537,
+      "step": 110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.988337220926358e-05,
+      "loss": 0.0705,
+      "step": 115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98778185049428e-05,
+      "loss": 0.076,
+      "step": 120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9872264800622017e-05,
+      "loss": 0.0569,
+      "step": 125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9866711096301233e-05,
+      "loss": 0.031,
+      "step": 130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9861157391980452e-05,
+      "loss": 0.0752,
+      "step": 135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.985560368765967e-05,
+      "loss": 0.07,
+      "step": 140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9850049983338887e-05,
+      "loss": 0.19,
+      "step": 145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9844496279018106e-05,
+      "loss": 0.3142,
+      "step": 150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984005331556148e-05,
+      "loss": 0.3666,
+      "step": 155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9834499611240698e-05,
+      "loss": 0.0794,
+      "step": 160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9828945906919917e-05,
+      "loss": 0.0642,
+      "step": 165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9823392202599133e-05,
+      "loss": 0.0725,
+      "step": 170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9817838498278352e-05,
+      "loss": 0.0734,
+      "step": 175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.981228479395757e-05,
+      "loss": 0.0704,
+      "step": 180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.980673108963679e-05,
+      "loss": 0.0496,
+      "step": 185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.980117738531601e-05,
+      "loss": 0.1102,
+      "step": 190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9795623680995225e-05,
+      "loss": 0.1329,
+      "step": 195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9790069976674444e-05,
+      "loss": 0.3449,
+      "step": 200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9784516272353663e-05,
+      "loss": 0.0658,
+      "step": 205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.977896256803288e-05,
+      "loss": 0.0666,
+      "step": 210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9773408863712098e-05,
+      "loss": 0.0651,
+      "step": 215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9767855159391317e-05,
+      "loss": 0.0854,
+      "step": 220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9762301455070533e-05,
+      "loss": 0.0707,
+      "step": 225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9756747750749752e-05,
+      "loss": 0.0457,
+      "step": 230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.975119404642897e-05,
+      "loss": 0.0872,
+      "step": 235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9745640342108187e-05,
+      "loss": 0.1035,
+      "step": 240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9740086637787406e-05,
+      "loss": 0.1467,
+      "step": 245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9734532933466625e-05,
+      "loss": 0.2719,
+      "step": 250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.972897922914584e-05,
+      "loss": 0.0747,
+      "step": 255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.972342552482506e-05,
+      "loss": 0.0612,
+      "step": 260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.971787182050428e-05,
+      "loss": 0.0653,
+      "step": 265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9712318116183495e-05,
+      "loss": 0.0722,
+      "step": 270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9706764411862714e-05,
+      "loss": 0.0797,
+      "step": 275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9701210707541933e-05,
+      "loss": 0.0353,
+      "step": 280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.969565700322115e-05,
+      "loss": 0.0656,
+      "step": 285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9690103298900368e-05,
+      "loss": 0.0801,
+      "step": 290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9684549594579587e-05,
+      "loss": 0.1436,
+      "step": 295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9678995890258806e-05,
+      "loss": 0.2955,
+      "step": 300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.967344218593802e-05,
+      "loss": 0.0663,
+      "step": 305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.966788848161724e-05,
+      "loss": 0.0629,
+      "step": 310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.966233477729646e-05,
+      "loss": 0.0624,
+      "step": 315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9656781072975676e-05,
+      "loss": 0.0891,
+      "step": 320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9651227368654895e-05,
+      "loss": 0.0625,
+      "step": 325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9645673664334114e-05,
+      "loss": 0.0364,
+      "step": 330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.964011996001333e-05,
+      "loss": 0.0835,
+      "step": 335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.963456625569255e-05,
+      "loss": 0.1232,
+      "step": 340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9629012551371768e-05,
+      "loss": 0.1574,
+      "step": 345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9623458847050983e-05,
+      "loss": 0.2321,
+      "step": 350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9617905142730203e-05,
+      "loss": 0.0607,
+      "step": 355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9612351438409422e-05,
+      "loss": 0.0635,
+      "step": 360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9606797734088637e-05,
+      "loss": 0.074,
+      "step": 365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9601244029767857e-05,
+      "loss": 0.0797,
+      "step": 370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9595690325447076e-05,
+      "loss": 0.0589,
+      "step": 375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.959013662112629e-05,
+      "loss": 0.0473,
+      "step": 380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958458291680551e-05,
+      "loss": 0.0881,
+      "step": 385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957902921248473e-05,
+      "loss": 0.0976,
+      "step": 390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9573475508163945e-05,
+      "loss": 0.1184,
+      "step": 395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9567921803843164e-05,
+      "loss": 0.2687,
+      "step": 400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9562368099522384e-05,
+      "loss": 0.0591,
+      "step": 405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9556814395201603e-05,
+      "loss": 0.0853,
+      "step": 410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.955126069088082e-05,
+      "loss": 0.072,
+      "step": 415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9545706986560037e-05,
+      "loss": 0.0887,
+      "step": 420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9540153282239257e-05,
+      "loss": 0.0519,
+      "step": 425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9534599577918472e-05,
+      "loss": 0.0556,
+      "step": 430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952904587359769e-05,
+      "loss": 0.0544,
+      "step": 435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952349216927691e-05,
+      "loss": 0.124,
+      "step": 440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9517938464956126e-05,
+      "loss": 0.1793,
+      "step": 445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9512384760635345e-05,
+      "loss": 0.2512,
+      "step": 450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9506831056314564e-05,
+      "loss": 0.0708,
+      "step": 455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950127735199378e-05,
+      "loss": 0.0921,
+      "step": 460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9495723647673e-05,
+      "loss": 0.0661,
+      "step": 465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.949016994335222e-05,
+      "loss": 0.0676,
+      "step": 470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9484616239031434e-05,
+      "loss": 0.0592,
+      "step": 475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9479062534710653e-05,
+      "loss": 0.052,
+      "step": 480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9473508830389872e-05,
+      "loss": 0.0671,
+      "step": 485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9467955126069088e-05,
+      "loss": 0.076,
+      "step": 490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9462401421748307e-05,
+      "loss": 0.1586,
+      "step": 495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9456847717427526e-05,
+      "loss": 0.2977,
+      "step": 500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9451294013106742e-05,
+      "loss": 0.0826,
+      "step": 505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944574030878596e-05,
+      "loss": 0.0645,
+      "step": 510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944018660446518e-05,
+      "loss": 0.0949,
+      "step": 515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.94346329001444e-05,
+      "loss": 0.0795,
+      "step": 520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942907919582362e-05,
+      "loss": 0.0774,
+      "step": 525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9423525491502834e-05,
+      "loss": 0.0422,
+      "step": 530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9417971787182053e-05,
+      "loss": 0.0781,
+      "step": 535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.941241808286127e-05,
+      "loss": 0.0998,
+      "step": 540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9406864378540488e-05,
+      "loss": 0.1149,
+      "step": 545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9401310674219707e-05,
+      "loss": 0.2637,
+      "step": 550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9395756969898923e-05,
+      "loss": 0.0631,
+      "step": 555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9390203265578142e-05,
+      "loss": 0.0629,
+      "step": 560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.938464956125736e-05,
+      "loss": 0.0595,
+      "step": 565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9379095856936577e-05,
+      "loss": 0.0849,
+      "step": 570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9373542152615796e-05,
+      "loss": 0.0792,
+      "step": 575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9367988448295015e-05,
+      "loss": 0.0548,
+      "step": 580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.936243474397423e-05,
+      "loss": 0.0595,
+      "step": 585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.935688103965345e-05,
+      "loss": 0.0836,
+      "step": 590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.935132733533267e-05,
+      "loss": 0.118,
+      "step": 595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9345773631011885e-05,
+      "loss": 0.3309,
+      "step": 600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9340219926691104e-05,
+      "loss": 0.0595,
+      "step": 605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334666222370323e-05,
+      "loss": 0.0707,
+      "step": 610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932911251804954e-05,
+      "loss": 0.0811,
+      "step": 615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323558813728758e-05,
+      "loss": 0.0749,
+      "step": 620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318005109407977e-05,
+      "loss": 0.0548,
+      "step": 625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9312451405087196e-05,
+      "loss": 0.0403,
+      "step": 630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9306897700766415e-05,
+      "loss": 0.069,
+      "step": 635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.930134399644563e-05,
+      "loss": 0.092,
+      "step": 640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.929579029212485e-05,
+      "loss": 0.1074,
+      "step": 645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.929023658780407e-05,
+      "loss": 0.2775,
+      "step": 650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284682883483285e-05,
+      "loss": 0.0768,
+      "step": 655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279129179162504e-05,
+      "loss": 0.0818,
+      "step": 660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273575474841723e-05,
+      "loss": 0.0752,
+      "step": 665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926802177052094e-05,
+      "loss": 0.0852,
+      "step": 670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262468066200158e-05,
+      "loss": 0.0708,
+      "step": 675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256914361879374e-05,
+      "loss": 0.0614,
+      "step": 680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251360657558593e-05,
+      "loss": 0.0653,
+      "step": 685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245806953237812e-05,
+      "loss": 0.1111,
+      "step": 690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240253248917028e-05,
+      "loss": 0.0991,
+      "step": 695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9234699544596247e-05,
+      "loss": 0.2361,
+      "step": 700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9229145840275466e-05,
+      "loss": 0.0592,
+      "step": 705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.922359213595468e-05,
+      "loss": 0.0742,
+      "step": 710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.92180384316339e-05,
+      "loss": 0.0636,
+      "step": 715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.921248472731312e-05,
+      "loss": 0.1021,
+      "step": 720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9206931022992335e-05,
+      "loss": 0.0656,
+      "step": 725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9201377318671555e-05,
+      "loss": 0.0385,
+      "step": 730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9195823614350774e-05,
+      "loss": 0.0805,
+      "step": 735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919026991002999e-05,
+      "loss": 0.1005,
+      "step": 740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9184716205709212e-05,
+      "loss": 0.1411,
+      "step": 745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9179162501388428e-05,
+      "loss": 0.3158,
+      "step": 750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9173608797067647e-05,
+      "loss": 0.0619,
+      "step": 755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9168055092746866e-05,
+      "loss": 0.0817,
+      "step": 760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.916250138842608e-05,
+      "loss": 0.0669,
+      "step": 765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.91569476841053e-05,
+      "loss": 0.0773,
+      "step": 770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.915139397978452e-05,
+      "loss": 0.0831,
+      "step": 775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9145840275463736e-05,
+      "loss": 0.0395,
+      "step": 780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9140286571142955e-05,
+      "loss": 0.0708,
+      "step": 785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9134732866822174e-05,
+      "loss": 0.0814,
+      "step": 790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.912917916250139e-05,
+      "loss": 0.1345,
+      "step": 795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.912362545818061e-05,
+      "loss": 0.341,
+      "step": 800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9118071753859824e-05,
+      "loss": 0.0611,
+      "step": 805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9112518049539043e-05,
+      "loss": 0.0628,
+      "step": 810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9106964345218263e-05,
+      "loss": 0.0665,
+      "step": 815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9101410640897478e-05,
+      "loss": 0.0793,
+      "step": 820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9095856936576697e-05,
+      "loss": 0.0545,
+      "step": 825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9090303232255916e-05,
+      "loss": 0.0474,
+      "step": 830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9084749527935132e-05,
+      "loss": 0.1069,
+      "step": 835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.907919582361435e-05,
+      "loss": 0.0886,
+      "step": 840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.907364211929357e-05,
+      "loss": 0.1219,
+      "step": 845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9068088414972786e-05,
+      "loss": 0.2779,
+      "step": 850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9062534710652005e-05,
+      "loss": 0.0628,
+      "step": 855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9056981006331224e-05,
+      "loss": 0.0615,
+      "step": 860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9051427302010443e-05,
+      "loss": 0.0576,
+      "step": 865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9045873597689663e-05,
+      "loss": 0.0787,
+      "step": 870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9040319893368878e-05,
+      "loss": 0.0577,
+      "step": 875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9034766189048097e-05,
+      "loss": 0.0466,
+      "step": 880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9029212484727317e-05,
+      "loss": 0.0917,
+      "step": 885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9023658780406532e-05,
+      "loss": 0.0905,
+      "step": 890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.901810507608575e-05,
+      "loss": 0.1263,
+      "step": 895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.901255137176497e-05,
+      "loss": 0.3009,
+      "step": 900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9006997667444186e-05,
+      "loss": 0.061,
+      "step": 905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9001443963123405e-05,
+      "loss": 0.0815,
+      "step": 910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8995890258802624e-05,
+      "loss": 0.0778,
+      "step": 915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.899033655448184e-05,
+      "loss": 0.0929,
+      "step": 920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.898478285016106e-05,
+      "loss": 0.0531,
+      "step": 925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8979229145840275e-05,
+      "loss": 0.0561,
+      "step": 930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8973675441519494e-05,
+      "loss": 0.091,
+      "step": 935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8968121737198713e-05,
+      "loss": 0.0851,
+      "step": 940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.896256803287793e-05,
+      "loss": 0.156,
+      "step": 945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8957014328557148e-05,
+      "loss": 0.3246,
+      "step": 950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8951460624236367e-05,
+      "loss": 0.0509,
+      "step": 955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8945906919915583e-05,
+      "loss": 0.0738,
+      "step": 960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8940353215594802e-05,
+      "loss": 0.0568,
+      "step": 965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.893479951127402e-05,
+      "loss": 0.0632,
+      "step": 970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.892924580695324e-05,
+      "loss": 0.0618,
+      "step": 975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.892369210263246e-05,
+      "loss": 0.0399,
+      "step": 980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8918138398311675e-05,
+      "loss": 0.0577,
+      "step": 985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8912584693990894e-05,
+      "loss": 0.1051,
+      "step": 990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8907030989670113e-05,
+      "loss": 0.1668,
+      "step": 995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.890147728534933e-05,
+      "loss": 0.2756,
+      "step": 1000
+    },
+    {
+      "epoch": 0.22,
+      "eval_loss": 0.03770424798130989,
+      "eval_runtime": 1663.1648,
+      "eval_samples_per_second": 9.631,
+      "eval_steps_per_second": 2.408,
+      "eval_wer": 0.08191489361702127,
+      "step": 1000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8895923581028548e-05,
+      "loss": 0.0562,
+      "step": 1005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8890369876707767e-05,
+      "loss": 0.0584,
+      "step": 1010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8884816172386983e-05,
+      "loss": 0.0629,
+      "step": 1015
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8879262468066202e-05,
+      "loss": 0.0782,
+      "step": 1020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.887370876374542e-05,
+      "loss": 0.0635,
+      "step": 1025
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8868155059424637e-05,
+      "loss": 0.056,
+      "step": 1030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8862601355103856e-05,
+      "loss": 0.0572,
+      "step": 1035
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8857047650783075e-05,
+      "loss": 0.091,
+      "step": 1040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.885149394646229e-05,
+      "loss": 0.1259,
+      "step": 1045
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.884594024214151e-05,
+      "loss": 0.2509,
+      "step": 1050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.884038653782073e-05,
+      "loss": 0.0615,
+      "step": 1055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8834832833499945e-05,
+      "loss": 0.0718,
+      "step": 1060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8829279129179164e-05,
+      "loss": 0.0776,
+      "step": 1065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.882372542485838e-05,
+      "loss": 0.0713,
+      "step": 1070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.88181717205376e-05,
+      "loss": 0.09,
+      "step": 1075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8812618016216818e-05,
+      "loss": 0.0354,
+      "step": 1080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8807064311896037e-05,
+      "loss": 0.0767,
+      "step": 1085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8801510607575256e-05,
+      "loss": 0.0888,
+      "step": 1090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8795956903254472e-05,
+      "loss": 0.0972,
+      "step": 1095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.879040319893369e-05,
+      "loss": 0.2538,
+      "step": 1100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.878484949461291e-05,
+      "loss": 0.0686,
+      "step": 1105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8779295790292126e-05,
+      "loss": 0.0704,
+      "step": 1110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8773742085971345e-05,
+      "loss": 0.0669,
+      "step": 1115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8768188381650564e-05,
+      "loss": 0.0973,
+      "step": 1120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.876263467732978e-05,
+      "loss": 0.0577,
+      "step": 1125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8757080973009e-05,
+      "loss": 0.062,
+      "step": 1130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8751527268688218e-05,
+      "loss": 0.0728,
+      "step": 1135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8745973564367434e-05,
+      "loss": 0.0972,
+      "step": 1140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8740419860046653e-05,
+      "loss": 0.1226,
+      "step": 1145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8734866155725872e-05,
+      "loss": 0.3245,
+      "step": 1150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8729312451405088e-05,
+      "loss": 0.0583,
+      "step": 1155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8723758747084307e-05,
+      "loss": 0.0518,
+      "step": 1160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8718205042763526e-05,
+      "loss": 0.0656,
+      "step": 1165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.871265133844274e-05,
+      "loss": 0.0784,
+      "step": 1170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.870709763412196e-05,
+      "loss": 0.0466,
+      "step": 1175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.870154392980118e-05,
+      "loss": 0.0414,
+      "step": 1180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8695990225480395e-05,
+      "loss": 0.0879,
+      "step": 1185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8690436521159615e-05,
+      "loss": 0.1077,
+      "step": 1190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8684882816838834e-05,
+      "loss": 0.1001,
+      "step": 1195
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8679329112518053e-05,
+      "loss": 0.3071,
+      "step": 1200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.867377540819727e-05,
+      "loss": 0.0635,
+      "step": 1205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8668221703876488e-05,
+      "loss": 0.0754,
+      "step": 1210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8662667999555707e-05,
+      "loss": 0.0631,
+      "step": 1215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8657114295234922e-05,
+      "loss": 0.0807,
+      "step": 1220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.865156059091414e-05,
+      "loss": 0.075,
+      "step": 1225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.864600688659336e-05,
+      "loss": 0.0461,
+      "step": 1230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8640453182272576e-05,
+      "loss": 0.0734,
+      "step": 1235
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8634899477951795e-05,
+      "loss": 0.1162,
+      "step": 1240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8629345773631015e-05,
+      "loss": 0.1143,
+      "step": 1245
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.862379206931023e-05,
+      "loss": 0.2313,
+      "step": 1250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.861823836498945e-05,
+      "loss": 0.0716,
+      "step": 1255
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.861268466066867e-05,
+      "loss": 0.0638,
+      "step": 1260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8607130956347884e-05,
+      "loss": 0.0612,
+      "step": 1265
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8601577252027103e-05,
+      "loss": 0.0792,
+      "step": 1270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8596023547706322e-05,
+      "loss": 0.0522,
+      "step": 1275
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8590469843385538e-05,
+      "loss": 0.0443,
+      "step": 1280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8584916139064757e-05,
+      "loss": 0.0811,
+      "step": 1285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8579362434743976e-05,
+      "loss": 0.0887,
+      "step": 1290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8573808730423192e-05,
+      "loss": 0.1432,
+      "step": 1295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.856825502610241e-05,
+      "loss": 0.3228,
+      "step": 1300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.856270132178163e-05,
+      "loss": 0.0638,
+      "step": 1305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.855714761746085e-05,
+      "loss": 0.0677,
+      "step": 1310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8551593913140065e-05,
+      "loss": 0.1845,
+      "step": 1315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8546040208819284e-05,
+      "loss": 0.0871,
+      "step": 1320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8540486504498503e-05,
+      "loss": 0.068,
+      "step": 1325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.853493280017772e-05,
+      "loss": 0.0723,
+      "step": 1330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8529379095856938e-05,
+      "loss": 0.0596,
+      "step": 1335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8523825391536157e-05,
+      "loss": 0.0825,
+      "step": 1340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8518271687215373e-05,
+      "loss": 0.1266,
+      "step": 1345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8512717982894592e-05,
+      "loss": 0.2559,
+      "step": 1350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.850716427857381e-05,
+      "loss": 0.0587,
+      "step": 1355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8501610574253027e-05,
+      "loss": 0.0642,
+      "step": 1360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8496056869932246e-05,
+      "loss": 0.072,
+      "step": 1365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8490503165611465e-05,
+      "loss": 0.0705,
+      "step": 1370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.848494946129068e-05,
+      "loss": 0.0696,
+      "step": 1375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.84793957569699e-05,
+      "loss": 0.042,
+      "step": 1380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.847384205264912e-05,
+      "loss": 0.0811,
+      "step": 1385
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8468288348328335e-05,
+      "loss": 0.097,
+      "step": 1390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8462734644007554e-05,
+      "loss": 0.1241,
+      "step": 1395
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8457180939686773e-05,
+      "loss": 0.2788,
+      "step": 1400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.845162723536599e-05,
+      "loss": 0.0723,
+      "step": 1405
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8446073531045208e-05,
+      "loss": 0.0652,
+      "step": 1410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8440519826724427e-05,
+      "loss": 0.0777,
+      "step": 1415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.8434966122403646e-05,
+      "loss": 0.0725,
+      "step": 1420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.8429412418082865e-05,
+      "loss": 0.0627,
+      "step": 1425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.842385871376208e-05,
+      "loss": 0.0649,
+      "step": 1430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.84183050094413e-05,
+      "loss": 0.0628,
+      "step": 1435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.841275130512052e-05,
+      "loss": 0.0962,
+      "step": 1440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.8407197600799735e-05,
+      "loss": 0.1234,
+      "step": 1445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.8401643896478954e-05,
+      "loss": 0.2994,
+      "step": 1450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.839609019215817e-05,
+      "loss": 0.0752,
+      "step": 1455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.839053648783739e-05,
+      "loss": 0.0683,
+      "step": 1460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8384982783516608e-05,
+      "loss": 0.0677,
+      "step": 1465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8379429079195824e-05,
+      "loss": 0.0921,
+      "step": 1470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8373875374875043e-05,
+      "loss": 0.0556,
+      "step": 1475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8368321670554262e-05,
+      "loss": 0.0395,
+      "step": 1480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8362767966233478e-05,
+      "loss": 0.0568,
+      "step": 1485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8357214261912697e-05,
+      "loss": 0.102,
+      "step": 1490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.8351660557591916e-05,
+      "loss": 0.1578,
+      "step": 1495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.834610685327113e-05,
+      "loss": 0.2777,
+      "step": 1500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.834055314895035e-05,
+      "loss": 0.0615,
+      "step": 1505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.833499944462957e-05,
+      "loss": 0.0677,
+      "step": 1510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8329445740308786e-05,
+      "loss": 0.0773,
+      "step": 1515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8323892035988005e-05,
+      "loss": 0.0882,
+      "step": 1520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8318338331667224e-05,
+      "loss": 0.0622,
+      "step": 1525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.831278462734644e-05,
+      "loss": 0.0563,
+      "step": 1530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8307230923025662e-05,
+      "loss": 0.0816,
+      "step": 1535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8301677218704878e-05,
+      "loss": 0.0812,
+      "step": 1540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8296123514384097e-05,
+      "loss": 0.1448,
+      "step": 1545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8290569810063316e-05,
+      "loss": 0.2633,
+      "step": 1550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.828501610574253e-05,
+      "loss": 0.0589,
+      "step": 1555
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.827946240142175e-05,
+      "loss": 0.0736,
+      "step": 1560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.827390869710097e-05,
+      "loss": 0.0742,
+      "step": 1565
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.8268354992780186e-05,
+      "loss": 0.072,
+      "step": 1570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.8262801288459405e-05,
+      "loss": 0.0666,
+      "step": 1575
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.825724758413862e-05,
+      "loss": 0.041,
+      "step": 1580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.825169387981784e-05,
+      "loss": 0.0543,
+      "step": 1585
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.824614017549706e-05,
+      "loss": 0.0576,
+      "step": 1590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.8240586471176274e-05,
+      "loss": 0.1169,
+      "step": 1595
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8235032766855493e-05,
+      "loss": 0.2606,
+      "step": 1600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8229479062534713e-05,
+      "loss": 0.0633,
+      "step": 1605
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.822392535821393e-05,
+      "loss": 0.0707,
+      "step": 1610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8218371653893147e-05,
+      "loss": 0.0842,
+      "step": 1615
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8212817949572367e-05,
+      "loss": 0.0702,
+      "step": 1620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8207264245251582e-05,
+      "loss": 0.0674,
+      "step": 1625
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.82017105409308e-05,
+      "loss": 0.04,
+      "step": 1630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.819615683661002e-05,
+      "loss": 0.0786,
+      "step": 1635
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8190603132289236e-05,
+      "loss": 0.0981,
+      "step": 1640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.818504942796846e-05,
+      "loss": 0.1502,
+      "step": 1645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8179495723647674e-05,
+      "loss": 0.3044,
+      "step": 1650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8173942019326894e-05,
+      "loss": 0.0581,
+      "step": 1655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8168388315006113e-05,
+      "loss": 0.0624,
+      "step": 1660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.816283461068533e-05,
+      "loss": 0.066,
+      "step": 1665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8157280906364547e-05,
+      "loss": 0.0712,
+      "step": 1670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8151727202043767e-05,
+      "loss": 0.073,
+      "step": 1675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8146173497722982e-05,
+      "loss": 0.0396,
+      "step": 1680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.81406197934022e-05,
+      "loss": 0.0545,
+      "step": 1685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.813506608908142e-05,
+      "loss": 0.0827,
+      "step": 1690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.8129512384760636e-05,
+      "loss": 0.1125,
+      "step": 1695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.8123958680439855e-05,
+      "loss": 0.2775,
+      "step": 1700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.811840497611907e-05,
+      "loss": 0.0644,
+      "step": 1705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.811285127179829e-05,
+      "loss": 0.0594,
+      "step": 1710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.810729756747751e-05,
+      "loss": 0.0738,
+      "step": 1715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.8101743863156725e-05,
+      "loss": 0.1034,
+      "step": 1720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.8096190158835944e-05,
+      "loss": 0.0555,
+      "step": 1725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.8090636454515163e-05,
+      "loss": 0.0358,
+      "step": 1730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.808508275019438e-05,
+      "loss": 0.0889,
+      "step": 1735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8079529045873598e-05,
+      "loss": 0.1131,
+      "step": 1740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8073975341552817e-05,
+      "loss": 0.1308,
+      "step": 1745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8068421637232033e-05,
+      "loss": 0.2393,
+      "step": 1750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8062867932911252e-05,
+      "loss": 0.0575,
+      "step": 1755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.805731422859047e-05,
+      "loss": 0.0651,
+      "step": 1760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.805176052426969e-05,
+      "loss": 0.0743,
+      "step": 1765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.804620681994891e-05,
+      "loss": 0.0716,
+      "step": 1770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8040653115628125e-05,
+      "loss": 0.067,
+      "step": 1775
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.8035099411307344e-05,
+      "loss": 0.049,
+      "step": 1780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.8029545706986563e-05,
+      "loss": 0.0699,
+      "step": 1785
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.802399200266578e-05,
+      "loss": 0.0733,
+      "step": 1790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.8018438298344998e-05,
+      "loss": 0.1237,
+      "step": 1795
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.8012884594024217e-05,
+      "loss": 0.291,
+      "step": 1800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.8007330889703433e-05,
+      "loss": 0.0661,
+      "step": 1805
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.8001777185382652e-05,
+      "loss": 0.0731,
+      "step": 1810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.799622348106187e-05,
+      "loss": 0.0665,
+      "step": 1815
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7990669776741087e-05,
+      "loss": 0.0952,
+      "step": 1820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7985116072420306e-05,
+      "loss": 0.0605,
+      "step": 1825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7979562368099525e-05,
+      "loss": 0.053,
+      "step": 1830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.797400866377874e-05,
+      "loss": 0.0644,
+      "step": 1835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.796845495945796e-05,
+      "loss": 0.0771,
+      "step": 1840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7962901255137176e-05,
+      "loss": 0.1272,
+      "step": 1845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7957347550816395e-05,
+      "loss": 0.2825,
+      "step": 1850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7951793846495614e-05,
+      "loss": 0.0873,
+      "step": 1855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.794624014217483e-05,
+      "loss": 0.0585,
+      "step": 1860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.794068643785405e-05,
+      "loss": 0.0703,
+      "step": 1865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7935132733533268e-05,
+      "loss": 0.0843,
+      "step": 1870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7929579029212487e-05,
+      "loss": 0.072,
+      "step": 1875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7924025324891706e-05,
+      "loss": 0.0398,
+      "step": 1880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7918471620570922e-05,
+      "loss": 0.0525,
+      "step": 1885
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.791291791625014e-05,
+      "loss": 0.0722,
+      "step": 1890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.790736421192936e-05,
+      "loss": 0.1567,
+      "step": 1895
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7901810507608576e-05,
+      "loss": 0.3214,
+      "step": 1900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7896256803287795e-05,
+      "loss": 0.0796,
+      "step": 1905
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7890703098967014e-05,
+      "loss": 0.0668,
+      "step": 1910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.788514939464623e-05,
+      "loss": 0.0728,
+      "step": 1915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.787959569032545e-05,
+      "loss": 0.0622,
+      "step": 1920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7874041986004668e-05,
+      "loss": 0.0511,
+      "step": 1925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7868488281683884e-05,
+      "loss": 0.0248,
+      "step": 1930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7862934577363103e-05,
+      "loss": 0.0608,
+      "step": 1935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7857380873042322e-05,
+      "loss": 0.0699,
+      "step": 1940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7851827168721538e-05,
+      "loss": 0.1303,
+      "step": 1945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7846273464400757e-05,
+      "loss": 0.2729,
+      "step": 1950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7840719760079976e-05,
+      "loss": 0.0608,
+      "step": 1955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.783516605575919e-05,
+      "loss": 0.0512,
+      "step": 1960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.782961235143841e-05,
+      "loss": 0.0928,
+      "step": 1965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7824058647117626e-05,
+      "loss": 0.0689,
+      "step": 1970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7818504942796845e-05,
+      "loss": 0.0411,
+      "step": 1975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7812951238476065e-05,
+      "loss": 0.0469,
+      "step": 1980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7807397534155284e-05,
+      "loss": 0.0803,
+      "step": 1985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7801843829834503e-05,
+      "loss": 0.0912,
+      "step": 1990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.779629012551372e-05,
+      "loss": 0.1424,
+      "step": 1995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7790736421192938e-05,
+      "loss": 0.2826,
+      "step": 2000
+    },
+    {
+      "epoch": 0.44,
+      "eval_loss": 0.031897641718387604,
+      "eval_runtime": 1519.6995,
+      "eval_samples_per_second": 10.54,
+      "eval_steps_per_second": 2.635,
+      "eval_wer": 0.07311575910566173,
+      "step": 2000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7785182716872157e-05,
+      "loss": 0.0643,
+      "step": 2005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7779629012551372e-05,
+      "loss": 0.0678,
+      "step": 2010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.777407530823059e-05,
+      "loss": 0.0624,
+      "step": 2015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.776852160390981e-05,
+      "loss": 0.0661,
+      "step": 2020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7762967899589026e-05,
+      "loss": 0.0395,
+      "step": 2025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7757414195268246e-05,
+      "loss": 0.0384,
+      "step": 2030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7751860490947465e-05,
+      "loss": 0.0549,
+      "step": 2035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.774630678662668e-05,
+      "loss": 0.0915,
+      "step": 2040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.77407530823059e-05,
+      "loss": 0.138,
+      "step": 2045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.773519937798512e-05,
+      "loss": 0.233,
+      "step": 2050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7729645673664334e-05,
+      "loss": 0.061,
+      "step": 2055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7724091969343553e-05,
+      "loss": 0.0652,
+      "step": 2060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7718538265022773e-05,
+      "loss": 0.0755,
+      "step": 2065
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7712984560701988e-05,
+      "loss": 0.0621,
+      "step": 2070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7707430856381207e-05,
+      "loss": 0.0524,
+      "step": 2075
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7701877152060426e-05,
+      "loss": 0.0556,
+      "step": 2080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.7696323447739642e-05,
+      "loss": 0.0715,
+      "step": 2085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.769076974341886e-05,
+      "loss": 0.1165,
+      "step": 2090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.768521603909808e-05,
+      "loss": 0.1427,
+      "step": 2095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.76796623347773e-05,
+      "loss": 0.2487,
+      "step": 2100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7674108630456515e-05,
+      "loss": 0.0394,
+      "step": 2105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7668554926135734e-05,
+      "loss": 0.0641,
+      "step": 2110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7663001221814953e-05,
+      "loss": 0.0677,
+      "step": 2115
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.765744751749417e-05,
+      "loss": 0.0832,
+      "step": 2120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.765189381317339e-05,
+      "loss": 0.0573,
+      "step": 2125
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7646340108852607e-05,
+      "loss": 0.0511,
+      "step": 2130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7640786404531823e-05,
+      "loss": 0.0587,
+      "step": 2135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7635232700211042e-05,
+      "loss": 0.1138,
+      "step": 2140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.762967899589026e-05,
+      "loss": 0.146,
+      "step": 2145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7624125291569477e-05,
+      "loss": 0.2942,
+      "step": 2150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7618571587248696e-05,
+      "loss": 0.0582,
+      "step": 2155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7613017882927915e-05,
+      "loss": 0.07,
+      "step": 2160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.760746417860713e-05,
+      "loss": 0.0832,
+      "step": 2165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.760191047428635e-05,
+      "loss": 0.0777,
+      "step": 2170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.759635676996557e-05,
+      "loss": 0.0456,
+      "step": 2175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7590803065644785e-05,
+      "loss": 0.0342,
+      "step": 2180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7585249361324004e-05,
+      "loss": 0.0539,
+      "step": 2185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7579695657003223e-05,
+      "loss": 0.1085,
+      "step": 2190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.757414195268244e-05,
+      "loss": 0.1241,
+      "step": 2195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7568588248361658e-05,
+      "loss": 0.2792,
+      "step": 2200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7563034544040877e-05,
+      "loss": 0.0814,
+      "step": 2205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7557480839720096e-05,
+      "loss": 0.0656,
+      "step": 2210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7551927135399315e-05,
+      "loss": 0.0663,
+      "step": 2215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.754637343107853e-05,
+      "loss": 0.0738,
+      "step": 2220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.754081972675775e-05,
+      "loss": 0.0639,
+      "step": 2225
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7535266022436966e-05,
+      "loss": 0.0404,
+      "step": 2230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7529712318116185e-05,
+      "loss": 0.0798,
+      "step": 2235
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7524158613795404e-05,
+      "loss": 0.0885,
+      "step": 2240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.751860490947462e-05,
+      "loss": 0.1042,
+      "step": 2245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.751305120515384e-05,
+      "loss": 0.2838,
+      "step": 2250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7507497500833058e-05,
+      "loss": 0.0763,
+      "step": 2255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7501943796512274e-05,
+      "loss": 0.0634,
+      "step": 2260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7496390092191493e-05,
+      "loss": 0.064,
+      "step": 2265
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.7490836387870712e-05,
+      "loss": 0.0696,
+      "step": 2270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7485282683549928e-05,
+      "loss": 0.0387,
+      "step": 2275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7479728979229147e-05,
+      "loss": 0.0425,
+      "step": 2280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7474175274908366e-05,
+      "loss": 0.0591,
+      "step": 2285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7468621570587582e-05,
+      "loss": 0.1065,
+      "step": 2290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.74630678662668e-05,
+      "loss": 0.1167,
+      "step": 2295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.745751416194602e-05,
+      "loss": 0.2663,
+      "step": 2300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7451960457625236e-05,
+      "loss": 0.0649,
+      "step": 2305
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7446406753304455e-05,
+      "loss": 0.0585,
+      "step": 2310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7440853048983674e-05,
+      "loss": 0.0613,
+      "step": 2315
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7435299344662893e-05,
+      "loss": 0.0824,
+      "step": 2320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7429745640342112e-05,
+      "loss": 0.0681,
+      "step": 2325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7424191936021328e-05,
+      "loss": 0.055,
+      "step": 2330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7418638231700547e-05,
+      "loss": 0.058,
+      "step": 2335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7413084527379766e-05,
+      "loss": 0.0787,
+      "step": 2340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7407530823058982e-05,
+      "loss": 0.1638,
+      "step": 2345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.74019771187382e-05,
+      "loss": 0.2191,
+      "step": 2350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7396423414417417e-05,
+      "loss": 0.0452,
+      "step": 2355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7390869710096636e-05,
+      "loss": 0.0674,
+      "step": 2360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.7385316005775855e-05,
+      "loss": 0.0645,
+      "step": 2365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.737976230145507e-05,
+      "loss": 0.077,
+      "step": 2370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.737420859713429e-05,
+      "loss": 0.078,
+      "step": 2375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.736865489281351e-05,
+      "loss": 0.0339,
+      "step": 2380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.7363101188492724e-05,
+      "loss": 0.0552,
+      "step": 2385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.7357547484171944e-05,
+      "loss": 0.078,
+      "step": 2390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.7351993779851163e-05,
+      "loss": 0.1273,
+      "step": 2395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.734644007553038e-05,
+      "loss": 0.2665,
+      "step": 2400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.7340886371209598e-05,
+      "loss": 0.066,
+      "step": 2405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.7335332666888817e-05,
+      "loss": 0.0541,
+      "step": 2410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.7329778962568032e-05,
+      "loss": 0.0743,
+      "step": 2415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.732422525824725e-05,
+      "loss": 0.0885,
+      "step": 2420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.731867155392647e-05,
+      "loss": 0.0679,
+      "step": 2425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.7313117849605686e-05,
+      "loss": 0.0531,
+      "step": 2430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.730756414528491e-05,
+      "loss": 0.0796,
+      "step": 2435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.7302010440964125e-05,
+      "loss": 0.122,
+      "step": 2440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.7296456736643344e-05,
+      "loss": 0.1142,
+      "step": 2445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.7290903032322563e-05,
+      "loss": 0.3012,
+      "step": 2450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.728534932800178e-05,
+      "loss": 0.0443,
+      "step": 2455
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.7279795623680998e-05,
+      "loss": 0.0694,
+      "step": 2460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.7274241919360217e-05,
+      "loss": 0.0836,
+      "step": 2465
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.7268688215039432e-05,
+      "loss": 0.0868,
+      "step": 2470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.726313451071865e-05,
+      "loss": 0.0479,
+      "step": 2475
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.7257580806397867e-05,
+      "loss": 0.0554,
+      "step": 2480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.7252027102077086e-05,
+      "loss": 0.0778,
+      "step": 2485
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.7246473397756305e-05,
+      "loss": 0.0608,
+      "step": 2490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.724091969343552e-05,
+      "loss": 0.1525,
+      "step": 2495
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.723536598911474e-05,
+      "loss": 0.2428,
+      "step": 2500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.722981228479396e-05,
+      "loss": 0.0611,
+      "step": 2505
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7224258580473175e-05,
+      "loss": 0.0587,
+      "step": 2510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7218704876152394e-05,
+      "loss": 0.081,
+      "step": 2515
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7213151171831613e-05,
+      "loss": 0.0868,
+      "step": 2520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.720759746751083e-05,
+      "loss": 0.0581,
+      "step": 2525
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7202043763190048e-05,
+      "loss": 0.0287,
+      "step": 2530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7196490058869267e-05,
+      "loss": 0.0833,
+      "step": 2535
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7190936354548483e-05,
+      "loss": 0.0729,
+      "step": 2540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.7185382650227706e-05,
+      "loss": 0.1018,
+      "step": 2545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.717982894590692e-05,
+      "loss": 0.2187,
+      "step": 2550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.717427524158614e-05,
+      "loss": 0.0624,
+      "step": 2555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.716872153726536e-05,
+      "loss": 0.0575,
+      "step": 2560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.7163167832944575e-05,
+      "loss": 0.0601,
+      "step": 2565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.7157614128623794e-05,
+      "loss": 0.0874,
+      "step": 2570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.7152060424303013e-05,
+      "loss": 0.0637,
+      "step": 2575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.714650671998223e-05,
+      "loss": 0.0546,
+      "step": 2580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.7140953015661448e-05,
+      "loss": 0.0618,
+      "step": 2585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.7135399311340667e-05,
+      "loss": 0.078,
+      "step": 2590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7129845607019883e-05,
+      "loss": 0.1608,
+      "step": 2595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7124291902699102e-05,
+      "loss": 0.3025,
+      "step": 2600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.711873819837832e-05,
+      "loss": 0.0628,
+      "step": 2605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7113184494057537e-05,
+      "loss": 0.0639,
+      "step": 2610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7107630789736756e-05,
+      "loss": 0.067,
+      "step": 2615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7102077085415972e-05,
+      "loss": 0.0703,
+      "step": 2620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.709652338109519e-05,
+      "loss": 0.0463,
+      "step": 2625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.709096967677441e-05,
+      "loss": 0.0536,
+      "step": 2630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7085415972453626e-05,
+      "loss": 0.0624,
+      "step": 2635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7079862268132845e-05,
+      "loss": 0.1174,
+      "step": 2640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7074308563812064e-05,
+      "loss": 0.1698,
+      "step": 2645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.706875485949128e-05,
+      "loss": 0.2327,
+      "step": 2650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.70632011551705e-05,
+      "loss": 0.0668,
+      "step": 2655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7057647450849718e-05,
+      "loss": 0.0581,
+      "step": 2660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7052093746528937e-05,
+      "loss": 0.078,
+      "step": 2665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7046540042208156e-05,
+      "loss": 0.0654,
+      "step": 2670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7040986337887372e-05,
+      "loss": 0.0614,
+      "step": 2675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.703543263356659e-05,
+      "loss": 0.0347,
+      "step": 2680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.702987892924581e-05,
+      "loss": 0.0583,
+      "step": 2685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7024325224925026e-05,
+      "loss": 0.0777,
+      "step": 2690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7018771520604245e-05,
+      "loss": 0.1457,
+      "step": 2695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7013217816283464e-05,
+      "loss": 0.2738,
+      "step": 2700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.700766411196268e-05,
+      "loss": 0.0544,
+      "step": 2705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.70021104076419e-05,
+      "loss": 0.0568,
+      "step": 2710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6996556703321118e-05,
+      "loss": 0.0622,
+      "step": 2715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6991002999000334e-05,
+      "loss": 0.0718,
+      "step": 2720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6985449294679553e-05,
+      "loss": 0.067,
+      "step": 2725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6979895590358772e-05,
+      "loss": 0.0367,
+      "step": 2730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6974341886037988e-05,
+      "loss": 0.0534,
+      "step": 2735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6968788181717207e-05,
+      "loss": 0.0859,
+      "step": 2740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6963234477396423e-05,
+      "loss": 0.163,
+      "step": 2745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.695768077307564e-05,
+      "loss": 0.2677,
+      "step": 2750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.695212706875486e-05,
+      "loss": 0.0598,
+      "step": 2755
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6946573364434076e-05,
+      "loss": 0.0671,
+      "step": 2760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6941019660113296e-05,
+      "loss": 0.0586,
+      "step": 2765
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6935465955792515e-05,
+      "loss": 0.0651,
+      "step": 2770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6929912251471734e-05,
+      "loss": 0.0607,
+      "step": 2775
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6924358547150953e-05,
+      "loss": 0.0306,
+      "step": 2780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.691880484283017e-05,
+      "loss": 0.0574,
+      "step": 2785
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6913251138509388e-05,
+      "loss": 0.0733,
+      "step": 2790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6907697434188607e-05,
+      "loss": 0.1255,
+      "step": 2795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6902143729867823e-05,
+      "loss": 0.2523,
+      "step": 2800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.689659002554704e-05,
+      "loss": 0.0643,
+      "step": 2805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.689103632122626e-05,
+      "loss": 0.067,
+      "step": 2810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6885482616905477e-05,
+      "loss": 0.0643,
+      "step": 2815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6879928912584696e-05,
+      "loss": 0.0736,
+      "step": 2820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6874375208263915e-05,
+      "loss": 0.0469,
+      "step": 2825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.686882150394313e-05,
+      "loss": 0.0458,
+      "step": 2830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.686326779962235e-05,
+      "loss": 0.0938,
+      "step": 2835
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.685771409530157e-05,
+      "loss": 0.092,
+      "step": 2840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6852160390980784e-05,
+      "loss": 0.1454,
+      "step": 2845
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6846606686660004e-05,
+      "loss": 0.2731,
+      "step": 2850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6841052982339223e-05,
+      "loss": 0.0584,
+      "step": 2855
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.683549927801844e-05,
+      "loss": 0.0719,
+      "step": 2860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.6829945573697657e-05,
+      "loss": 0.0624,
+      "step": 2865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.6824391869376873e-05,
+      "loss": 0.0697,
+      "step": 2870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.6818838165056092e-05,
+      "loss": 0.0554,
+      "step": 2875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.681328446073531e-05,
+      "loss": 0.039,
+      "step": 2880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.680773075641453e-05,
+      "loss": 0.0539,
+      "step": 2885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.680217705209375e-05,
+      "loss": 0.1008,
+      "step": 2890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.6796623347772965e-05,
+      "loss": 0.1097,
+      "step": 2895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.6791069643452184e-05,
+      "loss": 0.2765,
+      "step": 2900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.6785515939131404e-05,
+      "loss": 0.0659,
+      "step": 2905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.677996223481062e-05,
+      "loss": 0.0568,
+      "step": 2910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.677440853048984e-05,
+      "loss": 0.0747,
+      "step": 2915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.6768854826169058e-05,
+      "loss": 0.0966,
+      "step": 2920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.6763301121848273e-05,
+      "loss": 0.0604,
+      "step": 2925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.6757747417527492e-05,
+      "loss": 0.0613,
+      "step": 2930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.675219371320671e-05,
+      "loss": 0.0651,
+      "step": 2935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.6746640008885927e-05,
+      "loss": 0.1019,
+      "step": 2940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.6741086304565146e-05,
+      "loss": 0.1018,
+      "step": 2945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.6735532600244365e-05,
+      "loss": 0.2294,
+      "step": 2950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.672997889592358e-05,
+      "loss": 0.0606,
+      "step": 2955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.67244251916028e-05,
+      "loss": 0.073,
+      "step": 2960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.671887148728202e-05,
+      "loss": 0.0618,
+      "step": 2965
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6713317782961235e-05,
+      "loss": 0.0874,
+      "step": 2970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6707764078640454e-05,
+      "loss": 0.0496,
+      "step": 2975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6702210374319673e-05,
+      "loss": 0.0604,
+      "step": 2980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.669665666999889e-05,
+      "loss": 0.0598,
+      "step": 2985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6691102965678108e-05,
+      "loss": 0.0987,
+      "step": 2990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6685549261357327e-05,
+      "loss": 0.1404,
+      "step": 2995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6679995557036546e-05,
+      "loss": 0.3028,
+      "step": 3000
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 0.028801320120692253,
+      "eval_runtime": 1486.3915,
+      "eval_samples_per_second": 10.776,
+      "eval_steps_per_second": 2.694,
+      "eval_wer": 0.06837360259646592,
+      "step": 3000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6674441852715762e-05,
+      "loss": 0.0601,
+      "step": 3005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.666888814839498e-05,
+      "loss": 0.0579,
+      "step": 3010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.66633344440742e-05,
+      "loss": 0.0793,
+      "step": 3015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6657780739753416e-05,
+      "loss": 0.0811,
+      "step": 3020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6652227035432635e-05,
+      "loss": 0.0676,
+      "step": 3025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6646673331111854e-05,
+      "loss": 0.0507,
+      "step": 3030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.664111962679107e-05,
+      "loss": 0.084,
+      "step": 3035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.663556592247029e-05,
+      "loss": 0.0814,
+      "step": 3040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6630012218149508e-05,
+      "loss": 0.1037,
+      "step": 3045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6624458513828724e-05,
+      "loss": 0.2606,
+      "step": 3050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6618904809507943e-05,
+      "loss": 0.0569,
+      "step": 3055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6613351105187162e-05,
+      "loss": 0.0585,
+      "step": 3060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6607797400866378e-05,
+      "loss": 0.075,
+      "step": 3065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6602243696545597e-05,
+      "loss": 0.0736,
+      "step": 3070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6596689992224816e-05,
+      "loss": 0.0509,
+      "step": 3075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6591136287904032e-05,
+      "loss": 0.0481,
+      "step": 3080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.658558258358325e-05,
+      "loss": 0.1091,
+      "step": 3085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.658002887926247e-05,
+      "loss": 0.1179,
+      "step": 3090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6574475174941686e-05,
+      "loss": 0.137,
+      "step": 3095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6568921470620905e-05,
+      "loss": 0.2706,
+      "step": 3100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6563367766300124e-05,
+      "loss": 0.059,
+      "step": 3105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6557814061979343e-05,
+      "loss": 0.0577,
+      "step": 3110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6552260357658562e-05,
+      "loss": 0.0918,
+      "step": 3115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6546706653337778e-05,
+      "loss": 0.0781,
+      "step": 3120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6541152949016997e-05,
+      "loss": 0.0477,
+      "step": 3125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.6535599244696213e-05,
+      "loss": 0.0428,
+      "step": 3130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.6530045540375432e-05,
+      "loss": 0.0577,
+      "step": 3135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.652449183605465e-05,
+      "loss": 0.0899,
+      "step": 3140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.6518938131733867e-05,
+      "loss": 0.125,
+      "step": 3145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.6513384427413086e-05,
+      "loss": 0.2548,
+      "step": 3150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.6507830723092305e-05,
+      "loss": 0.0755,
+      "step": 3155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.650227701877152e-05,
+      "loss": 0.0682,
+      "step": 3160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.649672331445074e-05,
+      "loss": 0.0739,
+      "step": 3165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.649116961012996e-05,
+      "loss": 0.0652,
+      "step": 3170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.6485615905809175e-05,
+      "loss": 0.0466,
+      "step": 3175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.6480062201488394e-05,
+      "loss": 0.0252,
+      "step": 3180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.6474508497167613e-05,
+      "loss": 0.0464,
+      "step": 3185
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.646895479284683e-05,
+      "loss": 0.0912,
+      "step": 3190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.6463401088526048e-05,
+      "loss": 0.1133,
+      "step": 3195
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.6457847384205267e-05,
+      "loss": 0.2558,
+      "step": 3200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.6452293679884482e-05,
+      "loss": 0.0589,
+      "step": 3205
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.64467399755637e-05,
+      "loss": 0.0601,
+      "step": 3210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.644118627124292e-05,
+      "loss": 0.0505,
+      "step": 3215
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.643563256692214e-05,
+      "loss": 0.0602,
+      "step": 3220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.643007886260136e-05,
+      "loss": 0.0408,
+      "step": 3225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.6424525158280575e-05,
+      "loss": 0.0463,
+      "step": 3230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.6418971453959794e-05,
+      "loss": 0.0633,
+      "step": 3235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.6413417749639013e-05,
+      "loss": 0.0986,
+      "step": 3240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.640786404531823e-05,
+      "loss": 0.169,
+      "step": 3245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.6402310340997448e-05,
+      "loss": 0.2716,
+      "step": 3250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.6396756636676663e-05,
+      "loss": 0.0464,
+      "step": 3255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.6391202932355883e-05,
+      "loss": 0.0606,
+      "step": 3260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.63856492280351e-05,
+      "loss": 0.0612,
+      "step": 3265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6380095523714317e-05,
+      "loss": 0.0781,
+      "step": 3270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6374541819393536e-05,
+      "loss": 0.0595,
+      "step": 3275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6368988115072756e-05,
+      "loss": 0.0541,
+      "step": 3280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.636343441075197e-05,
+      "loss": 0.0477,
+      "step": 3285
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.635788070643119e-05,
+      "loss": 0.0782,
+      "step": 3290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.635232700211041e-05,
+      "loss": 0.0875,
+      "step": 3295
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6346773297789625e-05,
+      "loss": 0.2821,
+      "step": 3300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6341219593468844e-05,
+      "loss": 0.0721,
+      "step": 3305
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6335665889148063e-05,
+      "loss": 0.0631,
+      "step": 3310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.633011218482728e-05,
+      "loss": 0.0766,
+      "step": 3315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6324558480506498e-05,
+      "loss": 0.0792,
+      "step": 3320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6319004776185717e-05,
+      "loss": 0.0714,
+      "step": 3325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6313451071864933e-05,
+      "loss": 0.0242,
+      "step": 3330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6307897367544156e-05,
+      "loss": 0.0505,
+      "step": 3335
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.630234366322337e-05,
+      "loss": 0.0919,
+      "step": 3340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.629678995890259e-05,
+      "loss": 0.1376,
+      "step": 3345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.629123625458181e-05,
+      "loss": 0.2847,
+      "step": 3350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6285682550261025e-05,
+      "loss": 0.0671,
+      "step": 3355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6280128845940244e-05,
+      "loss": 0.0627,
+      "step": 3360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6274575141619463e-05,
+      "loss": 0.0757,
+      "step": 3365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.626902143729868e-05,
+      "loss": 0.0665,
+      "step": 3370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.62634677329779e-05,
+      "loss": 0.0607,
+      "step": 3375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6257914028657117e-05,
+      "loss": 0.0382,
+      "step": 3380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6252360324336333e-05,
+      "loss": 0.0726,
+      "step": 3385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6246806620015552e-05,
+      "loss": 0.0613,
+      "step": 3390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6241252915694768e-05,
+      "loss": 0.1511,
+      "step": 3395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6235699211373987e-05,
+      "loss": 0.2754,
+      "step": 3400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6230145507053206e-05,
+      "loss": 0.0749,
+      "step": 3405
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6224591802732422e-05,
+      "loss": 0.0522,
+      "step": 3410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.621903809841164e-05,
+      "loss": 0.0615,
+      "step": 3415
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.621348439409086e-05,
+      "loss": 0.0684,
+      "step": 3420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6207930689770076e-05,
+      "loss": 0.0597,
+      "step": 3425
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6202376985449295e-05,
+      "loss": 0.0613,
+      "step": 3430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6196823281128514e-05,
+      "loss": 0.0498,
+      "step": 3435
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.619126957680773e-05,
+      "loss": 0.1044,
+      "step": 3440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6185715872486952e-05,
+      "loss": 0.1196,
+      "step": 3445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6180162168166168e-05,
+      "loss": 0.2592,
+      "step": 3450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6174608463845387e-05,
+      "loss": 0.0569,
+      "step": 3455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6169054759524606e-05,
+      "loss": 0.0635,
+      "step": 3460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6163501055203822e-05,
+      "loss": 0.0751,
+      "step": 3465
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.615794735088304e-05,
+      "loss": 0.0712,
+      "step": 3470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.615239364656226e-05,
+      "loss": 0.0576,
+      "step": 3475
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6146839942241476e-05,
+      "loss": 0.0494,
+      "step": 3480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6141286237920695e-05,
+      "loss": 0.0426,
+      "step": 3485
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6135732533599914e-05,
+      "loss": 0.0876,
+      "step": 3490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.613017882927913e-05,
+      "loss": 0.1097,
+      "step": 3495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.612462512495835e-05,
+      "loss": 0.2556,
+      "step": 3500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6119071420637568e-05,
+      "loss": 0.0587,
+      "step": 3505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6113517716316784e-05,
+      "loss": 0.0571,
+      "step": 3510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6107964011996003e-05,
+      "loss": 0.0689,
+      "step": 3515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.610241030767522e-05,
+      "loss": 0.087,
+      "step": 3520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6096856603354438e-05,
+      "loss": 0.0577,
+      "step": 3525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6091302899033657e-05,
+      "loss": 0.0368,
+      "step": 3530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6085749194712873e-05,
+      "loss": 0.0706,
+      "step": 3535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6080195490392092e-05,
+      "loss": 0.0958,
+      "step": 3540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.607464178607131e-05,
+      "loss": 0.1481,
+      "step": 3545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6069088081750527e-05,
+      "loss": 0.2628,
+      "step": 3550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6063534377429746e-05,
+      "loss": 0.0594,
+      "step": 3555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6057980673108965e-05,
+      "loss": 0.0708,
+      "step": 3560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6052426968788184e-05,
+      "loss": 0.0699,
+      "step": 3565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6046873264467403e-05,
+      "loss": 0.079,
+      "step": 3570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.604131956014662e-05,
+      "loss": 0.0646,
+      "step": 3575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6035765855825838e-05,
+      "loss": 0.0237,
+      "step": 3580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6030212151505057e-05,
+      "loss": 0.0569,
+      "step": 3585
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6024658447184273e-05,
+      "loss": 0.0804,
+      "step": 3590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6019104742863492e-05,
+      "loss": 0.1337,
+      "step": 3595
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.601355103854271e-05,
+      "loss": 0.2835,
+      "step": 3600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6007997334221927e-05,
+      "loss": 0.0517,
+      "step": 3605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6002443629901146e-05,
+      "loss": 0.0752,
+      "step": 3610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.5996889925580365e-05,
+      "loss": 0.0774,
+      "step": 3615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.599133622125958e-05,
+      "loss": 0.0782,
+      "step": 3620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.59857825169388e-05,
+      "loss": 0.0725,
+      "step": 3625
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.598022881261802e-05,
+      "loss": 0.0323,
+      "step": 3630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.5974675108297234e-05,
+      "loss": 0.0379,
+      "step": 3635
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.5969121403976454e-05,
+      "loss": 0.0987,
+      "step": 3640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.596356769965567e-05,
+      "loss": 0.1611,
+      "step": 3645
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.595801399533489e-05,
+      "loss": 0.2173,
+      "step": 3650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.5952460291014108e-05,
+      "loss": 0.06,
+      "step": 3655
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.5946906586693323e-05,
+      "loss": 0.0687,
+      "step": 3660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.5941352882372542e-05,
+      "loss": 0.0605,
+      "step": 3665
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.593579917805176e-05,
+      "loss": 0.0795,
+      "step": 3670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.593024547373098e-05,
+      "loss": 0.055,
+      "step": 3675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.59246917694102e-05,
+      "loss": 0.0471,
+      "step": 3680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5919138065089415e-05,
+      "loss": 0.0987,
+      "step": 3685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5913584360768635e-05,
+      "loss": 0.0836,
+      "step": 3690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5908030656447854e-05,
+      "loss": 0.1381,
+      "step": 3695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.590247695212707e-05,
+      "loss": 0.1834,
+      "step": 3700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.589692324780629e-05,
+      "loss": 0.056,
+      "step": 3705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5891369543485508e-05,
+      "loss": 0.0702,
+      "step": 3710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5885815839164723e-05,
+      "loss": 0.0847,
+      "step": 3715
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5880262134843942e-05,
+      "loss": 0.0811,
+      "step": 3720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.587470843052316e-05,
+      "loss": 0.0428,
+      "step": 3725
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5869154726202377e-05,
+      "loss": 0.0389,
+      "step": 3730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5863601021881596e-05,
+      "loss": 0.034,
+      "step": 3735
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5858047317560815e-05,
+      "loss": 0.0998,
+      "step": 3740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.585249361324003e-05,
+      "loss": 0.1014,
+      "step": 3745
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.584693990891925e-05,
+      "loss": 0.3207,
+      "step": 3750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.584138620459847e-05,
+      "loss": 0.0583,
+      "step": 3755
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5835832500277685e-05,
+      "loss": 0.0473,
+      "step": 3760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5830278795956904e-05,
+      "loss": 0.0672,
+      "step": 3765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5824725091636123e-05,
+      "loss": 0.0781,
+      "step": 3770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.581917138731534e-05,
+      "loss": 0.05,
+      "step": 3775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5813617682994558e-05,
+      "loss": 0.0915,
+      "step": 3780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5808063978673777e-05,
+      "loss": 0.0452,
+      "step": 3785
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5802510274352996e-05,
+      "loss": 0.1039,
+      "step": 3790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5796956570032212e-05,
+      "loss": 0.1082,
+      "step": 3795
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.579140286571143e-05,
+      "loss": 0.2446,
+      "step": 3800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.578584916139065e-05,
+      "loss": 0.0746,
+      "step": 3805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5780295457069866e-05,
+      "loss": 0.0597,
+      "step": 3810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5774741752749085e-05,
+      "loss": 0.0712,
+      "step": 3815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5769188048428304e-05,
+      "loss": 0.0817,
+      "step": 3820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.576363434410752e-05,
+      "loss": 0.0501,
+      "step": 3825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.575808063978674e-05,
+      "loss": 0.0435,
+      "step": 3830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5752526935465958e-05,
+      "loss": 0.0788,
+      "step": 3835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5746973231145174e-05,
+      "loss": 0.1014,
+      "step": 3840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5741419526824393e-05,
+      "loss": 0.1645,
+      "step": 3845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5735865822503612e-05,
+      "loss": 0.2483,
+      "step": 3850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5730312118182828e-05,
+      "loss": 0.0494,
+      "step": 3855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5724758413862047e-05,
+      "loss": 0.0643,
+      "step": 3860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5719204709541266e-05,
+      "loss": 0.0578,
+      "step": 3865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5713651005220482e-05,
+      "loss": 0.0807,
+      "step": 3870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.57080973008997e-05,
+      "loss": 0.0529,
+      "step": 3875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.570254359657892e-05,
+      "loss": 0.0336,
+      "step": 3880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5696989892258136e-05,
+      "loss": 0.0602,
+      "step": 3885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5691436187937355e-05,
+      "loss": 0.0739,
+      "step": 3890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5685882483616574e-05,
+      "loss": 0.1081,
+      "step": 3895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5680328779295793e-05,
+      "loss": 0.2437,
+      "step": 3900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.567477507497501e-05,
+      "loss": 0.0522,
+      "step": 3905
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5669221370654228e-05,
+      "loss": 0.06,
+      "step": 3910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5663667666333447e-05,
+      "loss": 0.0567,
+      "step": 3915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5658113962012663e-05,
+      "loss": 0.0744,
+      "step": 3920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5652560257691882e-05,
+      "loss": 0.0553,
+      "step": 3925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.56470065533711e-05,
+      "loss": 0.0255,
+      "step": 3930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5641452849050317e-05,
+      "loss": 0.0635,
+      "step": 3935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5635899144729536e-05,
+      "loss": 0.0678,
+      "step": 3940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5630345440408755e-05,
+      "loss": 0.1481,
+      "step": 3945
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.562479173608797e-05,
+      "loss": 0.2179,
+      "step": 3950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.561923803176719e-05,
+      "loss": 0.0837,
+      "step": 3955
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.561368432744641e-05,
+      "loss": 0.0608,
+      "step": 3960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5608130623125625e-05,
+      "loss": 0.0804,
+      "step": 3965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5602576918804844e-05,
+      "loss": 0.06,
+      "step": 3970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5597023214484063e-05,
+      "loss": 0.0551,
+      "step": 3975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.559146951016328e-05,
+      "loss": 0.0316,
+      "step": 3980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5585915805842498e-05,
+      "loss": 0.0555,
+      "step": 3985
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5580362101521717e-05,
+      "loss": 0.1162,
+      "step": 3990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5574808397200933e-05,
+      "loss": 0.1445,
+      "step": 3995
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.556925469288015e-05,
+      "loss": 0.2618,
+      "step": 4000
+    },
+    {
+      "epoch": 0.89,
+      "eval_loss": 0.02628185786306858,
+      "eval_runtime": 1516.0724,
+      "eval_samples_per_second": 10.565,
+      "eval_steps_per_second": 2.642,
+      "eval_wer": 0.06339704291381176,
+      "step": 4000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.556370098855937e-05,
+      "loss": 0.075,
+      "step": 4005
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.555814728423859e-05,
+      "loss": 0.0611,
+      "step": 4010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.555259357991781e-05,
+      "loss": 0.0646,
+      "step": 4015
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5547039875597025e-05,
+      "loss": 0.0671,
+      "step": 4020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5541486171276244e-05,
+      "loss": 0.0683,
+      "step": 4025
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.553593246695546e-05,
+      "loss": 0.0467,
+      "step": 4030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.553037876263468e-05,
+      "loss": 0.0603,
+      "step": 4035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5524825058313898e-05,
+      "loss": 0.091,
+      "step": 4040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5519271353993113e-05,
+      "loss": 0.1234,
+      "step": 4045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5513717649672333e-05,
+      "loss": 0.3092,
+      "step": 4050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5508163945351552e-05,
+      "loss": 0.0705,
+      "step": 4055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5502610241030767e-05,
+      "loss": 0.0534,
+      "step": 4060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5497056536709987e-05,
+      "loss": 0.0698,
+      "step": 4065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5491502832389206e-05,
+      "loss": 0.0805,
+      "step": 4070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.548594912806842e-05,
+      "loss": 0.0862,
+      "step": 4075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.548039542374764e-05,
+      "loss": 0.0272,
+      "step": 4080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.547484171942686e-05,
+      "loss": 0.0797,
+      "step": 4085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5469288015106075e-05,
+      "loss": 0.0854,
+      "step": 4090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5463734310785294e-05,
+      "loss": 0.1223,
+      "step": 4095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5458180606464514e-05,
+      "loss": 0.2367,
+      "step": 4100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.545262690214373e-05,
+      "loss": 0.0653,
+      "step": 4105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.544707319782295e-05,
+      "loss": 0.0644,
+      "step": 4110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5441519493502167e-05,
+      "loss": 0.0741,
+      "step": 4115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5435965789181387e-05,
+      "loss": 0.0816,
+      "step": 4120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5430412084860606e-05,
+      "loss": 0.0441,
+      "step": 4125
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.542485838053982e-05,
+      "loss": 0.0505,
+      "step": 4130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.541930467621904e-05,
+      "loss": 0.0723,
+      "step": 4135
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.541375097189826e-05,
+      "loss": 0.1136,
+      "step": 4140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5408197267577475e-05,
+      "loss": 0.1288,
+      "step": 4145
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5402643563256694e-05,
+      "loss": 0.2245,
+      "step": 4150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5397089858935914e-05,
+      "loss": 0.048,
+      "step": 4155
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.539153615461513e-05,
+      "loss": 0.0569,
+      "step": 4160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.538598245029435e-05,
+      "loss": 0.0754,
+      "step": 4165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5380428745973564e-05,
+      "loss": 0.0687,
+      "step": 4170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5374875041652783e-05,
+      "loss": 0.0487,
+      "step": 4175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5369321337332002e-05,
+      "loss": 0.0285,
+      "step": 4180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5363767633011218e-05,
+      "loss": 0.0442,
+      "step": 4185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5358213928690437e-05,
+      "loss": 0.079,
+      "step": 4190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5352660224369656e-05,
+      "loss": 0.0937,
+      "step": 4195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5347106520048872e-05,
+      "loss": 0.2476,
+      "step": 4200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.534155281572809e-05,
+      "loss": 0.066,
+      "step": 4205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.533599911140731e-05,
+      "loss": 0.0694,
+      "step": 4210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5330445407086526e-05,
+      "loss": 0.0593,
+      "step": 4215
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5324891702765745e-05,
+      "loss": 0.0716,
+      "step": 4220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5319337998444964e-05,
+      "loss": 0.0648,
+      "step": 4225
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.531378429412418e-05,
+      "loss": 0.0386,
+      "step": 4230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5308230589803402e-05,
+      "loss": 0.0627,
+      "step": 4235
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5302676885482618e-05,
+      "loss": 0.0974,
+      "step": 4240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5297123181161837e-05,
+      "loss": 0.1132,
+      "step": 4245
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5291569476841056e-05,
+      "loss": 0.2465,
+      "step": 4250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5286015772520272e-05,
+      "loss": 0.0451,
+      "step": 4255
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.528046206819949e-05,
+      "loss": 0.0612,
+      "step": 4260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.527490836387871e-05,
+      "loss": 0.0671,
+      "step": 4265
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5269354659557926e-05,
+      "loss": 0.0863,
+      "step": 4270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5263800955237145e-05,
+      "loss": 0.0755,
+      "step": 4275
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5258247250916363e-05,
+      "loss": 0.0419,
+      "step": 4280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.525269354659558e-05,
+      "loss": 0.0571,
+      "step": 4285
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5247139842274799e-05,
+      "loss": 0.0785,
+      "step": 4290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5241586137954016e-05,
+      "loss": 0.1475,
+      "step": 4295
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5236032433633234e-05,
+      "loss": 0.2645,
+      "step": 4300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5230478729312453e-05,
+      "loss": 0.0792,
+      "step": 4305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.522492502499167e-05,
+      "loss": 0.0676,
+      "step": 4310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5219371320670888e-05,
+      "loss": 0.0655,
+      "step": 4315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5213817616350107e-05,
+      "loss": 0.0657,
+      "step": 4320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5208263912029324e-05,
+      "loss": 0.0507,
+      "step": 4325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5202710207708542e-05,
+      "loss": 0.0484,
+      "step": 4330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.519715650338776e-05,
+      "loss": 0.0565,
+      "step": 4335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5191602799066978e-05,
+      "loss": 0.0963,
+      "step": 4340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5186049094746197e-05,
+      "loss": 0.1451,
+      "step": 4345
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5180495390425417e-05,
+      "loss": 0.2916,
+      "step": 4350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5174941686104634e-05,
+      "loss": 0.0605,
+      "step": 4355
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5169387981783851e-05,
+      "loss": 0.0559,
+      "step": 4360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.516383427746307e-05,
+      "loss": 0.0625,
+      "step": 4365
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5158280573142288e-05,
+      "loss": 0.0783,
+      "step": 4370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5152726868821505e-05,
+      "loss": 0.0404,
+      "step": 4375
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5147173164500724e-05,
+      "loss": 0.0495,
+      "step": 4380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5141619460179942e-05,
+      "loss": 0.0431,
+      "step": 4385
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.513606575585916e-05,
+      "loss": 0.0795,
+      "step": 4390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5131622792402534e-05,
+      "loss": 0.1607,
+      "step": 4395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5126069088081751e-05,
+      "loss": 0.272,
+      "step": 4400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.512051538376097e-05,
+      "loss": 0.0484,
+      "step": 4405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5114961679440187e-05,
+      "loss": 0.0539,
+      "step": 4410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5109407975119405e-05,
+      "loss": 0.0769,
+      "step": 4415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5103854270798624e-05,
+      "loss": 0.0838,
+      "step": 4420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5098300566477841e-05,
+      "loss": 0.0391,
+      "step": 4425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5092746862157059e-05,
+      "loss": 0.0769,
+      "step": 4430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5087193157836278e-05,
+      "loss": 0.064,
+      "step": 4435
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5081639453515495e-05,
+      "loss": 0.08,
+      "step": 4440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5076085749194713e-05,
+      "loss": 0.1842,
+      "step": 4445
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5070532044873932e-05,
+      "loss": 0.2785,
+      "step": 4450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.506497834055315e-05,
+      "loss": 0.0656,
+      "step": 4455
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5059424636232368e-05,
+      "loss": 0.0557,
+      "step": 4460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5053870931911588e-05,
+      "loss": 0.0621,
+      "step": 4465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5048317227590805e-05,
+      "loss": 0.0784,
+      "step": 4470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5042763523270022e-05,
+      "loss": 0.0381,
+      "step": 4475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5037209818949241e-05,
+      "loss": 0.0325,
+      "step": 4480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5031656114628459e-05,
+      "loss": 0.0567,
+      "step": 4485
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5026102410307676e-05,
+      "loss": 0.0855,
+      "step": 4490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5020548705986895e-05,
+      "loss": 0.1069,
+      "step": 4495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5014995001666113e-05,
+      "loss": 0.2458,
+      "step": 4500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500944129734533e-05,
+      "loss": 0.1176,
+      "step": 4505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500388759302455e-05,
+      "loss": 0.0481,
+      "step": 4510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4998333888703767e-05,
+      "loss": 0.0561,
+      "step": 4515
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4992780184382984e-05,
+      "loss": 0.0628,
+      "step": 4520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4987226480062203e-05,
+      "loss": 0.0773,
+      "step": 4525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.498167277574142e-05,
+      "loss": 0.0454,
+      "step": 4530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4976119071420638e-05,
+      "loss": 0.0468,
+      "step": 4535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4970565367099856e-05,
+      "loss": 0.0477,
+      "step": 4540
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4965011662779075e-05,
+      "loss": 0.0878,
+      "step": 4545
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4959457958458292e-05,
+      "loss": 0.1237,
+      "step": 4550
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.495390425413751e-05,
+      "loss": 0.2383,
+      "step": 4555
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4948350549816729e-05,
+      "loss": 0.0537,
+      "step": 4560
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4942796845495946e-05,
+      "loss": 0.061,
+      "step": 4565
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4937243141175167e-05,
+      "loss": 0.0656,
+      "step": 4570
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4931689436854384e-05,
+      "loss": 0.0805,
+      "step": 4575
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4926135732533602e-05,
+      "loss": 0.0405,
+      "step": 4580
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.492058202821282e-05,
+      "loss": 0.0304,
+      "step": 4585
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4915028323892038e-05,
+      "loss": 0.0568,
+      "step": 4590
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4909474619571256e-05,
+      "loss": 0.0737,
+      "step": 4595
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4903920915250473e-05,
+      "loss": 0.1461,
+      "step": 4600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4898367210929692e-05,
+      "loss": 0.1985,
+      "step": 4605
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.489281350660891e-05,
+      "loss": 0.0559,
+      "step": 4610
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4887259802288127e-05,
+      "loss": 0.0694,
+      "step": 4615
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4881706097967346e-05,
+      "loss": 0.0697,
+      "step": 4620
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4876152393646564e-05,
+      "loss": 0.072,
+      "step": 4625
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4870598689325781e-05,
+      "loss": 0.0342,
+      "step": 4630
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4865044985005e-05,
+      "loss": 0.0554,
+      "step": 4635
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4859491280684217e-05,
+      "loss": 0.0751,
+      "step": 4640
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4853937576363435e-05,
+      "loss": 0.0851,
+      "step": 4645
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4848383872042654e-05,
+      "loss": 0.1625,
+      "step": 4650
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4842830167721871e-05,
+      "loss": 0.241,
+      "step": 4655
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4837276463401089e-05,
+      "loss": 0.0503,
+      "step": 4660
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4831722759080306e-05,
+      "loss": 0.0554,
+      "step": 4665
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4826169054759525e-05,
+      "loss": 0.0603,
+      "step": 4670
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4820615350438743e-05,
+      "loss": 0.0735,
+      "step": 4675
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.481506164611796e-05,
+      "loss": 0.0417,
+      "step": 4680
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4809507941797181e-05,
+      "loss": 0.0449,
+      "step": 4685
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4803954237476398e-05,
+      "loss": 0.0689,
+      "step": 4690
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4798400533155618e-05,
+      "loss": 0.092,
+      "step": 4695
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4792846828834835e-05,
+      "loss": 0.147,
+      "step": 4700
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4787293124514052e-05,
+      "loss": 0.2515,
+      "step": 4705
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4781739420193271e-05,
+      "loss": 0.0508,
+      "step": 4710
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4776185715872489e-05,
+      "loss": 0.0561,
+      "step": 4715
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4770632011551706e-05,
+      "loss": 0.0715,
+      "step": 4720
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4765078307230924e-05,
+      "loss": 0.0838,
+      "step": 4725
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4759524602910143e-05,
+      "loss": 0.0473,
+      "step": 4730
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.475397089858936e-05,
+      "loss": 0.0424,
+      "step": 4735
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4748417194268578e-05,
+      "loss": 0.0627,
+      "step": 4740
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4742863489947797e-05,
+      "loss": 0.0875,
+      "step": 4745
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4737309785627014e-05,
+      "loss": 0.1237,
+      "step": 4750
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4731756081306232e-05,
+      "loss": 0.1914,
+      "step": 4755
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.472620237698545e-05,
+      "loss": 0.0656,
+      "step": 4760
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4720648672664668e-05,
+      "loss": 0.084,
+      "step": 4765
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4715094968343886e-05,
+      "loss": 0.071,
+      "step": 4770
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4709541264023105e-05,
+      "loss": 0.072,
+      "step": 4775
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4703987559702322e-05,
+      "loss": 0.0505,
+      "step": 4780
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.469843385538154e-05,
+      "loss": 0.0653,
+      "step": 4785
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4692880151060757e-05,
+      "loss": 0.0866,
+      "step": 4790
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4687326446739978e-05,
+      "loss": 0.0888,
+      "step": 4795
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4681772742419195e-05,
+      "loss": 0.1337,
+      "step": 4800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4676219038098414e-05,
+      "loss": 0.237,
+      "step": 4805
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4670665333777632e-05,
+      "loss": 0.0537,
+      "step": 4810
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4665111629456849e-05,
+      "loss": 0.0461,
+      "step": 4815
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4659557925136068e-05,
+      "loss": 0.0756,
+      "step": 4820
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4654004220815286e-05,
+      "loss": 0.0743,
+      "step": 4825
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4648450516494503e-05,
+      "loss": 0.041,
+      "step": 4830
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4642896812173722e-05,
+      "loss": 0.0518,
+      "step": 4835
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.463734310785294e-05,
+      "loss": 0.0612,
+      "step": 4840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4631789403532157e-05,
+      "loss": 0.0879,
+      "step": 4845
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4626235699211374e-05,
+      "loss": 0.1324,
+      "step": 4850
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4620681994890593e-05,
+      "loss": 0.2705,
+      "step": 4855
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4615128290569811e-05,
+      "loss": 0.0515,
+      "step": 4860
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4609574586249028e-05,
+      "loss": 0.045,
+      "step": 4865
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4604020881928247e-05,
+      "loss": 0.0674,
+      "step": 4870
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4598467177607465e-05,
+      "loss": 0.0663,
+      "step": 4875
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4592913473286682e-05,
+      "loss": 0.0595,
+      "step": 4880
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4587359768965901e-05,
+      "loss": 0.065,
+      "step": 4885
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4581806064645119e-05,
+      "loss": 0.0522,
+      "step": 4890
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4576252360324336e-05,
+      "loss": 0.071,
+      "step": 4895
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4570698656003555e-05,
+      "loss": 0.1409,
+      "step": 4900
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4565144951682773e-05,
+      "loss": 0.258,
+      "step": 4905
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4559591247361994e-05,
+      "loss": 0.0533,
+      "step": 4910
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4554037543041211e-05,
+      "loss": 0.0512,
+      "step": 4915
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4548483838720428e-05,
+      "loss": 0.0652,
+      "step": 4920
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4542930134399646e-05,
+      "loss": 0.0724,
+      "step": 4925
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4537376430078865e-05,
+      "loss": 0.0417,
+      "step": 4930
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4531822725758082e-05,
+      "loss": 0.0413,
+      "step": 4935
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.45262690214373e-05,
+      "loss": 0.0652,
+      "step": 4940
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4520715317116519e-05,
+      "loss": 0.0827,
+      "step": 4945
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4515161612795736e-05,
+      "loss": 0.1547,
+      "step": 4950
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4509607908474954e-05,
+      "loss": 0.2925,
+      "step": 4955
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4504054204154173e-05,
+      "loss": 0.05,
+      "step": 4960
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.449850049983339e-05,
+      "loss": 0.0655,
+      "step": 4965
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4492946795512608e-05,
+      "loss": 0.0726,
+      "step": 4970
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4487393091191827e-05,
+      "loss": 0.0921,
+      "step": 4975
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4481839386871044e-05,
+      "loss": 0.0373,
+      "step": 4980
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4476285682550262e-05,
+      "loss": 0.0472,
+      "step": 4985
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4470731978229479e-05,
+      "loss": 0.0557,
+      "step": 4990
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4465178273908698e-05,
+      "loss": 0.0792,
+      "step": 4995
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4459624569587915e-05,
+      "loss": 0.1228,
+      "step": 5000
+    },
+    {
+      "epoch": 1.11,
+      "eval_loss": 0.025642985478043556,
+      "eval_runtime": 1623.1239,
+      "eval_samples_per_second": 9.869,
+      "eval_steps_per_second": 2.467,
+      "eval_wer": 0.05847457627118644,
+      "step": 5000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4454070865267133e-05,
+      "loss": 0.1829,
+      "step": 5005
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4448517160946352e-05,
+      "loss": 0.0529,
+      "step": 5010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.444296345662557e-05,
+      "loss": 0.0578,
+      "step": 5015
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.443740975230479e-05,
+      "loss": 0.0595,
+      "step": 5020
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4431856047984008e-05,
+      "loss": 0.0758,
+      "step": 5025
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4426302343663225e-05,
+      "loss": 0.0501,
+      "step": 5030
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4420748639342444e-05,
+      "loss": 0.0306,
+      "step": 5035
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4415194935021662e-05,
+      "loss": 0.0634,
+      "step": 5040
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4409641230700879e-05,
+      "loss": 0.0965,
+      "step": 5045
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4404087526380096e-05,
+      "loss": 0.1357,
+      "step": 5050
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4398533822059316e-05,
+      "loss": 0.2046,
+      "step": 5055
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4392980117738533e-05,
+      "loss": 0.0575,
+      "step": 5060
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.438742641341775e-05,
+      "loss": 0.0486,
+      "step": 5065
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.438187270909697e-05,
+      "loss": 0.0591,
+      "step": 5070
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4376319004776187e-05,
+      "loss": 0.0884,
+      "step": 5075
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4370765300455404e-05,
+      "loss": 0.0497,
+      "step": 5080
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4365211596134623e-05,
+      "loss": 0.0501,
+      "step": 5085
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4359657891813841e-05,
+      "loss": 0.0595,
+      "step": 5090
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4354104187493058e-05,
+      "loss": 0.0982,
+      "step": 5095
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4348550483172277e-05,
+      "loss": 0.1428,
+      "step": 5100
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4342996778851495e-05,
+      "loss": 0.1831,
+      "step": 5105
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.4337443074530712e-05,
+      "loss": 0.0531,
+      "step": 5110
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.433188937020993e-05,
+      "loss": 0.0627,
+      "step": 5115
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4326335665889149e-05,
+      "loss": 0.0787,
+      "step": 5120
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4320781961568366e-05,
+      "loss": 0.0766,
+      "step": 5125
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4315228257247584e-05,
+      "loss": 0.036,
+      "step": 5130
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4309674552926804e-05,
+      "loss": 0.0342,
+      "step": 5135
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4304120848606022e-05,
+      "loss": 0.0441,
+      "step": 5140
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4298567144285241e-05,
+      "loss": 0.0834,
+      "step": 5145
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4293013439964458e-05,
+      "loss": 0.1365,
+      "step": 5150
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4287459735643676e-05,
+      "loss": 0.2148,
+      "step": 5155
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4281906031322895e-05,
+      "loss": 0.0561,
+      "step": 5160
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4276352327002112e-05,
+      "loss": 0.0556,
+      "step": 5165
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.427079862268133e-05,
+      "loss": 0.0722,
+      "step": 5170
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4265244918360547e-05,
+      "loss": 0.0796,
+      "step": 5175
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4259691214039766e-05,
+      "loss": 0.0415,
+      "step": 5180
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4254137509718984e-05,
+      "loss": 0.0411,
+      "step": 5185
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4248583805398201e-05,
+      "loss": 0.0701,
+      "step": 5190
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.424303010107742e-05,
+      "loss": 0.0766,
+      "step": 5195
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4237476396756638e-05,
+      "loss": 0.1355,
+      "step": 5200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4231922692435855e-05,
+      "loss": 0.2661,
+      "step": 5205
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4226368988115074e-05,
+      "loss": 0.0618,
+      "step": 5210
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4220815283794292e-05,
+      "loss": 0.0588,
+      "step": 5215
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4215261579473509e-05,
+      "loss": 0.0718,
+      "step": 5220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4209707875152728e-05,
+      "loss": 0.0711,
+      "step": 5225
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4204154170831945e-05,
+      "loss": 0.0412,
+      "step": 5230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4198600466511163e-05,
+      "loss": 0.0306,
+      "step": 5235
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.419304676219038e-05,
+      "loss": 0.0682,
+      "step": 5240
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4187493057869601e-05,
+      "loss": 0.0834,
+      "step": 5245
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4181939353548819e-05,
+      "loss": 0.162,
+      "step": 5250
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4176385649228038e-05,
+      "loss": 0.1732,
+      "step": 5255
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4170831944907255e-05,
+      "loss": 0.0465,
+      "step": 5260
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4165278240586472e-05,
+      "loss": 0.0454,
+      "step": 5265
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4159724536265692e-05,
+      "loss": 0.0548,
+      "step": 5270
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4154170831944909e-05,
+      "loss": 0.0754,
+      "step": 5275
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4148617127624126e-05,
+      "loss": 0.0277,
+      "step": 5280
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4143063423303346e-05,
+      "loss": 0.0464,
+      "step": 5285
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4137509718982563e-05,
+      "loss": 0.0484,
+      "step": 5290
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.413195601466178e-05,
+      "loss": 0.0785,
+      "step": 5295
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4126402310341e-05,
+      "loss": 0.1549,
+      "step": 5300
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4120848606020217e-05,
+      "loss": 0.211,
+      "step": 5305
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4115294901699434e-05,
+      "loss": 0.0674,
+      "step": 5310
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4109741197378652e-05,
+      "loss": 0.0547,
+      "step": 5315
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.410418749305787e-05,
+      "loss": 0.0717,
+      "step": 5320
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4098633788737088e-05,
+      "loss": 0.0609,
+      "step": 5325
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4093080084416306e-05,
+      "loss": 0.0531,
+      "step": 5330
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4087526380095525e-05,
+      "loss": 0.0577,
+      "step": 5335
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4081972675774742e-05,
+      "loss": 0.0668,
+      "step": 5340
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.407641897145396e-05,
+      "loss": 0.0744,
+      "step": 5345
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4070865267133179e-05,
+      "loss": 0.1749,
+      "step": 5350
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4065311562812396e-05,
+      "loss": 0.2388,
+      "step": 5355
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4059757858491617e-05,
+      "loss": 0.0661,
+      "step": 5360
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4054204154170834e-05,
+      "loss": 0.0487,
+      "step": 5365
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4048650449850052e-05,
+      "loss": 0.063,
+      "step": 5370
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.404309674552927e-05,
+      "loss": 0.0858,
+      "step": 5375
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4037543041208488e-05,
+      "loss": 0.04,
+      "step": 5380
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.4031989336887706e-05,
+      "loss": 0.0554,
+      "step": 5385
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.4026435632566923e-05,
+      "loss": 0.067,
+      "step": 5390
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.4020881928246142e-05,
+      "loss": 0.0839,
+      "step": 5395
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.401532822392536e-05,
+      "loss": 0.145,
+      "step": 5400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.4009774519604577e-05,
+      "loss": 0.2298,
+      "step": 5405
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.4004220815283796e-05,
+      "loss": 0.0486,
+      "step": 5410
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3998667110963014e-05,
+      "loss": 0.0512,
+      "step": 5415
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3993113406642231e-05,
+      "loss": 0.0598,
+      "step": 5420
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.398755970232145e-05,
+      "loss": 0.0815,
+      "step": 5425
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3982005998000668e-05,
+      "loss": 0.0434,
+      "step": 5430
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3976452293679885e-05,
+      "loss": 0.042,
+      "step": 5435
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3970898589359102e-05,
+      "loss": 0.0529,
+      "step": 5440
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3965344885038321e-05,
+      "loss": 0.1064,
+      "step": 5445
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3959791180717539e-05,
+      "loss": 0.1096,
+      "step": 5450
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3955348217260913e-05,
+      "loss": 0.2322,
+      "step": 5455
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.394979451294013e-05,
+      "loss": 0.0509,
+      "step": 5460
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.394424080861935e-05,
+      "loss": 0.0428,
+      "step": 5465
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3938687104298567e-05,
+      "loss": 0.0545,
+      "step": 5470
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3933133399977788e-05,
+      "loss": 0.0696,
+      "step": 5475
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3927579695657005e-05,
+      "loss": 0.0515,
+      "step": 5480
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3922025991336223e-05,
+      "loss": 0.0323,
+      "step": 5485
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3916472287015442e-05,
+      "loss": 0.0458,
+      "step": 5490
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.391091858269466e-05,
+      "loss": 0.0884,
+      "step": 5495
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3905364878373877e-05,
+      "loss": 0.1321,
+      "step": 5500
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3899811174053096e-05,
+      "loss": 0.2134,
+      "step": 5505
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.3894257469732313e-05,
+      "loss": 0.0635,
+      "step": 5510
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.388870376541153e-05,
+      "loss": 0.0599,
+      "step": 5515
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3883150061090748e-05,
+      "loss": 0.0554,
+      "step": 5520
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3877596356769967e-05,
+      "loss": 0.0782,
+      "step": 5525
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3872042652449185e-05,
+      "loss": 0.0407,
+      "step": 5530
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3866488948128402e-05,
+      "loss": 0.0392,
+      "step": 5535
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3860935243807621e-05,
+      "loss": 0.0689,
+      "step": 5540
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3855381539486839e-05,
+      "loss": 0.0712,
+      "step": 5545
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3849827835166056e-05,
+      "loss": 0.1202,
+      "step": 5550
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3844274130845275e-05,
+      "loss": 0.2686,
+      "step": 5555
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.3838720426524492e-05,
+      "loss": 0.0499,
+      "step": 5560
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.383316672220371e-05,
+      "loss": 0.0577,
+      "step": 5565
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3827613017882929e-05,
+      "loss": 0.0743,
+      "step": 5570
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3822059313562146e-05,
+      "loss": 0.0798,
+      "step": 5575
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3816505609241364e-05,
+      "loss": 0.0499,
+      "step": 5580
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3810951904920585e-05,
+      "loss": 0.0425,
+      "step": 5585
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3805398200599802e-05,
+      "loss": 0.0563,
+      "step": 5590
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.379984449627902e-05,
+      "loss": 0.1166,
+      "step": 5595
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3794290791958239e-05,
+      "loss": 0.1276,
+      "step": 5600
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.3788737087637456e-05,
+      "loss": 0.2692,
+      "step": 5605
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3783183383316673e-05,
+      "loss": 0.0618,
+      "step": 5610
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3777629678995893e-05,
+      "loss": 0.0562,
+      "step": 5615
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.377207597467511e-05,
+      "loss": 0.0561,
+      "step": 5620
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3766522270354327e-05,
+      "loss": 0.0589,
+      "step": 5625
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3760968566033546e-05,
+      "loss": 0.0693,
+      "step": 5630
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3755414861712764e-05,
+      "loss": 0.038,
+      "step": 5635
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3749861157391981e-05,
+      "loss": 0.0571,
+      "step": 5640
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3744307453071199e-05,
+      "loss": 0.0979,
+      "step": 5645
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.3738753748750418e-05,
+      "loss": 0.1353,
+      "step": 5650
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.3733200044429635e-05,
+      "loss": 0.2911,
+      "step": 5655
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.3727646340108853e-05,
+      "loss": 0.0515,
+      "step": 5660
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.3722092635788072e-05,
+      "loss": 0.0397,
+      "step": 5665
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.371653893146729e-05,
+      "loss": 0.0614,
+      "step": 5670
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.3710985227146507e-05,
+      "loss": 0.0709,
+      "step": 5675
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.3705431522825726e-05,
+      "loss": 0.0434,
+      "step": 5680
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.3699877818504943e-05,
+      "loss": 0.0507,
+      "step": 5685
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.369432411418416e-05,
+      "loss": 0.065,
+      "step": 5690
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.368877040986338e-05,
+      "loss": 0.093,
+      "step": 5695
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3683216705542599e-05,
+      "loss": 0.1364,
+      "step": 5700
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3677663001221816e-05,
+      "loss": 0.2214,
+      "step": 5705
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3672109296901035e-05,
+      "loss": 0.0524,
+      "step": 5710
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3666555592580253e-05,
+      "loss": 0.057,
+      "step": 5715
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.366100188825947e-05,
+      "loss": 0.0689,
+      "step": 5720
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.365544818393869e-05,
+      "loss": 0.0724,
+      "step": 5725
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3649894479617907e-05,
+      "loss": 0.0357,
+      "step": 5730
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3644340775297124e-05,
+      "loss": 0.0546,
+      "step": 5735
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.3638787070976343e-05,
+      "loss": 0.0399,
+      "step": 5740
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.363323336665556e-05,
+      "loss": 0.0761,
+      "step": 5745
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3627679662334778e-05,
+      "loss": 0.1236,
+      "step": 5750
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3622125958013997e-05,
+      "loss": 0.2194,
+      "step": 5755
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3616572253693215e-05,
+      "loss": 0.0527,
+      "step": 5760
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3611018549372432e-05,
+      "loss": 0.0511,
+      "step": 5765
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.360546484505165e-05,
+      "loss": 0.0623,
+      "step": 5770
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3599911140730868e-05,
+      "loss": 0.0928,
+      "step": 5775
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3594357436410086e-05,
+      "loss": 0.0341,
+      "step": 5780
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.3588803732089303e-05,
+      "loss": 0.0301,
+      "step": 5785
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3583250027768522e-05,
+      "loss": 0.0598,
+      "step": 5790
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.357769632344774e-05,
+      "loss": 0.0784,
+      "step": 5795
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3572142619126957e-05,
+      "loss": 0.105,
+      "step": 5800
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3566588914806176e-05,
+      "loss": 0.1934,
+      "step": 5805
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3561035210485395e-05,
+      "loss": 0.0476,
+      "step": 5810
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3555481506164615e-05,
+      "loss": 0.0501,
+      "step": 5815
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3549927801843832e-05,
+      "loss": 0.051,
+      "step": 5820
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.354437409752305e-05,
+      "loss": 0.069,
+      "step": 5825
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.3538820393202269e-05,
+      "loss": 0.0514,
+      "step": 5830
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3533266688881486e-05,
+      "loss": 0.0221,
+      "step": 5835
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3527712984560703e-05,
+      "loss": 0.0751,
+      "step": 5840
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.352215928023992e-05,
+      "loss": 0.054,
+      "step": 5845
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.351660557591914e-05,
+      "loss": 0.1389,
+      "step": 5850
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3511051871598357e-05,
+      "loss": 0.2364,
+      "step": 5855
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3505498167277575e-05,
+      "loss": 0.0559,
+      "step": 5860
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3499944462956794e-05,
+      "loss": 0.0585,
+      "step": 5865
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3494390758636011e-05,
+      "loss": 0.0487,
+      "step": 5870
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.3488837054315229e-05,
+      "loss": 0.0511,
+      "step": 5875
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3483283349994448e-05,
+      "loss": 0.0385,
+      "step": 5880
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3477729645673665e-05,
+      "loss": 0.0402,
+      "step": 5885
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3472175941352883e-05,
+      "loss": 0.0548,
+      "step": 5890
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3466622237032102e-05,
+      "loss": 0.0803,
+      "step": 5895
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.346106853271132e-05,
+      "loss": 0.1173,
+      "step": 5900
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3455514828390537e-05,
+      "loss": 0.2381,
+      "step": 5905
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3449961124069754e-05,
+      "loss": 0.0526,
+      "step": 5910
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.3444407419748973e-05,
+      "loss": 0.057,
+      "step": 5915
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.343885371542819e-05,
+      "loss": 0.0642,
+      "step": 5920
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3433300011107411e-05,
+      "loss": 0.0614,
+      "step": 5925
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3427746306786629e-05,
+      "loss": 0.0382,
+      "step": 5930
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3422192602465846e-05,
+      "loss": 0.0547,
+      "step": 5935
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3416638898145065e-05,
+      "loss": 0.0837,
+      "step": 5940
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3411085193824283e-05,
+      "loss": 0.1043,
+      "step": 5945
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.34055314895035e-05,
+      "loss": 0.1402,
+      "step": 5950
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.339997778518272e-05,
+      "loss": 0.2662,
+      "step": 5955
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3394424080861937e-05,
+      "loss": 0.0508,
+      "step": 5960
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.3388870376541154e-05,
+      "loss": 0.0588,
+      "step": 5965
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3383316672220371e-05,
+      "loss": 0.0576,
+      "step": 5970
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.337776296789959e-05,
+      "loss": 0.071,
+      "step": 5975
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3372209263578808e-05,
+      "loss": 0.0376,
+      "step": 5980
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3366655559258025e-05,
+      "loss": 0.0315,
+      "step": 5985
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3361101854937245e-05,
+      "loss": 0.051,
+      "step": 5990
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3355548150616462e-05,
+      "loss": 0.0765,
+      "step": 5995
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.334999444629568e-05,
+      "loss": 0.0941,
+      "step": 6000
+    },
+    {
+      "epoch": 1.33,
+      "eval_loss": 0.030542902648448944,
+      "eval_runtime": 1648.7542,
+      "eval_samples_per_second": 9.715,
+      "eval_steps_per_second": 2.429,
+      "eval_wer": 0.05883519653804544,
+      "step": 6000
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3344440741974898e-05,
+      "loss": 0.2763,
+      "step": 6005
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.3338887037654116e-05,
+      "loss": 0.0546,
+      "step": 6010
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.0576,
+      "step": 6015
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3327779629012552e-05,
+      "loss": 0.0585,
+      "step": 6020
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.332222592469177e-05,
+      "loss": 0.0754,
+      "step": 6025
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3316672220370987e-05,
+      "loss": 0.0387,
+      "step": 6030
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3311118516050208e-05,
+      "loss": 0.0752,
+      "step": 6035
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3305564811729425e-05,
+      "loss": 0.0611,
+      "step": 6040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3300011107408643e-05,
+      "loss": 0.0682,
+      "step": 6045
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.3294457403087862e-05,
+      "loss": 0.1438,
+      "step": 6050
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.328890369876708e-05,
+      "loss": 0.2599,
+      "step": 6055
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3283349994446297e-05,
+      "loss": 0.0516,
+      "step": 6060
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3277796290125516e-05,
+      "loss": 0.0507,
+      "step": 6065
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3272242585804733e-05,
+      "loss": 0.0508,
+      "step": 6070
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.326668888148395e-05,
+      "loss": 0.056,
+      "step": 6075
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.326113517716317e-05,
+      "loss": 0.0744,
+      "step": 6080
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3255581472842387e-05,
+      "loss": 0.0335,
+      "step": 6085
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3250027768521605e-05,
+      "loss": 0.058,
+      "step": 6090
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3244474064200822e-05,
+      "loss": 0.0825,
+      "step": 6095
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.3238920359880041e-05,
+      "loss": 0.1488,
+      "step": 6100
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3233366655559259e-05,
+      "loss": 0.2172,
+      "step": 6105
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3227812951238476e-05,
+      "loss": 0.048,
+      "step": 6110
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3222259246917695e-05,
+      "loss": 0.0531,
+      "step": 6115
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3216705542596913e-05,
+      "loss": 0.0591,
+      "step": 6120
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.321115183827613e-05,
+      "loss": 0.0553,
+      "step": 6125
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3205598133955349e-05,
+      "loss": 0.0437,
+      "step": 6130
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3200044429634567e-05,
+      "loss": 0.0289,
+      "step": 6135
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3194490725313784e-05,
+      "loss": 0.0689,
+      "step": 6140
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.3188937020993003e-05,
+      "loss": 0.0661,
+      "step": 6145
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3183383316672222e-05,
+      "loss": 0.1261,
+      "step": 6150
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.317782961235144e-05,
+      "loss": 0.2211,
+      "step": 6155
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3172275908030659e-05,
+      "loss": 0.0593,
+      "step": 6160
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3166722203709876e-05,
+      "loss": 0.0571,
+      "step": 6165
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3161168499389094e-05,
+      "loss": 0.0496,
+      "step": 6170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3155614795068313e-05,
+      "loss": 0.0684,
+      "step": 6175
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.315006109074753e-05,
+      "loss": 0.0417,
+      "step": 6180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3144507386426747e-05,
+      "loss": 0.046,
+      "step": 6185
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.3138953682105967e-05,
+      "loss": 0.0337,
+      "step": 6190
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3133399977785184e-05,
+      "loss": 0.0794,
+      "step": 6195
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3127846273464401e-05,
+      "loss": 0.1318,
+      "step": 6200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.312229256914362e-05,
+      "loss": 0.2393,
+      "step": 6205
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3116738864822838e-05,
+      "loss": 0.0457,
+      "step": 6210
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3111185160502055e-05,
+      "loss": 0.059,
+      "step": 6215
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3105631456181274e-05,
+      "loss": 0.0673,
+      "step": 6220
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3100077751860492e-05,
+      "loss": 0.0694,
+      "step": 6225
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.309452404753971e-05,
+      "loss": 0.0415,
+      "step": 6230
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3088970343218927e-05,
+      "loss": 0.0356,
+      "step": 6235
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3083416638898146e-05,
+      "loss": 0.0516,
+      "step": 6240
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3077862934577363e-05,
+      "loss": 0.0789,
+      "step": 6245
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.307230923025658e-05,
+      "loss": 0.1,
+      "step": 6250
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.30667555259358e-05,
+      "loss": 0.1971,
+      "step": 6255
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3061201821615019e-05,
+      "loss": 0.0445,
+      "step": 6260
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3055648117294238e-05,
+      "loss": 0.0516,
+      "step": 6265
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3050094412973455e-05,
+      "loss": 0.0603,
+      "step": 6270
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3044540708652673e-05,
+      "loss": 0.08,
+      "step": 6275
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.3038987004331892e-05,
+      "loss": 0.0526,
+      "step": 6280
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.303343330001111e-05,
+      "loss": 0.0418,
+      "step": 6285
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.3027879595690327e-05,
+      "loss": 0.069,
+      "step": 6290
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.3022325891369544e-05,
+      "loss": 0.0599,
+      "step": 6295
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.3016772187048763e-05,
+      "loss": 0.1168,
+      "step": 6300
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.301121848272798e-05,
+      "loss": 0.2823,
+      "step": 6305
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.3005664778407198e-05,
+      "loss": 0.0544,
+      "step": 6310
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.3000111074086417e-05,
+      "loss": 0.0629,
+      "step": 6315
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.2994557369765635e-05,
+      "loss": 0.0759,
+      "step": 6320
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.2989003665444852e-05,
+      "loss": 0.0727,
+      "step": 6325
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2983449961124071e-05,
+      "loss": 0.0166,
+      "step": 6330
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2977896256803289e-05,
+      "loss": 0.0548,
+      "step": 6335
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2972342552482506e-05,
+      "loss": 0.0705,
+      "step": 6340
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2966788848161725e-05,
+      "loss": 0.072,
+      "step": 6345
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2961235143840943e-05,
+      "loss": 0.1583,
+      "step": 6350
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.295568143952016e-05,
+      "loss": 0.23,
+      "step": 6355
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2950127735199377e-05,
+      "loss": 0.0518,
+      "step": 6360
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2944574030878597e-05,
+      "loss": 0.0393,
+      "step": 6365
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.2939020326557814e-05,
+      "loss": 0.0608,
+      "step": 6370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2933466622237035e-05,
+      "loss": 0.0886,
+      "step": 6375
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2927912917916252e-05,
+      "loss": 0.0577,
+      "step": 6380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.292235921359547e-05,
+      "loss": 0.0472,
+      "step": 6385
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2916805509274689e-05,
+      "loss": 0.0695,
+      "step": 6390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2911251804953906e-05,
+      "loss": 0.0634,
+      "step": 6395
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2905698100633124e-05,
+      "loss": 0.1555,
+      "step": 6400
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2900144396312343e-05,
+      "loss": 0.2493,
+      "step": 6405
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.289459069199156e-05,
+      "loss": 0.061,
+      "step": 6410
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.2889036987670777e-05,
+      "loss": 0.0531,
+      "step": 6415
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2883483283349995e-05,
+      "loss": 0.0687,
+      "step": 6420
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2877929579029214e-05,
+      "loss": 0.0809,
+      "step": 6425
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2872375874708431e-05,
+      "loss": 0.0413,
+      "step": 6430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2866822170387649e-05,
+      "loss": 0.0552,
+      "step": 6435
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2861268466066868e-05,
+      "loss": 0.0368,
+      "step": 6440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2855714761746085e-05,
+      "loss": 0.072,
+      "step": 6445
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2850161057425303e-05,
+      "loss": 0.1336,
+      "step": 6450
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.2844607353104522e-05,
+      "loss": 0.1995,
+      "step": 6455
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.283905364878374e-05,
+      "loss": 0.0497,
+      "step": 6460
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2833499944462957e-05,
+      "loss": 0.0526,
+      "step": 6465
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2827946240142176e-05,
+      "loss": 0.0511,
+      "step": 6470
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2822392535821393e-05,
+      "loss": 0.0682,
+      "step": 6475
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.281683883150061e-05,
+      "loss": 0.0339,
+      "step": 6480
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2811285127179831e-05,
+      "loss": 0.034,
+      "step": 6485
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2805731422859049e-05,
+      "loss": 0.0787,
+      "step": 6490
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2800177718538266e-05,
+      "loss": 0.0714,
+      "step": 6495
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2794624014217485e-05,
+      "loss": 0.1622,
+      "step": 6500
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.2789070309896703e-05,
+      "loss": 0.2268,
+      "step": 6505
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.278351660557592e-05,
+      "loss": 0.0522,
+      "step": 6510
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.277796290125514e-05,
+      "loss": 0.0609,
+      "step": 6515
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.2772409196934357e-05,
+      "loss": 0.0565,
+      "step": 6520
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.2766855492613574e-05,
+      "loss": 0.06,
+      "step": 6525
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.2761301788292793e-05,
+      "loss": 0.0442,
+      "step": 6530
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.275574808397201e-05,
+      "loss": 0.0316,
+      "step": 6535
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.2750194379651228e-05,
+      "loss": 0.0788,
+      "step": 6540
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.2744640675330446e-05,
+      "loss": 0.0731,
+      "step": 6545
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.2739086971009665e-05,
+      "loss": 0.1229,
+      "step": 6550
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2733533266688882e-05,
+      "loss": 0.2654,
+      "step": 6555
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.27279795623681e-05,
+      "loss": 0.0448,
+      "step": 6560
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2722425858047319e-05,
+      "loss": 0.0403,
+      "step": 6565
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2716872153726536e-05,
+      "loss": 0.0539,
+      "step": 6570
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2711318449405753e-05,
+      "loss": 0.071,
+      "step": 6575
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2705764745084973e-05,
+      "loss": 0.0368,
+      "step": 6580
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.270021104076419e-05,
+      "loss": 0.0374,
+      "step": 6585
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2694657336443407e-05,
+      "loss": 0.0525,
+      "step": 6590
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2689103632122626e-05,
+      "loss": 0.097,
+      "step": 6595
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2683549927801846e-05,
+      "loss": 0.1366,
+      "step": 6600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2677996223481065e-05,
+      "loss": 0.2219,
+      "step": 6605
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2672442519160282e-05,
+      "loss": 0.044,
+      "step": 6610
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.26668888148395e-05,
+      "loss": 0.0426,
+      "step": 6615
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2661335110518717e-05,
+      "loss": 0.0491,
+      "step": 6620
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2655781406197936e-05,
+      "loss": 0.0736,
+      "step": 6625
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2650227701877153e-05,
+      "loss": 0.0255,
+      "step": 6630
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2644673997556371e-05,
+      "loss": 0.0312,
+      "step": 6635
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.263912029323559e-05,
+      "loss": 0.0603,
+      "step": 6640
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2633566588914807e-05,
+      "loss": 0.0887,
+      "step": 6645
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2628012884594025e-05,
+      "loss": 0.1525,
+      "step": 6650
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2622459180273244e-05,
+      "loss": 0.2394,
+      "step": 6655
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2616905475952461e-05,
+      "loss": 0.05,
+      "step": 6660
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2611351771631679e-05,
+      "loss": 0.0505,
+      "step": 6665
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2605798067310898e-05,
+      "loss": 0.0684,
+      "step": 6670
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2600244362990115e-05,
+      "loss": 0.0635,
+      "step": 6675
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.2594690658669333e-05,
+      "loss": 0.0348,
+      "step": 6680
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.258913695434855e-05,
+      "loss": 0.0427,
+      "step": 6685
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.258358325002777e-05,
+      "loss": 0.0642,
+      "step": 6690
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2578029545706987e-05,
+      "loss": 0.0918,
+      "step": 6695
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2572475841386204e-05,
+      "loss": 0.1498,
+      "step": 6700
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2566922137065423e-05,
+      "loss": 0.2131,
+      "step": 6705
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2561368432744642e-05,
+      "loss": 0.0363,
+      "step": 6710
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2555814728423861e-05,
+      "loss": 0.0616,
+      "step": 6715
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2550261024103079e-05,
+      "loss": 0.0785,
+      "step": 6720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2544707319782296e-05,
+      "loss": 0.0657,
+      "step": 6725
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.2539153615461515e-05,
+      "loss": 0.0637,
+      "step": 6730
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2533599911140733e-05,
+      "loss": 0.04,
+      "step": 6735
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.252804620681995e-05,
+      "loss": 0.0709,
+      "step": 6740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2522492502499168e-05,
+      "loss": 0.0971,
+      "step": 6745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2516938798178387e-05,
+      "loss": 0.1094,
+      "step": 6750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2511385093857604e-05,
+      "loss": 0.2893,
+      "step": 6755
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2505831389536822e-05,
+      "loss": 0.0442,
+      "step": 6760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.250027768521604e-05,
+      "loss": 0.0489,
+      "step": 6765
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2494723980895258e-05,
+      "loss": 0.056,
+      "step": 6770
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2489170276574475e-05,
+      "loss": 0.073,
+      "step": 6775
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2483616572253695e-05,
+      "loss": 0.0495,
+      "step": 6780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2478062867932912e-05,
+      "loss": 0.0237,
+      "step": 6785
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.247250916361213e-05,
+      "loss": 0.0724,
+      "step": 6790
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2466955459291349e-05,
+      "loss": 0.0929,
+      "step": 6795
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2461401754970566e-05,
+      "loss": 0.1426,
+      "step": 6800
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2455848050649783e-05,
+      "loss": 0.2286,
+      "step": 6805
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2450294346329e-05,
+      "loss": 0.05,
+      "step": 6810
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.244474064200822e-05,
+      "loss": 0.0473,
+      "step": 6815
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.2439186937687437e-05,
+      "loss": 0.051,
+      "step": 6820
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2433633233366658e-05,
+      "loss": 0.0817,
+      "step": 6825
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2428079529045876e-05,
+      "loss": 0.0203,
+      "step": 6830
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2422525824725093e-05,
+      "loss": 0.0397,
+      "step": 6835
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2416972120404312e-05,
+      "loss": 0.0618,
+      "step": 6840
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.241141841608353e-05,
+      "loss": 0.1006,
+      "step": 6845
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2405864711762747e-05,
+      "loss": 0.134,
+      "step": 6850
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2400311007441966e-05,
+      "loss": 0.312,
+      "step": 6855
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2394757303121183e-05,
+      "loss": 0.0491,
+      "step": 6860
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.2389203598800401e-05,
+      "loss": 0.0496,
+      "step": 6865
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2383649894479618e-05,
+      "loss": 0.0553,
+      "step": 6870
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2378096190158837e-05,
+      "loss": 0.0667,
+      "step": 6875
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2372542485838055e-05,
+      "loss": 0.0376,
+      "step": 6880
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2366988781517272e-05,
+      "loss": 0.0353,
+      "step": 6885
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2361435077196491e-05,
+      "loss": 0.0599,
+      "step": 6890
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2355881372875709e-05,
+      "loss": 0.1023,
+      "step": 6895
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2350327668554926e-05,
+      "loss": 0.1497,
+      "step": 6900
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2344773964234145e-05,
+      "loss": 0.2177,
+      "step": 6905
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.2339220259913363e-05,
+      "loss": 0.0432,
+      "step": 6910
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.233366655559258e-05,
+      "loss": 0.0474,
+      "step": 6915
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.23281128512718e-05,
+      "loss": 0.0568,
+      "step": 6920
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.2322559146951017e-05,
+      "loss": 0.0701,
+      "step": 6925
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.2317005442630234e-05,
+      "loss": 0.0396,
+      "step": 6930
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.2311451738309455e-05,
+      "loss": 0.0298,
+      "step": 6935
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.2305898033988672e-05,
+      "loss": 0.0596,
+      "step": 6940
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.230034432966789e-05,
+      "loss": 0.0562,
+      "step": 6945
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.2294790625347109e-05,
+      "loss": 0.1235,
+      "step": 6950
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.2289236921026326e-05,
+      "loss": 0.2481,
+      "step": 6955
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2283683216705544e-05,
+      "loss": 0.0567,
+      "step": 6960
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2278129512384763e-05,
+      "loss": 0.0562,
+      "step": 6965
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.227257580806398e-05,
+      "loss": 0.066,
+      "step": 6970
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2267022103743198e-05,
+      "loss": 0.075,
+      "step": 6975
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2261468399422417e-05,
+      "loss": 0.0479,
+      "step": 6980
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2255914695101634e-05,
+      "loss": 0.0332,
+      "step": 6985
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2250360990780852e-05,
+      "loss": 0.0799,
+      "step": 6990
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2244807286460069e-05,
+      "loss": 0.0816,
+      "step": 6995
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.2239253582139288e-05,
+      "loss": 0.1204,
+      "step": 7000
+    },
+    {
+      "epoch": 1.55,
+      "eval_loss": 0.027117447927594185,
+      "eval_runtime": 1647.1953,
+      "eval_samples_per_second": 9.724,
+      "eval_steps_per_second": 2.431,
+      "eval_wer": 0.056058420483231154,
+      "step": 7000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.2233699877818505e-05,
+      "loss": 0.2059,
+      "step": 7005
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.2228146173497723e-05,
+      "loss": 0.0512,
+      "step": 7010
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.2222592469176942e-05,
+      "loss": 0.0544,
+      "step": 7015
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.221703876485616e-05,
+      "loss": 0.0522,
+      "step": 7020
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.2211485060535377e-05,
+      "loss": 0.0632,
+      "step": 7025
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.2205931356214596e-05,
+      "loss": 0.0477,
+      "step": 7030
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.2200377651893813e-05,
+      "loss": 0.0509,
+      "step": 7035
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.219482394757303e-05,
+      "loss": 0.0441,
+      "step": 7040
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.218927024325225e-05,
+      "loss": 0.0859,
+      "step": 7045
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2183716538931469e-05,
+      "loss": 0.1346,
+      "step": 7050
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2178162834610688e-05,
+      "loss": 0.2427,
+      "step": 7055
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2172609130289906e-05,
+      "loss": 0.0402,
+      "step": 7060
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2167055425969123e-05,
+      "loss": 0.0498,
+      "step": 7065
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.216150172164834e-05,
+      "loss": 0.0684,
+      "step": 7070
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.215594801732756e-05,
+      "loss": 0.0651,
+      "step": 7075
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2150394313006777e-05,
+      "loss": 0.0335,
+      "step": 7080
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2144840608685994e-05,
+      "loss": 0.0369,
+      "step": 7085
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.2139286904365213e-05,
+      "loss": 0.0722,
+      "step": 7090
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.213373320004443e-05,
+      "loss": 0.0873,
+      "step": 7095
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2128179495723648e-05,
+      "loss": 0.1654,
+      "step": 7100
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2122625791402867e-05,
+      "loss": 0.2407,
+      "step": 7105
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2117072087082085e-05,
+      "loss": 0.0455,
+      "step": 7110
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2111518382761302e-05,
+      "loss": 0.0693,
+      "step": 7115
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2105964678440521e-05,
+      "loss": 0.0544,
+      "step": 7120
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2100410974119739e-05,
+      "loss": 0.0736,
+      "step": 7125
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2094857269798956e-05,
+      "loss": 0.0414,
+      "step": 7130
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.2089303565478174e-05,
+      "loss": 0.0357,
+      "step": 7135
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2083749861157393e-05,
+      "loss": 0.0553,
+      "step": 7140
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.207819615683661e-05,
+      "loss": 0.1045,
+      "step": 7145
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2072642452515827e-05,
+      "loss": 0.1349,
+      "step": 7150
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2067088748195047e-05,
+      "loss": 0.2134,
+      "step": 7155
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2061535043874266e-05,
+      "loss": 0.04,
+      "step": 7160
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2055981339553485e-05,
+      "loss": 0.0551,
+      "step": 7165
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2050427635232702e-05,
+      "loss": 0.0551,
+      "step": 7170
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.204487393091192e-05,
+      "loss": 0.0632,
+      "step": 7175
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.2039320226591139e-05,
+      "loss": 0.0415,
+      "step": 7180
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.2033766522270356e-05,
+      "loss": 0.031,
+      "step": 7185
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.2028212817949574e-05,
+      "loss": 0.0463,
+      "step": 7190
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.2022659113628791e-05,
+      "loss": 0.0627,
+      "step": 7195
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.201710540930801e-05,
+      "loss": 0.1171,
+      "step": 7200
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.2011551704987228e-05,
+      "loss": 0.2093,
+      "step": 7205
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.2005998000666445e-05,
+      "loss": 0.0492,
+      "step": 7210
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.2000444296345664e-05,
+      "loss": 0.053,
+      "step": 7215
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.1994890592024881e-05,
+      "loss": 0.1845,
+      "step": 7220
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.1989336887704099e-05,
+      "loss": 0.0706,
+      "step": 7225
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1983783183383318e-05,
+      "loss": 0.0388,
+      "step": 7230
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1978229479062535e-05,
+      "loss": 0.0338,
+      "step": 7235
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1972675774741753e-05,
+      "loss": 0.0615,
+      "step": 7240
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1967122070420972e-05,
+      "loss": 0.0866,
+      "step": 7245
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.196156836610019e-05,
+      "loss": 0.1337,
+      "step": 7250
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1956014661779407e-05,
+      "loss": 0.195,
+      "step": 7255
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1950460957458624e-05,
+      "loss": 0.041,
+      "step": 7260
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.1944907253137843e-05,
+      "loss": 0.0488,
+      "step": 7265
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.193935354881706e-05,
+      "loss": 0.0548,
+      "step": 7270
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1933799844496282e-05,
+      "loss": 0.0665,
+      "step": 7275
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1928246140175499e-05,
+      "loss": 0.047,
+      "step": 7280
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1922692435854716e-05,
+      "loss": 0.0419,
+      "step": 7285
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1917138731533935e-05,
+      "loss": 0.0452,
+      "step": 7290
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1911585027213153e-05,
+      "loss": 0.0886,
+      "step": 7295
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.190603132289237e-05,
+      "loss": 0.1262,
+      "step": 7300
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.190047761857159e-05,
+      "loss": 0.2426,
+      "step": 7305
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1894923914250807e-05,
+      "loss": 0.0561,
+      "step": 7310
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.1889370209930024e-05,
+      "loss": 0.0536,
+      "step": 7315
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1883816505609242e-05,
+      "loss": 0.0539,
+      "step": 7320
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.187826280128846e-05,
+      "loss": 0.0682,
+      "step": 7325
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1872709096967678e-05,
+      "loss": 0.0371,
+      "step": 7330
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1867155392646896e-05,
+      "loss": 0.048,
+      "step": 7335
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1861601688326115e-05,
+      "loss": 0.0471,
+      "step": 7340
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1856047984005332e-05,
+      "loss": 0.0978,
+      "step": 7345
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.185049427968455e-05,
+      "loss": 0.1092,
+      "step": 7350
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1844940575363769e-05,
+      "loss": 0.2146,
+      "step": 7355
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1839386871042986e-05,
+      "loss": 0.0475,
+      "step": 7360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1833833166722204e-05,
+      "loss": 0.0544,
+      "step": 7365
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1828279462401423e-05,
+      "loss": 0.0608,
+      "step": 7370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.182272575808064e-05,
+      "loss": 0.0653,
+      "step": 7375
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1817172053759857e-05,
+      "loss": 0.0313,
+      "step": 7380
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1811618349439078e-05,
+      "loss": 0.0391,
+      "step": 7385
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1806064645118296e-05,
+      "loss": 0.0607,
+      "step": 7390
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1800510940797513e-05,
+      "loss": 0.0793,
+      "step": 7395
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.1794957236476732e-05,
+      "loss": 0.1472,
+      "step": 7400
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.178940353215595e-05,
+      "loss": 0.2493,
+      "step": 7405
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1783849827835167e-05,
+      "loss": 0.0423,
+      "step": 7410
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1778296123514386e-05,
+      "loss": 0.0552,
+      "step": 7415
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1772742419193604e-05,
+      "loss": 0.0591,
+      "step": 7420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1767188714872821e-05,
+      "loss": 0.0556,
+      "step": 7425
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.176163501055204e-05,
+      "loss": 0.0475,
+      "step": 7430
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1756081306231258e-05,
+      "loss": 0.024,
+      "step": 7435
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1750527601910475e-05,
+      "loss": 0.0507,
+      "step": 7440
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1744973897589694e-05,
+      "loss": 0.0699,
+      "step": 7445
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.1739420193268911e-05,
+      "loss": 0.1329,
+      "step": 7450
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1733866488948129e-05,
+      "loss": 0.1718,
+      "step": 7455
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1728312784627346e-05,
+      "loss": 0.053,
+      "step": 7460
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1722759080306565e-05,
+      "loss": 0.0483,
+      "step": 7465
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1717205375985783e-05,
+      "loss": 0.0645,
+      "step": 7470
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1711651671665e-05,
+      "loss": 0.0706,
+      "step": 7475
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.170609796734422e-05,
+      "loss": 0.0368,
+      "step": 7480
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1700544263023437e-05,
+      "loss": 0.0408,
+      "step": 7485
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1694990558702654e-05,
+      "loss": 0.0493,
+      "step": 7490
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.1689436854381873e-05,
+      "loss": 0.0625,
+      "step": 7495
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1683883150061092e-05,
+      "loss": 0.145,
+      "step": 7500
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1678329445740311e-05,
+      "loss": 0.1962,
+      "step": 7505
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1672775741419529e-05,
+      "loss": 0.0513,
+      "step": 7510
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1667222037098746e-05,
+      "loss": 0.0551,
+      "step": 7515
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1661668332777964e-05,
+      "loss": 0.0613,
+      "step": 7520
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1656114628457183e-05,
+      "loss": 0.0678,
+      "step": 7525
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.16505609241364e-05,
+      "loss": 0.0293,
+      "step": 7530
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1646117960679775e-05,
+      "loss": 0.0313,
+      "step": 7535
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1640564256358992e-05,
+      "loss": 0.0564,
+      "step": 7540
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1635010552038211e-05,
+      "loss": 0.0582,
+      "step": 7545
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1629456847717428e-05,
+      "loss": 0.1424,
+      "step": 7550
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1623903143396646e-05,
+      "loss": 0.2103,
+      "step": 7555
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1618349439075865e-05,
+      "loss": 0.0474,
+      "step": 7560
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1612795734755082e-05,
+      "loss": 0.0472,
+      "step": 7565
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.16072420304343e-05,
+      "loss": 0.0701,
+      "step": 7570
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1601688326113519e-05,
+      "loss": 0.0647,
+      "step": 7575
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1596134621792736e-05,
+      "loss": 0.0501,
+      "step": 7580
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.1590580917471954e-05,
+      "loss": 0.0285,
+      "step": 7585
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1585027213151173e-05,
+      "loss": 0.0568,
+      "step": 7590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.157947350883039e-05,
+      "loss": 0.0781,
+      "step": 7595
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1573919804509608e-05,
+      "loss": 0.1475,
+      "step": 7600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1568366100188825e-05,
+      "loss": 0.2475,
+      "step": 7605
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1562812395868044e-05,
+      "loss": 0.0549,
+      "step": 7610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1557258691547263e-05,
+      "loss": 0.0523,
+      "step": 7615
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1551704987226482e-05,
+      "loss": 0.0553,
+      "step": 7620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.15461512829057e-05,
+      "loss": 0.0525,
+      "step": 7625
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.1540597578584917e-05,
+      "loss": 0.0455,
+      "step": 7630
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1535043874264136e-05,
+      "loss": 0.0344,
+      "step": 7635
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1529490169943354e-05,
+      "loss": 0.0545,
+      "step": 7640
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1523936465622571e-05,
+      "loss": 0.0578,
+      "step": 7645
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.151838276130179e-05,
+      "loss": 0.1248,
+      "step": 7650
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1512829056981008e-05,
+      "loss": 0.2677,
+      "step": 7655
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1507275352660225e-05,
+      "loss": 0.0539,
+      "step": 7660
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1501721648339443e-05,
+      "loss": 0.0385,
+      "step": 7665
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.1496167944018662e-05,
+      "loss": 0.0455,
+      "step": 7670
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.149061423969788e-05,
+      "loss": 0.0601,
+      "step": 7675
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1485060535377097e-05,
+      "loss": 0.0525,
+      "step": 7680
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1479506831056316e-05,
+      "loss": 0.0452,
+      "step": 7685
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1473953126735533e-05,
+      "loss": 0.0465,
+      "step": 7690
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.146839942241475e-05,
+      "loss": 0.0823,
+      "step": 7695
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.146284571809397e-05,
+      "loss": 0.0875,
+      "step": 7700
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1457292013773187e-05,
+      "loss": 0.2495,
+      "step": 7705
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1451738309452404e-05,
+      "loss": 0.0484,
+      "step": 7710
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1446184605131624e-05,
+      "loss": 0.0515,
+      "step": 7715
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1440630900810841e-05,
+      "loss": 0.0504,
+      "step": 7720
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.143507719649006e-05,
+      "loss": 0.0628,
+      "step": 7725
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.142952349216928e-05,
+      "loss": 0.056,
+      "step": 7730
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1423969787848497e-05,
+      "loss": 0.0475,
+      "step": 7735
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1418416083527714e-05,
+      "loss": 0.0462,
+      "step": 7740
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1412862379206933e-05,
+      "loss": 0.0591,
+      "step": 7745
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.140730867488615e-05,
+      "loss": 0.1507,
+      "step": 7750
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1401754970565368e-05,
+      "loss": 0.1823,
+      "step": 7755
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1396201266244587e-05,
+      "loss": 0.048,
+      "step": 7760
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1390647561923805e-05,
+      "loss": 0.0697,
+      "step": 7765
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.1385093857603022e-05,
+      "loss": 0.0787,
+      "step": 7770
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1379540153282241e-05,
+      "loss": 0.0586,
+      "step": 7775
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1373986448961458e-05,
+      "loss": 0.0459,
+      "step": 7780
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1368432744640676e-05,
+      "loss": 0.0367,
+      "step": 7785
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1362879040319893e-05,
+      "loss": 0.0471,
+      "step": 7790
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1357325335999112e-05,
+      "loss": 0.0985,
+      "step": 7795
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.135177163167833e-05,
+      "loss": 0.124,
+      "step": 7800
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1346217927357547e-05,
+      "loss": 0.2438,
+      "step": 7805
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1340664223036766e-05,
+      "loss": 0.0481,
+      "step": 7810
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.1335110518715984e-05,
+      "loss": 0.0572,
+      "step": 7815
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.1329556814395201e-05,
+      "loss": 0.0509,
+      "step": 7820
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.132400311007442e-05,
+      "loss": 0.0615,
+      "step": 7825
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.1318449405753638e-05,
+      "loss": 0.0505,
+      "step": 7830
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.1312895701432855e-05,
+      "loss": 0.0426,
+      "step": 7835
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.1307341997112076e-05,
+      "loss": 0.0757,
+      "step": 7840
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.1301788292791293e-05,
+      "loss": 0.0897,
+      "step": 7845
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.129623458847051e-05,
+      "loss": 0.1247,
+      "step": 7850
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.129068088414973e-05,
+      "loss": 0.2354,
+      "step": 7855
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.1285127179828947e-05,
+      "loss": 0.0326,
+      "step": 7860
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1279573475508165e-05,
+      "loss": 0.0504,
+      "step": 7865
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1274019771187384e-05,
+      "loss": 0.0707,
+      "step": 7870
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1268466066866601e-05,
+      "loss": 0.0679,
+      "step": 7875
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1262912362545819e-05,
+      "loss": 0.0325,
+      "step": 7880
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1257358658225038e-05,
+      "loss": 0.0275,
+      "step": 7885
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1251804953904255e-05,
+      "loss": 0.0711,
+      "step": 7890
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1246251249583473e-05,
+      "loss": 0.069,
+      "step": 7895
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1240697545262692e-05,
+      "loss": 0.1488,
+      "step": 7900
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.1235143840941909e-05,
+      "loss": 0.2312,
+      "step": 7905
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1229590136621127e-05,
+      "loss": 0.062,
+      "step": 7910
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1224036432300344e-05,
+      "loss": 0.0652,
+      "step": 7915
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1218482727979563e-05,
+      "loss": 0.0546,
+      "step": 7920
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.121292902365878e-05,
+      "loss": 0.0632,
+      "step": 7925
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1207375319337998e-05,
+      "loss": 0.0475,
+      "step": 7930
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1201821615017217e-05,
+      "loss": 0.0366,
+      "step": 7935
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1196267910696434e-05,
+      "loss": 0.0815,
+      "step": 7940
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1190714206375652e-05,
+      "loss": 0.1118,
+      "step": 7945
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1185160502054873e-05,
+      "loss": 0.1409,
+      "step": 7950
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.117960679773409e-05,
+      "loss": 0.1908,
+      "step": 7955
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.117405309341331e-05,
+      "loss": 0.0456,
+      "step": 7960
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1168499389092527e-05,
+      "loss": 0.0601,
+      "step": 7965
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1162945684771744e-05,
+      "loss": 0.0422,
+      "step": 7970
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1157391980450963e-05,
+      "loss": 0.0773,
+      "step": 7975
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.115183827613018e-05,
+      "loss": 0.0419,
+      "step": 7980
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1146284571809398e-05,
+      "loss": 0.0283,
+      "step": 7985
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1140730867488615e-05,
+      "loss": 0.0736,
+      "step": 7990
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1135177163167834e-05,
+      "loss": 0.0577,
+      "step": 7995
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1129623458847052e-05,
+      "loss": 0.0972,
+      "step": 8000
+    },
+    {
+      "epoch": 1.78,
+      "eval_loss": 0.02691551484167576,
+      "eval_runtime": 1624.5375,
+      "eval_samples_per_second": 9.86,
+      "eval_steps_per_second": 2.465,
+      "eval_wer": 0.05488640461593942,
+      "step": 8000
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.112406975452627e-05,
+      "loss": 0.2562,
+      "step": 8005
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1118516050205488e-05,
+      "loss": 0.0461,
+      "step": 8010
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1112962345884706e-05,
+      "loss": 0.0541,
+      "step": 8015
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1107408641563923e-05,
+      "loss": 0.0502,
+      "step": 8020
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1101854937243142e-05,
+      "loss": 0.0555,
+      "step": 8025
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.109630123292236e-05,
+      "loss": 0.0367,
+      "step": 8030
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1090747528601577e-05,
+      "loss": 0.0316,
+      "step": 8035
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1085193824280796e-05,
+      "loss": 0.0562,
+      "step": 8040
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1079640119960014e-05,
+      "loss": 0.0894,
+      "step": 8045
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1074086415639231e-05,
+      "loss": 0.1359,
+      "step": 8050
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1068532711318449e-05,
+      "loss": 0.2298,
+      "step": 8055
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1062979006997668e-05,
+      "loss": 0.0501,
+      "step": 8060
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1057425302676887e-05,
+      "loss": 0.0622,
+      "step": 8065
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1051871598356106e-05,
+      "loss": 0.0437,
+      "step": 8070
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.1046317894035323e-05,
+      "loss": 0.0666,
+      "step": 8075
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.104076418971454e-05,
+      "loss": 0.0445,
+      "step": 8080
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.103521048539376e-05,
+      "loss": 0.0316,
+      "step": 8085
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1029656781072977e-05,
+      "loss": 0.0617,
+      "step": 8090
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1024103076752195e-05,
+      "loss": 0.0841,
+      "step": 8095
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1018549372431414e-05,
+      "loss": 0.1618,
+      "step": 8100
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1012995668110631e-05,
+      "loss": 0.1914,
+      "step": 8105
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1007441963789849e-05,
+      "loss": 0.0455,
+      "step": 8110
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1001888259469066e-05,
+      "loss": 0.0549,
+      "step": 8115
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0996334555148285e-05,
+      "loss": 0.056,
+      "step": 8120
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0990780850827503e-05,
+      "loss": 0.0736,
+      "step": 8125
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.098522714650672e-05,
+      "loss": 0.0356,
+      "step": 8130
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0979673442185939e-05,
+      "loss": 0.0463,
+      "step": 8135
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0974119737865157e-05,
+      "loss": 0.0422,
+      "step": 8140
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0968566033544374e-05,
+      "loss": 0.0712,
+      "step": 8145
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0963012329223593e-05,
+      "loss": 0.1427,
+      "step": 8150
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.095745862490281e-05,
+      "loss": 0.1629,
+      "step": 8155
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0951904920582028e-05,
+      "loss": 0.0498,
+      "step": 8160
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0946351216261247e-05,
+      "loss": 0.0657,
+      "step": 8165
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0940797511940464e-05,
+      "loss": 0.0585,
+      "step": 8170
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.0935243807619683e-05,
+      "loss": 0.0732,
+      "step": 8175
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0929690103298903e-05,
+      "loss": 0.0308,
+      "step": 8180
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.092413639897812e-05,
+      "loss": 0.0381,
+      "step": 8185
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0918582694657337e-05,
+      "loss": 0.0627,
+      "step": 8190
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0913028990336557e-05,
+      "loss": 0.0753,
+      "step": 8195
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0907475286015774e-05,
+      "loss": 0.1047,
+      "step": 8200
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0901921581694991e-05,
+      "loss": 0.2234,
+      "step": 8205
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.089636787737421e-05,
+      "loss": 0.0495,
+      "step": 8210
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0890814173053428e-05,
+      "loss": 0.0519,
+      "step": 8215
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.0885260468732645e-05,
+      "loss": 0.0506,
+      "step": 8220
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.0879706764411864e-05,
+      "loss": 0.0738,
+      "step": 8225
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.0874153060091082e-05,
+      "loss": 0.035,
+      "step": 8230
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.08685993557703e-05,
+      "loss": 0.0431,
+      "step": 8235
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.0863045651449517e-05,
+      "loss": 0.0685,
+      "step": 8240
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.0857491947128736e-05,
+      "loss": 0.0888,
+      "step": 8245
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.0851938242807953e-05,
+      "loss": 0.096,
+      "step": 8250
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.084638453848717e-05,
+      "loss": 0.2263,
+      "step": 8255
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.084083083416639e-05,
+      "loss": 0.0437,
+      "step": 8260
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.0835277129845607e-05,
+      "loss": 0.0499,
+      "step": 8265
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0829723425524825e-05,
+      "loss": 0.042,
+      "step": 8270
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0824169721204044e-05,
+      "loss": 0.0815,
+      "step": 8275
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0818616016883261e-05,
+      "loss": 0.0343,
+      "step": 8280
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0813062312562479e-05,
+      "loss": 0.0542,
+      "step": 8285
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.08075086082417e-05,
+      "loss": 0.0467,
+      "step": 8290
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0801954903920917e-05,
+      "loss": 0.0638,
+      "step": 8295
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0796401199600134e-05,
+      "loss": 0.128,
+      "step": 8300
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.0790847495279353e-05,
+      "loss": 0.2118,
+      "step": 8305
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.078529379095857e-05,
+      "loss": 0.065,
+      "step": 8310
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0779740086637788e-05,
+      "loss": 0.0498,
+      "step": 8315
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0774186382317007e-05,
+      "loss": 0.0554,
+      "step": 8320
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0768632677996225e-05,
+      "loss": 0.0695,
+      "step": 8325
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0763078973675442e-05,
+      "loss": 0.0296,
+      "step": 8330
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0757525269354661e-05,
+      "loss": 0.0304,
+      "step": 8335
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0751971565033879e-05,
+      "loss": 0.059,
+      "step": 8340
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0746417860713096e-05,
+      "loss": 0.0986,
+      "step": 8345
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0740864156392315e-05,
+      "loss": 0.1288,
+      "step": 8350
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0735310452071533e-05,
+      "loss": 0.2048,
+      "step": 8355
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.072975674775075e-05,
+      "loss": 0.0604,
+      "step": 8360
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0724203043429969e-05,
+      "loss": 0.0435,
+      "step": 8365
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0718649339109186e-05,
+      "loss": 0.0495,
+      "step": 8370
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0713095634788404e-05,
+      "loss": 0.0666,
+      "step": 8375
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0707541930467621e-05,
+      "loss": 0.129,
+      "step": 8380
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.070198822614684e-05,
+      "loss": 0.0341,
+      "step": 8385
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0696434521826058e-05,
+      "loss": 0.0519,
+      "step": 8390
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0690880817505275e-05,
+      "loss": 0.0908,
+      "step": 8395
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.0685327113184496e-05,
+      "loss": 0.1108,
+      "step": 8400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0679773408863713e-05,
+      "loss": 0.2783,
+      "step": 8405
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0674219704542933e-05,
+      "loss": 0.0503,
+      "step": 8410
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.066866600022215e-05,
+      "loss": 0.0536,
+      "step": 8415
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0663112295901367e-05,
+      "loss": 0.0616,
+      "step": 8420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0657558591580587e-05,
+      "loss": 0.0612,
+      "step": 8425
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0652004887259804e-05,
+      "loss": 0.0254,
+      "step": 8430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0646451182939021e-05,
+      "loss": 0.0353,
+      "step": 8435
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0640897478618239e-05,
+      "loss": 0.0457,
+      "step": 8440
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.0635343774297458e-05,
+      "loss": 0.0738,
+      "step": 8445
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0629790069976675e-05,
+      "loss": 0.1255,
+      "step": 8450
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0624236365655893e-05,
+      "loss": 0.2515,
+      "step": 8455
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0618682661335112e-05,
+      "loss": 0.0552,
+      "step": 8460
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.061312895701433e-05,
+      "loss": 0.051,
+      "step": 8465
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0607575252693547e-05,
+      "loss": 0.0572,
+      "step": 8470
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0602021548372766e-05,
+      "loss": 0.0655,
+      "step": 8475
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0596467844051983e-05,
+      "loss": 0.0399,
+      "step": 8480
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.05909141397312e-05,
+      "loss": 0.0396,
+      "step": 8485
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.058536043541042e-05,
+      "loss": 0.056,
+      "step": 8490
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0579806731089637e-05,
+      "loss": 0.0787,
+      "step": 8495
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0574253026768855e-05,
+      "loss": 0.1596,
+      "step": 8500
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0568699322448072e-05,
+      "loss": 0.1823,
+      "step": 8505
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0563145618127291e-05,
+      "loss": 0.061,
+      "step": 8510
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.055759191380651e-05,
+      "loss": 0.0387,
+      "step": 8515
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.055203820948573e-05,
+      "loss": 0.052,
+      "step": 8520
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0546484505164947e-05,
+      "loss": 0.0669,
+      "step": 8525
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0540930800844164e-05,
+      "loss": 0.0329,
+      "step": 8530
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0535377096523383e-05,
+      "loss": 0.0452,
+      "step": 8535
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.05298233922026e-05,
+      "loss": 0.049,
+      "step": 8540
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0524269687881818e-05,
+      "loss": 0.0844,
+      "step": 8545
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0518715983561037e-05,
+      "loss": 0.1493,
+      "step": 8550
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0513162279240255e-05,
+      "loss": 0.1364,
+      "step": 8555
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0507608574919472e-05,
+      "loss": 0.0461,
+      "step": 8560
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.050205487059869e-05,
+      "loss": 0.0651,
+      "step": 8565
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0496501166277909e-05,
+      "loss": 0.0619,
+      "step": 8570
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0490947461957126e-05,
+      "loss": 0.0763,
+      "step": 8575
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0485393757636343e-05,
+      "loss": 0.0373,
+      "step": 8580
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0479840053315562e-05,
+      "loss": 0.0356,
+      "step": 8585
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.047428634899478e-05,
+      "loss": 0.0485,
+      "step": 8590
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0468732644673997e-05,
+      "loss": 0.0943,
+      "step": 8595
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0463178940353216e-05,
+      "loss": 0.0908,
+      "step": 8600
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0457625236032434e-05,
+      "loss": 0.2225,
+      "step": 8605
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0452071531711651e-05,
+      "loss": 0.0463,
+      "step": 8610
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.044651782739087e-05,
+      "loss": 0.0471,
+      "step": 8615
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0440964123070088e-05,
+      "loss": 0.0716,
+      "step": 8620
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.0435410418749307e-05,
+      "loss": 0.0744,
+      "step": 8625
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0429856714428526e-05,
+      "loss": 0.0426,
+      "step": 8630
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0424303010107743e-05,
+      "loss": 0.0453,
+      "step": 8635
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.041874930578696e-05,
+      "loss": 0.0447,
+      "step": 8640
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.041319560146618e-05,
+      "loss": 0.073,
+      "step": 8645
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0407641897145397e-05,
+      "loss": 0.107,
+      "step": 8650
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0402088192824615e-05,
+      "loss": 0.2009,
+      "step": 8655
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0396534488503834e-05,
+      "loss": 0.0427,
+      "step": 8660
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0390980784183051e-05,
+      "loss": 0.059,
+      "step": 8665
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.0385427079862269e-05,
+      "loss": 0.0649,
+      "step": 8670
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0379873375541488e-05,
+      "loss": 0.0644,
+      "step": 8675
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0374319671220705e-05,
+      "loss": 0.0317,
+      "step": 8680
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0368765966899923e-05,
+      "loss": 0.0297,
+      "step": 8685
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.036321226257914e-05,
+      "loss": 0.0412,
+      "step": 8690
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.035765855825836e-05,
+      "loss": 0.0692,
+      "step": 8695
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0352104853937577e-05,
+      "loss": 0.1505,
+      "step": 8700
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0346551149616794e-05,
+      "loss": 0.2174,
+      "step": 8705
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0340997445296013e-05,
+      "loss": 0.0442,
+      "step": 8710
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.033544374097523e-05,
+      "loss": 0.0537,
+      "step": 8715
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0329890036654448e-05,
+      "loss": 0.045,
+      "step": 8720
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0324336332333667e-05,
+      "loss": 0.0738,
+      "step": 8725
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0318782628012885e-05,
+      "loss": 0.0282,
+      "step": 8730
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0313228923692102e-05,
+      "loss": 0.0412,
+      "step": 8735
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0307675219371323e-05,
+      "loss": 0.0702,
+      "step": 8740
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.030212151505054e-05,
+      "loss": 0.0747,
+      "step": 8745
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.029656781072976e-05,
+      "loss": 0.1086,
+      "step": 8750
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0291014106408977e-05,
+      "loss": 0.1858,
+      "step": 8755
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.0285460402088194e-05,
+      "loss": 0.0432,
+      "step": 8760
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0279906697767412e-05,
+      "loss": 0.0466,
+      "step": 8765
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.027435299344663e-05,
+      "loss": 0.046,
+      "step": 8770
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0268799289125848e-05,
+      "loss": 0.0686,
+      "step": 8775
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0263245584805065e-05,
+      "loss": 0.0414,
+      "step": 8780
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0257691880484285e-05,
+      "loss": 0.0472,
+      "step": 8785
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0252138176163502e-05,
+      "loss": 0.0663,
+      "step": 8790
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.024658447184272e-05,
+      "loss": 0.0782,
+      "step": 8795
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0241030767521939e-05,
+      "loss": 0.1499,
+      "step": 8800
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.0235477063201156e-05,
+      "loss": 0.1651,
+      "step": 8805
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0229923358880373e-05,
+      "loss": 0.05,
+      "step": 8810
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0224369654559592e-05,
+      "loss": 0.043,
+      "step": 8815
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.021881595023881e-05,
+      "loss": 0.046,
+      "step": 8820
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0213262245918027e-05,
+      "loss": 0.0564,
+      "step": 8825
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0207708541597245e-05,
+      "loss": 0.0452,
+      "step": 8830
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0202154837276464e-05,
+      "loss": 0.0388,
+      "step": 8835
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0196601132955681e-05,
+      "loss": 0.0564,
+      "step": 8840
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.0191047428634899e-05,
+      "loss": 0.0718,
+      "step": 8845
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.018549372431412e-05,
+      "loss": 0.0818,
+      "step": 8850
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0179940019993337e-05,
+      "loss": 0.1738,
+      "step": 8855
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0174386315672556e-05,
+      "loss": 0.039,
+      "step": 8860
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0168832611351773e-05,
+      "loss": 0.0535,
+      "step": 8865
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.016327890703099e-05,
+      "loss": 0.0531,
+      "step": 8870
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.015772520271021e-05,
+      "loss": 0.0662,
+      "step": 8875
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0152171498389427e-05,
+      "loss": 0.0336,
+      "step": 8880
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0146617794068645e-05,
+      "loss": 0.0307,
+      "step": 8885
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0141064089747862e-05,
+      "loss": 0.0746,
+      "step": 8890
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.0135510385427081e-05,
+      "loss": 0.0757,
+      "step": 8895
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0129956681106299e-05,
+      "loss": 0.1042,
+      "step": 8900
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0124402976785516e-05,
+      "loss": 0.2255,
+      "step": 8905
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0118849272464735e-05,
+      "loss": 0.0505,
+      "step": 8910
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0113295568143953e-05,
+      "loss": 0.0374,
+      "step": 8915
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.010774186382317e-05,
+      "loss": 0.0561,
+      "step": 8920
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.010218815950239e-05,
+      "loss": 0.0733,
+      "step": 8925
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0096634455181607e-05,
+      "loss": 0.0327,
+      "step": 8930
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0091080750860824e-05,
+      "loss": 0.0451,
+      "step": 8935
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0085527046540043e-05,
+      "loss": 0.0552,
+      "step": 8940
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.007997334221926e-05,
+      "loss": 0.0904,
+      "step": 8945
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0074419637898478e-05,
+      "loss": 0.1209,
+      "step": 8950
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0068865933577695e-05,
+      "loss": 0.2235,
+      "step": 8955
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0063312229256914e-05,
+      "loss": 0.0499,
+      "step": 8960
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0057758524936134e-05,
+      "loss": 0.0525,
+      "step": 8965
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0052204820615353e-05,
+      "loss": 0.0573,
+      "step": 8970
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.004665111629457e-05,
+      "loss": 0.0502,
+      "step": 8975
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0041097411973788e-05,
+      "loss": 0.0509,
+      "step": 8980
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.0035543707653007e-05,
+      "loss": 0.0369,
+      "step": 8985
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0029990003332224e-05,
+      "loss": 0.0497,
+      "step": 8990
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0024436299011441e-05,
+      "loss": 0.0941,
+      "step": 8995
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.001888259469066e-05,
+      "loss": 0.1341,
+      "step": 9000
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.031054431572556496,
+      "eval_runtime": 1776.0483,
+      "eval_samples_per_second": 9.019,
+      "eval_steps_per_second": 2.255,
+      "eval_wer": 0.05692390912369275,
+      "step": 9000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0013328890369878e-05,
+      "loss": 0.1905,
+      "step": 9005
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0007775186049095e-05,
+      "loss": 0.086,
+      "step": 9010
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0002221481728313e-05,
+      "loss": 0.0447,
+      "step": 9015
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.996667777407532e-06,
+      "loss": 0.0493,
+      "step": 9020
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.99111407308675e-06,
+      "loss": 0.0647,
+      "step": 9025
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.985560368765967e-06,
+      "loss": 0.0673,
+      "step": 9030
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.980006664445186e-06,
+      "loss": 0.0262,
+      "step": 9035
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.974452960124403e-06,
+      "loss": 0.0408,
+      "step": 9040
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.96889925580362e-06,
+      "loss": 0.0359,
+      "step": 9045
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.96334555148284e-06,
+      "loss": 0.0901,
+      "step": 9050
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.957791847162059e-06,
+      "loss": 0.1306,
+      "step": 9055
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.952238142841276e-06,
+      "loss": 0.1773,
+      "step": 9060
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.946684438520494e-06,
+      "loss": 0.0395,
+      "step": 9065
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.941130734199713e-06,
+      "loss": 0.0617,
+      "step": 9070
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.93557702987893e-06,
+      "loss": 0.0703,
+      "step": 9075
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.930023325558148e-06,
+      "loss": 0.0717,
+      "step": 9080
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.924469621237365e-06,
+      "loss": 0.0327,
+      "step": 9085
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.918915916916584e-06,
+      "loss": 0.0282,
+      "step": 9090
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.913362212595802e-06,
+      "loss": 0.0657,
+      "step": 9095
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.907808508275019e-06,
+      "loss": 0.096,
+      "step": 9100
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.902254803954238e-06,
+      "loss": 0.1246,
+      "step": 9105
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.896701099633457e-06,
+      "loss": 0.165,
+      "step": 9110
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.891147395312675e-06,
+      "loss": 0.0473,
+      "step": 9115
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.885593690991892e-06,
+      "loss": 0.0549,
+      "step": 9120
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.880039986671111e-06,
+      "loss": 0.0478,
+      "step": 9125
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.874486282350329e-06,
+      "loss": 0.06,
+      "step": 9130
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.868932578029546e-06,
+      "loss": 0.0778,
+      "step": 9135
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.863378873708765e-06,
+      "loss": 0.026,
+      "step": 9140
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.857825169387983e-06,
+      "loss": 0.0542,
+      "step": 9145
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.8522714650672e-06,
+      "loss": 0.0593,
+      "step": 9150
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.846717760746417e-06,
+      "loss": 0.1238,
+      "step": 9155
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.841164056425637e-06,
+      "loss": 0.27,
+      "step": 9160
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.835610352104856e-06,
+      "loss": 0.0471,
+      "step": 9165
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.830056647784073e-06,
+      "loss": 0.0415,
+      "step": 9170
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.82450294346329e-06,
+      "loss": 0.0672,
+      "step": 9175
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.81894923914251e-06,
+      "loss": 0.06,
+      "step": 9180
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.813395534821727e-06,
+      "loss": 0.0386,
+      "step": 9185
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.807841830500944e-06,
+      "loss": 0.0364,
+      "step": 9190
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.802288126180164e-06,
+      "loss": 0.0606,
+      "step": 9195
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.796734421859381e-06,
+      "loss": 0.0754,
+      "step": 9200
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.791180717538598e-06,
+      "loss": 0.1789,
+      "step": 9205
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.785627013217816e-06,
+      "loss": 0.1736,
+      "step": 9210
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.780073308897035e-06,
+      "loss": 0.0445,
+      "step": 9215
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.774519604576254e-06,
+      "loss": 0.0505,
+      "step": 9220
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.768965900255471e-06,
+      "loss": 0.0537,
+      "step": 9225
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.763412195934689e-06,
+      "loss": 0.0676,
+      "step": 9230
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.757858491613908e-06,
+      "loss": 0.0233,
+      "step": 9235
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.752304787293125e-06,
+      "loss": 0.0296,
+      "step": 9240
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.746751082972343e-06,
+      "loss": 0.064,
+      "step": 9245
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.741197378651562e-06,
+      "loss": 0.0623,
+      "step": 9250
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.73564367433078e-06,
+      "loss": 0.1334,
+      "step": 9255
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.730089970009997e-06,
+      "loss": 0.1873,
+      "step": 9260
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.724536265689216e-06,
+      "loss": 0.0442,
+      "step": 9265
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.718982561368433e-06,
+      "loss": 0.0727,
+      "step": 9270
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.713428857047652e-06,
+      "loss": 0.0578,
+      "step": 9275
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.70787515272687e-06,
+      "loss": 0.0843,
+      "step": 9280
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.702321448406087e-06,
+      "loss": 0.0391,
+      "step": 9285
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.696767744085306e-06,
+      "loss": 0.048,
+      "step": 9290
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.691214039764524e-06,
+      "loss": 0.0437,
+      "step": 9295
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.685660335443741e-06,
+      "loss": 0.1035,
+      "step": 9300
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.68010663112296e-06,
+      "loss": 0.1373,
+      "step": 9305
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.674552926802178e-06,
+      "loss": 0.1879,
+      "step": 9310
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.668999222481395e-06,
+      "loss": 0.046,
+      "step": 9315
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.663445518160614e-06,
+      "loss": 0.0452,
+      "step": 9320
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.657891813839832e-06,
+      "loss": 0.0588,
+      "step": 9325
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.65233810951905e-06,
+      "loss": 0.064,
+      "step": 9330
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.646784405198268e-06,
+      "loss": 0.0263,
+      "step": 9335
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.641230700877486e-06,
+      "loss": 0.0523,
+      "step": 9340
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.635676996556705e-06,
+      "loss": 0.0356,
+      "step": 9345
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.630123292235922e-06,
+      "loss": 0.1031,
+      "step": 9350
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.62456958791514e-06,
+      "loss": 0.0794,
+      "step": 9355
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.619015883594359e-06,
+      "loss": 0.1389,
+      "step": 9360
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.613462179273576e-06,
+      "loss": 0.0473,
+      "step": 9365
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.607908474952793e-06,
+      "loss": 0.0536,
+      "step": 9370
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.602354770632013e-06,
+      "loss": 0.0799,
+      "step": 9375
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.59680106631123e-06,
+      "loss": 0.0752,
+      "step": 9380
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.591247361990449e-06,
+      "loss": 0.0223,
+      "step": 9385
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.585693657669667e-06,
+      "loss": 0.0483,
+      "step": 9390
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.580139953348886e-06,
+      "loss": 0.0621,
+      "step": 9395
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.574586249028103e-06,
+      "loss": 0.1037,
+      "step": 9400
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.56903254470732e-06,
+      "loss": 0.1416,
+      "step": 9405
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.563478840386538e-06,
+      "loss": 0.2099,
+      "step": 9410
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.557925136065757e-06,
+      "loss": 0.0368,
+      "step": 9415
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.552371431744974e-06,
+      "loss": 0.0412,
+      "step": 9420
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.546817727424192e-06,
+      "loss": 0.0459,
+      "step": 9425
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.541264023103411e-06,
+      "loss": 0.0565,
+      "step": 9430
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.535710318782628e-06,
+      "loss": 0.032,
+      "step": 9435
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.530156614461847e-06,
+      "loss": 0.0197,
+      "step": 9440
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.524602910141065e-06,
+      "loss": 0.0458,
+      "step": 9445
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.519049205820284e-06,
+      "loss": 0.1146,
+      "step": 9450
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.513495501499501e-06,
+      "loss": 0.1045,
+      "step": 9455
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.507941797178719e-06,
+      "loss": 0.1781,
+      "step": 9460
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.502388092857936e-06,
+      "loss": 0.0475,
+      "step": 9465
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.496834388537155e-06,
+      "loss": 0.0519,
+      "step": 9470
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.491280684216373e-06,
+      "loss": 0.0505,
+      "step": 9475
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.48572697989559e-06,
+      "loss": 0.0645,
+      "step": 9480
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.48017327557481e-06,
+      "loss": 0.0322,
+      "step": 9485
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.474619571254027e-06,
+      "loss": 0.0276,
+      "step": 9490
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.469065866933244e-06,
+      "loss": 0.0621,
+      "step": 9495
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.463512162612463e-06,
+      "loss": 0.0801,
+      "step": 9500
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.457958458291682e-06,
+      "loss": 0.1139,
+      "step": 9505
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.4524047539709e-06,
+      "loss": 0.2403,
+      "step": 9510
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.446851049650117e-06,
+      "loss": 0.0613,
+      "step": 9515
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.441297345329336e-06,
+      "loss": 0.0411,
+      "step": 9520
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.435743641008554e-06,
+      "loss": 0.0473,
+      "step": 9525
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.430189936687771e-06,
+      "loss": 0.0688,
+      "step": 9530
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.424636232366989e-06,
+      "loss": 0.0235,
+      "step": 9535
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.419082528046208e-06,
+      "loss": 0.0447,
+      "step": 9540
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.413528823725425e-06,
+      "loss": 0.0735,
+      "step": 9545
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.407975119404642e-06,
+      "loss": 0.0568,
+      "step": 9550
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.402421415083862e-06,
+      "loss": 0.1302,
+      "step": 9555
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.39686771076308e-06,
+      "loss": 0.2016,
+      "step": 9560
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.391314006442298e-06,
+      "loss": 0.0482,
+      "step": 9565
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.385760302121516e-06,
+      "loss": 0.0516,
+      "step": 9570
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.380206597800735e-06,
+      "loss": 0.069,
+      "step": 9575
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.374652893479952e-06,
+      "loss": 0.0547,
+      "step": 9580
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.36909918915917e-06,
+      "loss": 0.0432,
+      "step": 9585
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.363545484838389e-06,
+      "loss": 0.0303,
+      "step": 9590
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.357991780517606e-06,
+      "loss": 0.0698,
+      "step": 9595
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.352438076196823e-06,
+      "loss": 0.1073,
+      "step": 9600
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.34688437187604e-06,
+      "loss": 0.2612,
+      "step": 9605
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.34133066755526e-06,
+      "loss": 0.2343,
+      "step": 9610
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.335776963234479e-06,
+      "loss": 0.0467,
+      "step": 9615
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.330223258913696e-06,
+      "loss": 0.066,
+      "step": 9620
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.324669554592914e-06,
+      "loss": 0.065,
+      "step": 9625
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.319115850272133e-06,
+      "loss": 0.0687,
+      "step": 9630
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.31356214595135e-06,
+      "loss": 0.0274,
+      "step": 9635
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.308008441630568e-06,
+      "loss": 0.0428,
+      "step": 9640
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.302454737309787e-06,
+      "loss": 0.0522,
+      "step": 9645
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.296901032989004e-06,
+      "loss": 0.0975,
+      "step": 9650
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.291347328668222e-06,
+      "loss": 0.1279,
+      "step": 9655
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.286904365211596e-06,
+      "loss": 0.349,
+      "step": 9660
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.281350660890815e-06,
+      "loss": 0.0397,
+      "step": 9665
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.275796956570033e-06,
+      "loss": 0.0422,
+      "step": 9670
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.270243252249252e-06,
+      "loss": 0.0728,
+      "step": 9675
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.264689547928469e-06,
+      "loss": 0.0686,
+      "step": 9680
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.259135843607687e-06,
+      "loss": 0.0351,
+      "step": 9685
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.253582139286906e-06,
+      "loss": 0.0441,
+      "step": 9690
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.248028434966123e-06,
+      "loss": 0.0521,
+      "step": 9695
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.24247473064534e-06,
+      "loss": 0.1045,
+      "step": 9700
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.23692102632456e-06,
+      "loss": 0.1269,
+      "step": 9705
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.231367322003777e-06,
+      "loss": 0.1917,
+      "step": 9710
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.225813617682994e-06,
+      "loss": 0.055,
+      "step": 9715
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.220259913362214e-06,
+      "loss": 0.0511,
+      "step": 9720
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.214706209041433e-06,
+      "loss": 0.05,
+      "step": 9725
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.20915250472065e-06,
+      "loss": 0.0545,
+      "step": 9730
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.203598800399867e-06,
+      "loss": 0.0298,
+      "step": 9735
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.198045096079085e-06,
+      "loss": 0.04,
+      "step": 9740
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.192491391758304e-06,
+      "loss": 0.0437,
+      "step": 9745
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.186937687437521e-06,
+      "loss": 0.0737,
+      "step": 9750
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.181383983116739e-06,
+      "loss": 0.1467,
+      "step": 9755
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.175830278795958e-06,
+      "loss": 0.1908,
+      "step": 9760
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.170276574475175e-06,
+      "loss": 0.04,
+      "step": 9765
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.164722870154393e-06,
+      "loss": 0.0577,
+      "step": 9770
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.159169165833612e-06,
+      "loss": 0.0421,
+      "step": 9775
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.153615461512831e-06,
+      "loss": 0.0639,
+      "step": 9780
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.148061757192048e-06,
+      "loss": 0.0347,
+      "step": 9785
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.142508052871266e-06,
+      "loss": 0.0356,
+      "step": 9790
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.136954348550485e-06,
+      "loss": 0.0634,
+      "step": 9795
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.131400644229702e-06,
+      "loss": 0.0863,
+      "step": 9800
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.12584693990892e-06,
+      "loss": 0.1212,
+      "step": 9805
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.120293235588137e-06,
+      "loss": 0.2039,
+      "step": 9810
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.114739531267356e-06,
+      "loss": 0.0492,
+      "step": 9815
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.109185826946574e-06,
+      "loss": 0.0367,
+      "step": 9820
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.103632122625791e-06,
+      "loss": 0.0523,
+      "step": 9825
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.09807841830501e-06,
+      "loss": 0.078,
+      "step": 9830
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.09252471398423e-06,
+      "loss": 0.0229,
+      "step": 9835
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.086971009663447e-06,
+      "loss": 0.0521,
+      "step": 9840
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.081417305342664e-06,
+      "loss": 0.1158,
+      "step": 9845
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.075863601021883e-06,
+      "loss": 0.0853,
+      "step": 9850
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.0703098967011e-06,
+      "loss": 0.1259,
+      "step": 9855
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.064756192380318e-06,
+      "loss": 0.1677,
+      "step": 9860
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.059202488059536e-06,
+      "loss": 0.0502,
+      "step": 9865
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.053648783738755e-06,
+      "loss": 0.0507,
+      "step": 9870
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.048095079417972e-06,
+      "loss": 0.0542,
+      "step": 9875
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.04254137509719e-06,
+      "loss": 0.0527,
+      "step": 9880
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.036987670776409e-06,
+      "loss": 0.0189,
+      "step": 9885
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.031433966455626e-06,
+      "loss": 0.0314,
+      "step": 9890
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.025880262134845e-06,
+      "loss": 0.0521,
+      "step": 9895
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.020326557814063e-06,
+      "loss": 0.1058,
+      "step": 9900
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.014772853493282e-06,
+      "loss": 0.0902,
+      "step": 9905
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.009219149172499e-06,
+      "loss": 0.2431,
+      "step": 9910
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.003665444851716e-06,
+      "loss": 0.0461,
+      "step": 9915
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 8.998111740530936e-06,
+      "loss": 0.0444,
+      "step": 9920
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 8.992558036210153e-06,
+      "loss": 0.0477,
+      "step": 9925
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 8.98700433188937e-06,
+      "loss": 0.0571,
+      "step": 9930
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.981450627568588e-06,
+      "loss": 0.0187,
+      "step": 9935
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.975896923247807e-06,
+      "loss": 0.0367,
+      "step": 9940
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.970343218927024e-06,
+      "loss": 0.0646,
+      "step": 9945
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.964789514606243e-06,
+      "loss": 0.0816,
+      "step": 9950
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.959235810285461e-06,
+      "loss": 0.1224,
+      "step": 9955
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.95368210596468e-06,
+      "loss": 0.213,
+      "step": 9960
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.948128401643897e-06,
+      "loss": 0.0452,
+      "step": 9965
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.942574697323115e-06,
+      "loss": 0.0515,
+      "step": 9970
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.937020993002334e-06,
+      "loss": 0.0666,
+      "step": 9975
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.931467288681551e-06,
+      "loss": 0.0711,
+      "step": 9980
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.925913584360769e-06,
+      "loss": 0.0262,
+      "step": 9985
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.920359880039988e-06,
+      "loss": 0.0458,
+      "step": 9990
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.914806175719205e-06,
+      "loss": 0.0612,
+      "step": 9995
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.909252471398423e-06,
+      "loss": 0.0612,
+      "step": 10000
+    },
+    {
+      "epoch": 2.22,
+      "eval_loss": 0.02773973159492016,
+      "eval_runtime": 1637.5521,
+      "eval_samples_per_second": 9.782,
+      "eval_steps_per_second": 2.446,
+      "eval_wer": 0.05380454381536243,
+      "step": 10000
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.903698767077642e-06,
+      "loss": 0.1273,
+      "step": 10005
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.89814506275686e-06,
+      "loss": 0.2514,
+      "step": 10010
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.892591358436078e-06,
+      "loss": 0.042,
+      "step": 10015
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 8.887037654115296e-06,
+      "loss": 0.046,
+      "step": 10020
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.881483949794513e-06,
+      "loss": 0.0349,
+      "step": 10025
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.875930245473732e-06,
+      "loss": 0.0712,
+      "step": 10030
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.87037654115295e-06,
+      "loss": 0.0255,
+      "step": 10035
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.864822836832167e-06,
+      "loss": 0.0399,
+      "step": 10040
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.859269132511386e-06,
+      "loss": 0.0587,
+      "step": 10045
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.853715428190604e-06,
+      "loss": 0.0592,
+      "step": 10050
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.848161723869821e-06,
+      "loss": 0.1063,
+      "step": 10055
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.84260801954904e-06,
+      "loss": 0.1707,
+      "step": 10060
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 8.837054315228258e-06,
+      "loss": 0.0568,
+      "step": 10065
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.831500610907477e-06,
+      "loss": 0.0598,
+      "step": 10070
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.825946906586694e-06,
+      "loss": 0.0522,
+      "step": 10075
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.820393202265912e-06,
+      "loss": 0.0673,
+      "step": 10080
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.81483949794513e-06,
+      "loss": 0.0369,
+      "step": 10085
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.809285793624348e-06,
+      "loss": 0.113,
+      "step": 10090
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.803732089303566e-06,
+      "loss": 0.058,
+      "step": 10095
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.798178384982785e-06,
+      "loss": 0.1069,
+      "step": 10100
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.792624680662002e-06,
+      "loss": 0.1291,
+      "step": 10105
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8.78707097634122e-06,
+      "loss": 0.2198,
+      "step": 10110
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.781517272020439e-06,
+      "loss": 0.049,
+      "step": 10115
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.775963567699658e-06,
+      "loss": 0.0528,
+      "step": 10120
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.770409863378875e-06,
+      "loss": 0.0516,
+      "step": 10125
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.764856159058093e-06,
+      "loss": 0.0529,
+      "step": 10130
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.75930245473731e-06,
+      "loss": 0.0327,
+      "step": 10135
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.753748750416529e-06,
+      "loss": 0.0571,
+      "step": 10140
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.748195046095746e-06,
+      "loss": 0.0464,
+      "step": 10145
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.742641341774964e-06,
+      "loss": 0.0832,
+      "step": 10150
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.737087637454183e-06,
+      "loss": 0.161,
+      "step": 10155
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.7315339331334e-06,
+      "loss": 0.1703,
+      "step": 10160
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.725980228812618e-06,
+      "loss": 0.0413,
+      "step": 10165
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.720426524491837e-06,
+      "loss": 0.0399,
+      "step": 10170
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.714872820171056e-06,
+      "loss": 0.049,
+      "step": 10175
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.709319115850273e-06,
+      "loss": 0.063,
+      "step": 10180
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.703765411529491e-06,
+      "loss": 0.0342,
+      "step": 10185
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.698211707208708e-06,
+      "loss": 0.0499,
+      "step": 10190
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.692658002887927e-06,
+      "loss": 0.0414,
+      "step": 10195
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 8.687104298567145e-06,
+      "loss": 0.0906,
+      "step": 10200
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.681550594246362e-06,
+      "loss": 0.1534,
+      "step": 10205
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.675996889925581e-06,
+      "loss": 0.2036,
+      "step": 10210
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.670443185604799e-06,
+      "loss": 0.047,
+      "step": 10215
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.664889481284016e-06,
+      "loss": 0.0445,
+      "step": 10220
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.659335776963235e-06,
+      "loss": 0.0565,
+      "step": 10225
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.653782072642454e-06,
+      "loss": 0.0916,
+      "step": 10230
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.648228368321672e-06,
+      "loss": 0.0206,
+      "step": 10235
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.64267466400089e-06,
+      "loss": 0.0521,
+      "step": 10240
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.637120959680108e-06,
+      "loss": 0.0541,
+      "step": 10245
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.631567255359326e-06,
+      "loss": 0.088,
+      "step": 10250
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.626013551038543e-06,
+      "loss": 0.163,
+      "step": 10255
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.62045984671776e-06,
+      "loss": 0.2409,
+      "step": 10260
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.61490614239698e-06,
+      "loss": 0.0452,
+      "step": 10265
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.609352438076197e-06,
+      "loss": 0.0558,
+      "step": 10270
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.603798733755415e-06,
+      "loss": 0.054,
+      "step": 10275
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.598245029434634e-06,
+      "loss": 0.0647,
+      "step": 10280
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.592691325113853e-06,
+      "loss": 0.029,
+      "step": 10285
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 8.58713762079307e-06,
+      "loss": 0.0239,
+      "step": 10290
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.581583916472288e-06,
+      "loss": 0.0507,
+      "step": 10295
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.576030212151507e-06,
+      "loss": 0.0889,
+      "step": 10300
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.570476507830724e-06,
+      "loss": 0.139,
+      "step": 10305
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.564922803509942e-06,
+      "loss": 0.1565,
+      "step": 10310
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.55936909918916e-06,
+      "loss": 0.0418,
+      "step": 10315
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.553815394868378e-06,
+      "loss": 0.0482,
+      "step": 10320
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.548261690547595e-06,
+      "loss": 0.0561,
+      "step": 10325
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.542707986226813e-06,
+      "loss": 0.0641,
+      "step": 10330
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.537154281906032e-06,
+      "loss": 0.0298,
+      "step": 10335
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.53160057758525e-06,
+      "loss": 0.0291,
+      "step": 10340
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.526046873264469e-06,
+      "loss": 0.0563,
+      "step": 10345
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.520493168943686e-06,
+      "loss": 0.0674,
+      "step": 10350
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.514939464622905e-06,
+      "loss": 0.1234,
+      "step": 10355
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.509385760302122e-06,
+      "loss": 0.1679,
+      "step": 10360
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.50383205598134e-06,
+      "loss": 0.0376,
+      "step": 10365
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.498278351660559e-06,
+      "loss": 0.0463,
+      "step": 10370
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.492724647339776e-06,
+      "loss": 0.0614,
+      "step": 10375
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.487170943018994e-06,
+      "loss": 0.0663,
+      "step": 10380
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.481617238698211e-06,
+      "loss": 0.0269,
+      "step": 10385
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.47606353437743e-06,
+      "loss": 0.0224,
+      "step": 10390
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.470509830056648e-06,
+      "loss": 0.0463,
+      "step": 10395
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.464956125735867e-06,
+      "loss": 0.1134,
+      "step": 10400
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.459402421415084e-06,
+      "loss": 0.1399,
+      "step": 10405
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.453848717094303e-06,
+      "loss": 0.2287,
+      "step": 10410
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.44829501277352e-06,
+      "loss": 0.0361,
+      "step": 10415
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.442741308452738e-06,
+      "loss": 0.0417,
+      "step": 10420
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 8.437187604131957e-06,
+      "loss": 0.0443,
+      "step": 10425
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.431633899811175e-06,
+      "loss": 0.0672,
+      "step": 10430
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.426080195490392e-06,
+      "loss": 0.0317,
+      "step": 10435
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.420526491169611e-06,
+      "loss": 0.0485,
+      "step": 10440
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.414972786848829e-06,
+      "loss": 0.0745,
+      "step": 10445
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.409419082528046e-06,
+      "loss": 0.0951,
+      "step": 10450
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.403865378207265e-06,
+      "loss": 0.1212,
+      "step": 10455
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.398311673886483e-06,
+      "loss": 0.1626,
+      "step": 10460
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.392757969565702e-06,
+      "loss": 0.0449,
+      "step": 10465
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.38720426524492e-06,
+      "loss": 0.0472,
+      "step": 10470
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.381650560924137e-06,
+      "loss": 0.0689,
+      "step": 10475
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.376096856603356e-06,
+      "loss": 0.0681,
+      "step": 10480
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.370543152282573e-06,
+      "loss": 0.0188,
+      "step": 10485
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.36498944796179e-06,
+      "loss": 0.0307,
+      "step": 10490
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.35943574364101e-06,
+      "loss": 0.0434,
+      "step": 10495
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.353882039320227e-06,
+      "loss": 0.0561,
+      "step": 10500
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.348328334999445e-06,
+      "loss": 0.1528,
+      "step": 10505
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.342774630678664e-06,
+      "loss": 0.1664,
+      "step": 10510
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.337220926357881e-06,
+      "loss": 0.0429,
+      "step": 10515
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.3316672220371e-06,
+      "loss": 0.0351,
+      "step": 10520
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.326113517716318e-06,
+      "loss": 0.0524,
+      "step": 10525
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.320559813395535e-06,
+      "loss": 0.0614,
+      "step": 10530
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.315006109074754e-06,
+      "loss": 0.0439,
+      "step": 10535
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.309452404753972e-06,
+      "loss": 0.0545,
+      "step": 10540
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.303898700433189e-06,
+      "loss": 0.0605,
+      "step": 10545
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.298344996112408e-06,
+      "loss": 0.0902,
+      "step": 10550
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.292791291791625e-06,
+      "loss": 0.1415,
+      "step": 10555
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 8.287237587470843e-06,
+      "loss": 0.1941,
+      "step": 10560
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.281683883150062e-06,
+      "loss": 0.0451,
+      "step": 10565
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.276130178829281e-06,
+      "loss": 0.0556,
+      "step": 10570
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.270576474508499e-06,
+      "loss": 0.0365,
+      "step": 10575
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.265022770187716e-06,
+      "loss": 0.0539,
+      "step": 10580
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.259469065866933e-06,
+      "loss": 0.0229,
+      "step": 10585
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.253915361546152e-06,
+      "loss": 0.0416,
+      "step": 10590
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.24836165722537e-06,
+      "loss": 0.0719,
+      "step": 10595
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.242807952904587e-06,
+      "loss": 0.0931,
+      "step": 10600
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.237254248583806e-06,
+      "loss": 0.1698,
+      "step": 10605
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.231700544263024e-06,
+      "loss": 0.1732,
+      "step": 10610
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.226146839942241e-06,
+      "loss": 0.0571,
+      "step": 10615
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.22059313562146e-06,
+      "loss": 0.054,
+      "step": 10620
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.21503943130068e-06,
+      "loss": 0.0428,
+      "step": 10625
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.209485726979897e-06,
+      "loss": 0.057,
+      "step": 10630
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.203932022659114e-06,
+      "loss": 0.0288,
+      "step": 10635
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.198378318338332e-06,
+      "loss": 0.0465,
+      "step": 10640
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.19282461401755e-06,
+      "loss": 0.0502,
+      "step": 10645
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 8.187270909696768e-06,
+      "loss": 0.0911,
+      "step": 10650
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.181717205375986e-06,
+      "loss": 0.1291,
+      "step": 10655
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.176163501055205e-06,
+      "loss": 0.1882,
+      "step": 10660
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.170609796734422e-06,
+      "loss": 0.0516,
+      "step": 10665
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.16505609241364e-06,
+      "loss": 0.0563,
+      "step": 10670
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.159502388092859e-06,
+      "loss": 0.0516,
+      "step": 10675
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.153948683772078e-06,
+      "loss": 0.0706,
+      "step": 10680
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.148394979451295e-06,
+      "loss": 0.0356,
+      "step": 10685
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.142841275130513e-06,
+      "loss": 0.0364,
+      "step": 10690
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.137287570809732e-06,
+      "loss": 0.058,
+      "step": 10695
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.13173386648895e-06,
+      "loss": 0.0787,
+      "step": 10700
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.126180162168167e-06,
+      "loss": 0.1034,
+      "step": 10705
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.120626457847384e-06,
+      "loss": 0.1872,
+      "step": 10710
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.115072753526603e-06,
+      "loss": 0.0454,
+      "step": 10715
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.10951904920582e-06,
+      "loss": 0.0487,
+      "step": 10720
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.103965344885038e-06,
+      "loss": 0.044,
+      "step": 10725
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.098411640564257e-06,
+      "loss": 0.056,
+      "step": 10730
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.092857936243476e-06,
+      "loss": 0.0462,
+      "step": 10735
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.087304231922694e-06,
+      "loss": 0.0394,
+      "step": 10740
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.081750527601911e-06,
+      "loss": 0.0572,
+      "step": 10745
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.07619682328113e-06,
+      "loss": 0.1031,
+      "step": 10750
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.070643118960348e-06,
+      "loss": 0.1049,
+      "step": 10755
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.065089414639565e-06,
+      "loss": 0.1768,
+      "step": 10760
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.059535710318784e-06,
+      "loss": 0.0497,
+      "step": 10765
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.053982005998001e-06,
+      "loss": 0.0383,
+      "step": 10770
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.048428301677219e-06,
+      "loss": 0.0467,
+      "step": 10775
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.042874597356436e-06,
+      "loss": 0.0466,
+      "step": 10780
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 8.037320893035655e-06,
+      "loss": 0.0311,
+      "step": 10785
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 8.031767188714873e-06,
+      "loss": 0.0402,
+      "step": 10790
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 8.026213484394092e-06,
+      "loss": 0.0521,
+      "step": 10795
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 8.02065978007331e-06,
+      "loss": 0.0901,
+      "step": 10800
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 8.015106075752528e-06,
+      "loss": 0.1498,
+      "step": 10805
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 8.009552371431746e-06,
+      "loss": 0.142,
+      "step": 10810
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 8.003998667110963e-06,
+      "loss": 0.0347,
+      "step": 10815
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.998444962790182e-06,
+      "loss": 0.0435,
+      "step": 10820
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.9928912584694e-06,
+      "loss": 0.05,
+      "step": 10825
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.987337554148617e-06,
+      "loss": 0.0871,
+      "step": 10830
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.981783849827835e-06,
+      "loss": 0.0235,
+      "step": 10835
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.976230145507054e-06,
+      "loss": 0.0541,
+      "step": 10840
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.970676441186271e-06,
+      "loss": 0.0476,
+      "step": 10845
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.96512273686549e-06,
+      "loss": 0.0829,
+      "step": 10850
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.959569032544708e-06,
+      "loss": 0.0758,
+      "step": 10855
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.954015328223927e-06,
+      "loss": 0.1808,
+      "step": 10860
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.948461623903144e-06,
+      "loss": 0.0401,
+      "step": 10865
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.942907919582362e-06,
+      "loss": 0.0434,
+      "step": 10870
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.93735421526158e-06,
+      "loss": 0.0482,
+      "step": 10875
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.931800510940798e-06,
+      "loss": 0.0612,
+      "step": 10880
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.926246806620016e-06,
+      "loss": 0.0289,
+      "step": 10885
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.920693102299235e-06,
+      "loss": 0.0222,
+      "step": 10890
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.915139397978452e-06,
+      "loss": 0.0789,
+      "step": 10895
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.90958569365767e-06,
+      "loss": 0.0593,
+      "step": 10900
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.904031989336889e-06,
+      "loss": 0.126,
+      "step": 10905
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.898478285016106e-06,
+      "loss": 0.1348,
+      "step": 10910
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.892924580695325e-06,
+      "loss": 0.0529,
+      "step": 10915
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.887370876374543e-06,
+      "loss": 0.0636,
+      "step": 10920
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.88181717205376e-06,
+      "loss": 0.0531,
+      "step": 10925
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.876263467732979e-06,
+      "loss": 0.0616,
+      "step": 10930
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.870709763412197e-06,
+      "loss": 0.0175,
+      "step": 10935
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.865156059091414e-06,
+      "loss": 0.049,
+      "step": 10940
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.859602354770633e-06,
+      "loss": 0.0604,
+      "step": 10945
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.85404865044985e-06,
+      "loss": 0.0956,
+      "step": 10950
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.848494946129068e-06,
+      "loss": 0.1084,
+      "step": 10955
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.842941241808287e-06,
+      "loss": 0.2724,
+      "step": 10960
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.837387537487504e-06,
+      "loss": 0.047,
+      "step": 10965
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.831833833166724e-06,
+      "loss": 0.0537,
+      "step": 10970
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.826280128845941e-06,
+      "loss": 0.0489,
+      "step": 10975
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.820726424525158e-06,
+      "loss": 0.0595,
+      "step": 10980
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.815172720204377e-06,
+      "loss": 0.0159,
+      "step": 10985
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.809619015883595e-06,
+      "loss": 0.0375,
+      "step": 10990
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.804065311562812e-06,
+      "loss": 0.0896,
+      "step": 10995
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.798511607242031e-06,
+      "loss": 0.0686,
+      "step": 11000
+    },
+    {
+      "epoch": 2.44,
+      "eval_loss": 0.026114122942090034,
+      "eval_runtime": 1481.1638,
+      "eval_samples_per_second": 10.814,
+      "eval_steps_per_second": 2.704,
+      "eval_wer": 0.04807068157230436,
+      "step": 11000
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.792957902921249e-06,
+      "loss": 0.1083,
+      "step": 11005
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.787404198600466e-06,
+      "loss": 0.2278,
+      "step": 11010
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.781850494279685e-06,
+      "loss": 0.0392,
+      "step": 11015
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.776296789958904e-06,
+      "loss": 0.0414,
+      "step": 11020
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.770743085638122e-06,
+      "loss": 0.0683,
+      "step": 11025
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.76518938131734e-06,
+      "loss": 0.0561,
+      "step": 11030
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.759635676996557e-06,
+      "loss": 0.0243,
+      "step": 11035
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.754081972675776e-06,
+      "loss": 0.0478,
+      "step": 11040
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.748528268354993e-06,
+      "loss": 0.0935,
+      "step": 11045
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.74297456403421e-06,
+      "loss": 0.0897,
+      "step": 11050
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.73742085971343e-06,
+      "loss": 0.1266,
+      "step": 11055
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.731867155392647e-06,
+      "loss": 0.1696,
+      "step": 11060
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.726313451071865e-06,
+      "loss": 0.048,
+      "step": 11065
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.720759746751084e-06,
+      "loss": 0.0464,
+      "step": 11070
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.715206042430303e-06,
+      "loss": 0.0628,
+      "step": 11075
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.70965233810952e-06,
+      "loss": 0.0685,
+      "step": 11080
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.704098633788738e-06,
+      "loss": 0.0473,
+      "step": 11085
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.698544929467957e-06,
+      "loss": 0.0387,
+      "step": 11090
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.692991225147174e-06,
+      "loss": 0.0516,
+      "step": 11095
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.687437520826392e-06,
+      "loss": 0.0742,
+      "step": 11100
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.681883816505609e-06,
+      "loss": 0.1191,
+      "step": 11105
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.676330112184828e-06,
+      "loss": 0.1816,
+      "step": 11110
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.670776407864046e-06,
+      "loss": 0.0489,
+      "step": 11115
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.665222703543263e-06,
+      "loss": 0.0539,
+      "step": 11120
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.659668999222482e-06,
+      "loss": 0.0513,
+      "step": 11125
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.654115294901701e-06,
+      "loss": 0.0669,
+      "step": 11130
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.648561590580919e-06,
+      "loss": 0.0365,
+      "step": 11135
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.643007886260136e-06,
+      "loss": 0.0347,
+      "step": 11140
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.637454181939355e-06,
+      "loss": 0.0705,
+      "step": 11145
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.631900477618573e-06,
+      "loss": 0.0786,
+      "step": 11150
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.62634677329779e-06,
+      "loss": 0.1167,
+      "step": 11155
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.620793068977008e-06,
+      "loss": 0.1719,
+      "step": 11160
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.6152393646562265e-06,
+      "loss": 0.0484,
+      "step": 11165
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.609685660335444e-06,
+      "loss": 0.051,
+      "step": 11170
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.604131956014662e-06,
+      "loss": 0.0457,
+      "step": 11175
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.59857825169388e-06,
+      "loss": 0.0676,
+      "step": 11180
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.593024547373099e-06,
+      "loss": 0.0212,
+      "step": 11185
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.587470843052317e-06,
+      "loss": 0.0362,
+      "step": 11190
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.581917138731535e-06,
+      "loss": 0.0536,
+      "step": 11195
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.576363434410753e-06,
+      "loss": 0.1078,
+      "step": 11200
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.570809730089971e-06,
+      "loss": 0.1361,
+      "step": 11205
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.565256025769188e-06,
+      "loss": 0.1706,
+      "step": 11210
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.559702321448407e-06,
+      "loss": 0.0384,
+      "step": 11215
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.554148617127625e-06,
+      "loss": 0.0547,
+      "step": 11220
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.548594912806842e-06,
+      "loss": 0.0447,
+      "step": 11225
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.5430412084860606e-06,
+      "loss": 0.0568,
+      "step": 11230
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.537487504165279e-06,
+      "loss": 0.0343,
+      "step": 11235
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.531933799844496e-06,
+      "loss": 0.0596,
+      "step": 11240
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.526380095523715e-06,
+      "loss": 0.0621,
+      "step": 11245
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.520826391202934e-06,
+      "loss": 0.0856,
+      "step": 11250
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.515272686882151e-06,
+      "loss": 0.1325,
+      "step": 11255
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.509718982561369e-06,
+      "loss": 0.1926,
+      "step": 11260
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.5041652782405876e-06,
+      "loss": 0.0331,
+      "step": 11265
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.498611573919805e-06,
+      "loss": 0.0486,
+      "step": 11270
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.493057869599023e-06,
+      "loss": 0.05,
+      "step": 11275
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.487504165278241e-06,
+      "loss": 0.0651,
+      "step": 11280
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.481950460957459e-06,
+      "loss": 0.0367,
+      "step": 11285
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.476396756636677e-06,
+      "loss": 0.1167,
+      "step": 11290
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.470843052315895e-06,
+      "loss": 0.0529,
+      "step": 11295
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.465289347995114e-06,
+      "loss": 0.06,
+      "step": 11300
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.459735643674332e-06,
+      "loss": 0.0941,
+      "step": 11305
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.454181939353549e-06,
+      "loss": 0.1887,
+      "step": 11310
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.448628235032768e-06,
+      "loss": 0.0478,
+      "step": 11315
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.443074530711986e-06,
+      "loss": 0.0515,
+      "step": 11320
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.437520826391203e-06,
+      "loss": 0.0532,
+      "step": 11325
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.431967122070422e-06,
+      "loss": 0.0595,
+      "step": 11330
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.426413417749639e-06,
+      "loss": 0.0399,
+      "step": 11335
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.420859713428857e-06,
+      "loss": 0.0214,
+      "step": 11340
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.4153060091080755e-06,
+      "loss": 0.0745,
+      "step": 11345
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.409752304787293e-06,
+      "loss": 0.0458,
+      "step": 11350
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.404198600466512e-06,
+      "loss": 0.0899,
+      "step": 11355
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.39864489614573e-06,
+      "loss": 0.1467,
+      "step": 11360
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.393091191824948e-06,
+      "loss": 0.0292,
+      "step": 11365
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.387537487504166e-06,
+      "loss": 0.0397,
+      "step": 11370
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.381983783183384e-06,
+      "loss": 0.0474,
+      "step": 11375
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.376430078862602e-06,
+      "loss": 0.1507,
+      "step": 11380
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.37087637454182e-06,
+      "loss": 0.025,
+      "step": 11385
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.365322670221038e-06,
+      "loss": 0.0246,
+      "step": 11390
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.359768965900256e-06,
+      "loss": 0.0494,
+      "step": 11395
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.354215261579474e-06,
+      "loss": 0.088,
+      "step": 11400
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.348661557258691e-06,
+      "loss": 0.1434,
+      "step": 11405
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.34310785293791e-06,
+      "loss": 0.2171,
+      "step": 11410
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.337554148617129e-06,
+      "loss": 0.0426,
+      "step": 11415
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.332000444296347e-06,
+      "loss": 0.0583,
+      "step": 11420
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.326446739975564e-06,
+      "loss": 0.0489,
+      "step": 11425
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.320893035654783e-06,
+      "loss": 0.0534,
+      "step": 11430
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.315339331334e-06,
+      "loss": 0.0313,
+      "step": 11435
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.309785627013218e-06,
+      "loss": 0.0626,
+      "step": 11440
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.3042319226924366e-06,
+      "loss": 0.0408,
+      "step": 11445
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.298678218371654e-06,
+      "loss": 0.0845,
+      "step": 11450
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.293124514050872e-06,
+      "loss": 0.1346,
+      "step": 11455
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.2875708097300905e-06,
+      "loss": 0.1647,
+      "step": 11460
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.282017105409308e-06,
+      "loss": 0.0505,
+      "step": 11465
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.276463401088527e-06,
+      "loss": 0.0432,
+      "step": 11470
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.270909696767745e-06,
+      "loss": 0.0581,
+      "step": 11475
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.265355992446963e-06,
+      "loss": 0.0727,
+      "step": 11480
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.259802288126181e-06,
+      "loss": 0.0276,
+      "step": 11485
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.254248583805399e-06,
+      "loss": 0.0371,
+      "step": 11490
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.248694879484617e-06,
+      "loss": 0.0593,
+      "step": 11495
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.243141175163835e-06,
+      "loss": 0.0982,
+      "step": 11500
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.237587470843052e-06,
+      "loss": 0.105,
+      "step": 11505
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.232033766522271e-06,
+      "loss": 0.1472,
+      "step": 11510
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.226480062201489e-06,
+      "loss": 0.047,
+      "step": 11515
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.220926357880706e-06,
+      "loss": 0.0393,
+      "step": 11520
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.215372653559925e-06,
+      "loss": 0.0521,
+      "step": 11525
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.209818949239144e-06,
+      "loss": 0.058,
+      "step": 11530
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.204265244918361e-06,
+      "loss": 0.0328,
+      "step": 11535
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.198711540597579e-06,
+      "loss": 0.028,
+      "step": 11540
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.193157836276798e-06,
+      "loss": 0.054,
+      "step": 11545
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.187604131956015e-06,
+      "loss": 0.0759,
+      "step": 11550
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.182050427635233e-06,
+      "loss": 0.107,
+      "step": 11555
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.176496723314451e-06,
+      "loss": 0.1543,
+      "step": 11560
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.170943018993669e-06,
+      "loss": 0.0459,
+      "step": 11565
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.165389314672887e-06,
+      "loss": 0.0432,
+      "step": 11570
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.159835610352105e-06,
+      "loss": 0.0383,
+      "step": 11575
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.154281906031324e-06,
+      "loss": 0.0713,
+      "step": 11580
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.148728201710542e-06,
+      "loss": 0.0327,
+      "step": 11585
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.14317449738976e-06,
+      "loss": 0.0478,
+      "step": 11590
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.137620793068978e-06,
+      "loss": 0.0472,
+      "step": 11595
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.132067088748196e-06,
+      "loss": 0.078,
+      "step": 11600
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.126513384427413e-06,
+      "loss": 0.147,
+      "step": 11605
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.120959680106632e-06,
+      "loss": 0.1828,
+      "step": 11610
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.11540597578585e-06,
+      "loss": 0.0425,
+      "step": 11615
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.109852271465067e-06,
+      "loss": 0.0376,
+      "step": 11620
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.104298567144286e-06,
+      "loss": 0.0532,
+      "step": 11625
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.098744862823503e-06,
+      "loss": 0.0723,
+      "step": 11630
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.093191158502722e-06,
+      "loss": 0.0256,
+      "step": 11635
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.08763745418194e-06,
+      "loss": 0.0768,
+      "step": 11640
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.082083749861159e-06,
+      "loss": 0.0618,
+      "step": 11645
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.076530045540376e-06,
+      "loss": 0.0802,
+      "step": 11650
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.070976341219594e-06,
+      "loss": 0.1172,
+      "step": 11655
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.065422636898812e-06,
+      "loss": 0.1576,
+      "step": 11660
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.05986893257803e-06,
+      "loss": 0.0512,
+      "step": 11665
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.054315228257248e-06,
+      "loss": 0.0426,
+      "step": 11670
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.048761523936466e-06,
+      "loss": 0.0546,
+      "step": 11675
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.043207819615684e-06,
+      "loss": 0.0572,
+      "step": 11680
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.037654115294902e-06,
+      "loss": 0.0237,
+      "step": 11685
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.03210041097412e-06,
+      "loss": 0.0461,
+      "step": 11690
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.026546706653339e-06,
+      "loss": 0.0456,
+      "step": 11695
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.020993002332557e-06,
+      "loss": 0.1001,
+      "step": 11700
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.0154392980117744e-06,
+      "loss": 0.1546,
+      "step": 11705
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.009885593690993e-06,
+      "loss": 0.2307,
+      "step": 11710
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.004331889370211e-06,
+      "loss": 0.0457,
+      "step": 11715
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 6.998778185049428e-06,
+      "loss": 0.0499,
+      "step": 11720
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 6.993224480728647e-06,
+      "loss": 0.044,
+      "step": 11725
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 6.987670776407864e-06,
+      "loss": 0.0576,
+      "step": 11730
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.982117072087082e-06,
+      "loss": 0.0265,
+      "step": 11735
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.976563367766301e-06,
+      "loss": 0.0591,
+      "step": 11740
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.971009663445518e-06,
+      "loss": 0.0447,
+      "step": 11745
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.965455959124737e-06,
+      "loss": 0.0919,
+      "step": 11750
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.959902254803955e-06,
+      "loss": 0.1142,
+      "step": 11755
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.954348550483173e-06,
+      "loss": 0.2168,
+      "step": 11760
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.948794846162391e-06,
+      "loss": 0.0416,
+      "step": 11765
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.943241141841609e-06,
+      "loss": 0.045,
+      "step": 11770
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.937687437520827e-06,
+      "loss": 0.0724,
+      "step": 11775
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.932133733200045e-06,
+      "loss": 0.0409,
+      "step": 11780
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.926580028879263e-06,
+      "loss": 0.031,
+      "step": 11785
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.921026324558481e-06,
+      "loss": 0.045,
+      "step": 11790
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.915472620237699e-06,
+      "loss": 0.0738,
+      "step": 11795
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.909918915916916e-06,
+      "loss": 0.0919,
+      "step": 11800
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.9043652115961355e-06,
+      "loss": 0.1124,
+      "step": 11805
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.898811507275354e-06,
+      "loss": 0.1611,
+      "step": 11810
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.893257802954572e-06,
+      "loss": 0.052,
+      "step": 11815
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.887704098633789e-06,
+      "loss": 0.0354,
+      "step": 11820
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.882150394313008e-06,
+      "loss": 0.0499,
+      "step": 11825
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.876596689992225e-06,
+      "loss": 0.0487,
+      "step": 11830
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.871042985671443e-06,
+      "loss": 0.0174,
+      "step": 11835
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.865489281350662e-06,
+      "loss": 0.0444,
+      "step": 11840
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.859935577029879e-06,
+      "loss": 0.0621,
+      "step": 11845
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.854381872709097e-06,
+      "loss": 0.0897,
+      "step": 11850
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.848828168388315e-06,
+      "loss": 0.1044,
+      "step": 11855
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.843274464067534e-06,
+      "loss": 0.1726,
+      "step": 11860
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.837720759746752e-06,
+      "loss": 0.0419,
+      "step": 11865
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.83216705542597e-06,
+      "loss": 0.0485,
+      "step": 11870
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.826613351105188e-06,
+      "loss": 0.045,
+      "step": 11875
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.821059646784406e-06,
+      "loss": 0.0468,
+      "step": 11880
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.8155059424636234e-06,
+      "loss": 0.0216,
+      "step": 11885
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.809952238142842e-06,
+      "loss": 0.0244,
+      "step": 11890
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.80439853382206e-06,
+      "loss": 0.0349,
+      "step": 11895
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.798844829501277e-06,
+      "loss": 0.0525,
+      "step": 11900
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.793291125180496e-06,
+      "loss": 0.121,
+      "step": 11905
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.787737420859714e-06,
+      "loss": 0.1991,
+      "step": 11910
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.782183716538931e-06,
+      "loss": 0.0461,
+      "step": 11915
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.7766300122181504e-06,
+      "loss": 0.0558,
+      "step": 11920
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.771076307897369e-06,
+      "loss": 0.0556,
+      "step": 11925
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.765522603576586e-06,
+      "loss": 0.0587,
+      "step": 11930
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.759968899255804e-06,
+      "loss": 0.0224,
+      "step": 11935
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.754415194935023e-06,
+      "loss": 0.0374,
+      "step": 11940
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.74886149061424e-06,
+      "loss": 0.0633,
+      "step": 11945
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.743307786293458e-06,
+      "loss": 0.0754,
+      "step": 11950
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.737754081972676e-06,
+      "loss": 0.112,
+      "step": 11955
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.732200377651894e-06,
+      "loss": 0.1445,
+      "step": 11960
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.726646673331112e-06,
+      "loss": 0.0516,
+      "step": 11965
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.72109296901033e-06,
+      "loss": 0.0642,
+      "step": 11970
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.715539264689549e-06,
+      "loss": 0.0505,
+      "step": 11975
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.709985560368767e-06,
+      "loss": 0.0666,
+      "step": 11980
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.7044318560479845e-06,
+      "loss": 0.0219,
+      "step": 11985
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.698878151727203e-06,
+      "loss": 0.0374,
+      "step": 11990
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.693324447406421e-06,
+      "loss": 0.054,
+      "step": 11995
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.6877707430856384e-06,
+      "loss": 0.0512,
+      "step": 12000
+    },
+    {
+      "epoch": 2.66,
+      "eval_loss": 0.027297571301460266,
+      "eval_runtime": 1504.2471,
+      "eval_samples_per_second": 10.649,
+      "eval_steps_per_second": 2.662,
+      "eval_wer": 0.04886404615939416,
+      "step": 12000
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.682217038764857e-06,
+      "loss": 0.0947,
+      "step": 12005
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.676663334444075e-06,
+      "loss": 0.2441,
+      "step": 12010
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.671109630123292e-06,
+      "loss": 0.0436,
+      "step": 12015
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.665555925802511e-06,
+      "loss": 0.0399,
+      "step": 12020
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.660002221481728e-06,
+      "loss": 0.043,
+      "step": 12025
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.654448517160947e-06,
+      "loss": 0.0726,
+      "step": 12030
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.6488948128401654e-06,
+      "loss": 0.0179,
+      "step": 12035
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.643341108519384e-06,
+      "loss": 0.0335,
+      "step": 12040
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.637787404198601e-06,
+      "loss": 0.0481,
+      "step": 12045
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.632233699877819e-06,
+      "loss": 0.0972,
+      "step": 12050
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.626679995557037e-06,
+      "loss": 0.1203,
+      "step": 12055
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.621126291236255e-06,
+      "loss": 0.1635,
+      "step": 12060
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.615572586915473e-06,
+      "loss": 0.0444,
+      "step": 12065
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.610018882594691e-06,
+      "loss": 0.0547,
+      "step": 12070
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.604465178273909e-06,
+      "loss": 0.0463,
+      "step": 12075
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.598911473953126e-06,
+      "loss": 0.0695,
+      "step": 12080
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.5933577696323455e-06,
+      "loss": 0.03,
+      "step": 12085
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 6.587804065311564e-06,
+      "loss": 0.0437,
+      "step": 12090
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.582250360990782e-06,
+      "loss": 0.093,
+      "step": 12095
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.5766966566699995e-06,
+      "loss": 0.1169,
+      "step": 12100
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.571142952349218e-06,
+      "loss": 0.1297,
+      "step": 12105
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.565589248028435e-06,
+      "loss": 0.1731,
+      "step": 12110
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.560035543707653e-06,
+      "loss": 0.0401,
+      "step": 12115
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.554481839386872e-06,
+      "loss": 0.052,
+      "step": 12120
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.548928135066089e-06,
+      "loss": 0.0621,
+      "step": 12125
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.543374430745307e-06,
+      "loss": 0.0597,
+      "step": 12130
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 6.537820726424526e-06,
+      "loss": 0.024,
+      "step": 12135
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.532267022103743e-06,
+      "loss": 0.0339,
+      "step": 12140
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.526713317782962e-06,
+      "loss": 0.0353,
+      "step": 12145
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.52115961346218e-06,
+      "loss": 0.0829,
+      "step": 12150
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.515605909141398e-06,
+      "loss": 0.1528,
+      "step": 12155
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.510052204820616e-06,
+      "loss": 0.17,
+      "step": 12160
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.504498500499834e-06,
+      "loss": 0.0466,
+      "step": 12165
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.498944796179052e-06,
+      "loss": 0.0395,
+      "step": 12170
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.49339109185827e-06,
+      "loss": 0.054,
+      "step": 12175
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 6.4878373875374875e-06,
+      "loss": 0.0791,
+      "step": 12180
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.482283683216706e-06,
+      "loss": 0.0365,
+      "step": 12185
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.476729978895924e-06,
+      "loss": 0.0325,
+      "step": 12190
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.471176274575141e-06,
+      "loss": 0.0417,
+      "step": 12195
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.4656225702543605e-06,
+      "loss": 0.0606,
+      "step": 12200
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.460068865933579e-06,
+      "loss": 0.1526,
+      "step": 12205
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.454515161612796e-06,
+      "loss": 0.1892,
+      "step": 12210
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.4489614572920145e-06,
+      "loss": 0.0338,
+      "step": 12215
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.443407752971233e-06,
+      "loss": 0.0417,
+      "step": 12220
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 6.43785404865045e-06,
+      "loss": 0.044,
+      "step": 12225
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.432300344329668e-06,
+      "loss": 0.0543,
+      "step": 12230
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.426746640008887e-06,
+      "loss": 0.0199,
+      "step": 12235
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.421192935688104e-06,
+      "loss": 0.0309,
+      "step": 12240
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.415639231367322e-06,
+      "loss": 0.0364,
+      "step": 12245
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.41008552704654e-06,
+      "loss": 0.065,
+      "step": 12250
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.404531822725759e-06,
+      "loss": 0.1225,
+      "step": 12255
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.398978118404977e-06,
+      "loss": 0.1474,
+      "step": 12260
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.393424414084195e-06,
+      "loss": 0.0441,
+      "step": 12265
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 6.387870709763413e-06,
+      "loss": 0.0497,
+      "step": 12270
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.382317005442631e-06,
+      "loss": 0.0622,
+      "step": 12275
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.3767633011218485e-06,
+      "loss": 0.0577,
+      "step": 12280
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.371209596801067e-06,
+      "loss": 0.0254,
+      "step": 12285
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.365655892480285e-06,
+      "loss": 0.0492,
+      "step": 12290
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.3601021881595024e-06,
+      "loss": 0.0552,
+      "step": 12295
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.354548483838721e-06,
+      "loss": 0.0734,
+      "step": 12300
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.348994779517938e-06,
+      "loss": 0.1585,
+      "step": 12305
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.343441075197157e-06,
+      "loss": 0.1431,
+      "step": 12310
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 6.3378873708763755e-06,
+      "loss": 0.0379,
+      "step": 12315
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.332333666555594e-06,
+      "loss": 0.0527,
+      "step": 12320
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.326779962234811e-06,
+      "loss": 0.0634,
+      "step": 12325
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.3212262579140294e-06,
+      "loss": 0.0696,
+      "step": 12330
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.315672553593247e-06,
+      "loss": 0.0178,
+      "step": 12335
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.310118849272465e-06,
+      "loss": 0.0509,
+      "step": 12340
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.304565144951683e-06,
+      "loss": 0.049,
+      "step": 12345
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.299011440630901e-06,
+      "loss": 0.0878,
+      "step": 12350
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.293457736310119e-06,
+      "loss": 0.1311,
+      "step": 12355
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 6.287904031989337e-06,
+      "loss": 0.2205,
+      "step": 12360
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.282350327668555e-06,
+      "loss": 0.0396,
+      "step": 12365
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.276796623347774e-06,
+      "loss": 0.0472,
+      "step": 12370
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.271242919026992e-06,
+      "loss": 0.063,
+      "step": 12375
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.2656892147062095e-06,
+      "loss": 0.0768,
+      "step": 12380
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.260135510385428e-06,
+      "loss": 0.0206,
+      "step": 12385
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.254581806064646e-06,
+      "loss": 0.043,
+      "step": 12390
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.2490281017438635e-06,
+      "loss": 0.0412,
+      "step": 12395
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.243474397423082e-06,
+      "loss": 0.0667,
+      "step": 12400
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.237920693102299e-06,
+      "loss": 0.1464,
+      "step": 12405
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.232366988781517e-06,
+      "loss": 0.1683,
+      "step": 12410
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.226813284460736e-06,
+      "loss": 0.0431,
+      "step": 12415
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.221259580139953e-06,
+      "loss": 0.0481,
+      "step": 12420
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.215705875819172e-06,
+      "loss": 0.0506,
+      "step": 12425
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.2101521714983905e-06,
+      "loss": 0.0541,
+      "step": 12430
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.204598467177608e-06,
+      "loss": 0.0131,
+      "step": 12435
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.199044762856826e-06,
+      "loss": 0.0361,
+      "step": 12440
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.193491058536044e-06,
+      "loss": 0.0694,
+      "step": 12445
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 6.187937354215262e-06,
+      "loss": 0.1017,
+      "step": 12450
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.18238364989448e-06,
+      "loss": 0.1186,
+      "step": 12455
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.176829945573698e-06,
+      "loss": 0.2217,
+      "step": 12460
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.171276241252916e-06,
+      "loss": 0.0337,
+      "step": 12465
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.165722536932134e-06,
+      "loss": 0.051,
+      "step": 12470
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.1601688326113515e-06,
+      "loss": 0.0507,
+      "step": 12475
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.1546151282905706e-06,
+      "loss": 0.0591,
+      "step": 12480
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.149061423969789e-06,
+      "loss": 0.0289,
+      "step": 12485
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.143507719649007e-06,
+      "loss": 0.0332,
+      "step": 12490
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 6.1379540153282245e-06,
+      "loss": 0.0479,
+      "step": 12495
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.132400311007443e-06,
+      "loss": 0.0733,
+      "step": 12500
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.12684660668666e-06,
+      "loss": 0.1136,
+      "step": 12505
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.1212929023658785e-06,
+      "loss": 0.1952,
+      "step": 12510
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.115739198045097e-06,
+      "loss": 0.0339,
+      "step": 12515
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.110185493724314e-06,
+      "loss": 0.0381,
+      "step": 12520
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.104631789403532e-06,
+      "loss": 0.0485,
+      "step": 12525
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.09907808508275e-06,
+      "loss": 0.0571,
+      "step": 12530
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.093524380761969e-06,
+      "loss": 0.0235,
+      "step": 12535
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 6.087970676441187e-06,
+      "loss": 0.0451,
+      "step": 12540
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.0824169721204055e-06,
+      "loss": 0.0668,
+      "step": 12545
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.076863267799623e-06,
+      "loss": 0.0908,
+      "step": 12550
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.071309563478841e-06,
+      "loss": 0.1115,
+      "step": 12555
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.065755859158059e-06,
+      "loss": 0.1646,
+      "step": 12560
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.060202154837277e-06,
+      "loss": 0.0389,
+      "step": 12565
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.054648450516495e-06,
+      "loss": 0.0438,
+      "step": 12570
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.0490947461957125e-06,
+      "loss": 0.0435,
+      "step": 12575
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.043541041874931e-06,
+      "loss": 0.0543,
+      "step": 12580
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.037987337554149e-06,
+      "loss": 0.0326,
+      "step": 12585
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 6.0324336332333664e-06,
+      "loss": 0.0364,
+      "step": 12590
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 6.0268799289125856e-06,
+      "loss": 0.0404,
+      "step": 12595
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 6.021326224591804e-06,
+      "loss": 0.0699,
+      "step": 12600
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 6.015772520271021e-06,
+      "loss": 0.0913,
+      "step": 12605
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 6.0102188159502395e-06,
+      "loss": 0.1676,
+      "step": 12610
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 6.004665111629458e-06,
+      "loss": 0.0517,
+      "step": 12615
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 5.999111407308675e-06,
+      "loss": 0.0489,
+      "step": 12620
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 5.9935577029878934e-06,
+      "loss": 0.0599,
+      "step": 12625
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 5.988003998667111e-06,
+      "loss": 0.0534,
+      "step": 12630
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.982450294346329e-06,
+      "loss": 0.033,
+      "step": 12635
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.976896590025547e-06,
+      "loss": 0.0376,
+      "step": 12640
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.971342885704765e-06,
+      "loss": 0.0595,
+      "step": 12645
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.965789181383984e-06,
+      "loss": 0.0848,
+      "step": 12650
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.960235477063202e-06,
+      "loss": 0.102,
+      "step": 12655
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.95468177274242e-06,
+      "loss": 0.1471,
+      "step": 12660
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.949128068421638e-06,
+      "loss": 0.0609,
+      "step": 12665
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.943574364100856e-06,
+      "loss": 0.0495,
+      "step": 12670
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 5.9380206597800735e-06,
+      "loss": 0.0433,
+      "step": 12675
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.932466955459292e-06,
+      "loss": 0.063,
+      "step": 12680
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.92691325113851e-06,
+      "loss": 0.0251,
+      "step": 12685
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.9213595468177275e-06,
+      "loss": 0.0174,
+      "step": 12690
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.915805842496946e-06,
+      "loss": 0.0487,
+      "step": 12695
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.910252138176163e-06,
+      "loss": 0.0875,
+      "step": 12700
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.904698433855382e-06,
+      "loss": 0.1031,
+      "step": 12705
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.8991447295346005e-06,
+      "loss": 0.1648,
+      "step": 12710
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.893591025213819e-06,
+      "loss": 0.0503,
+      "step": 12715
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 5.888037320893036e-06,
+      "loss": 0.038,
+      "step": 12720
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.8824836165722545e-06,
+      "loss": 0.0515,
+      "step": 12725
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.876929912251472e-06,
+      "loss": 0.0725,
+      "step": 12730
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.87137620793069e-06,
+      "loss": 0.024,
+      "step": 12735
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.865822503609908e-06,
+      "loss": 0.0456,
+      "step": 12740
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.860268799289126e-06,
+      "loss": 0.0348,
+      "step": 12745
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.854715094968344e-06,
+      "loss": 0.07,
+      "step": 12750
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.849161390647562e-06,
+      "loss": 0.1404,
+      "step": 12755
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.843607686326781e-06,
+      "loss": 0.1914,
+      "step": 12760
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 5.838053982005999e-06,
+      "loss": 0.0364,
+      "step": 12765
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.832500277685217e-06,
+      "loss": 0.0448,
+      "step": 12770
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.8269465733644346e-06,
+      "loss": 0.052,
+      "step": 12775
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.821392869043653e-06,
+      "loss": 0.0514,
+      "step": 12780
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.815839164722871e-06,
+      "loss": 0.0416,
+      "step": 12785
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.8102854604020885e-06,
+      "loss": 0.0452,
+      "step": 12790
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.804731756081307e-06,
+      "loss": 0.0417,
+      "step": 12795
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.799178051760524e-06,
+      "loss": 0.082,
+      "step": 12800
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.7936243474397425e-06,
+      "loss": 0.1487,
+      "step": 12805
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 5.788070643118961e-06,
+      "loss": 0.1806,
+      "step": 12810
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.782516938798178e-06,
+      "loss": 0.0392,
+      "step": 12815
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.776963234477397e-06,
+      "loss": 0.0449,
+      "step": 12820
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.7714095301566155e-06,
+      "loss": 0.0488,
+      "step": 12825
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.765855825835833e-06,
+      "loss": 0.2127,
+      "step": 12830
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.760302121515051e-06,
+      "loss": 0.0278,
+      "step": 12835
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.7547484171942695e-06,
+      "loss": 0.0365,
+      "step": 12840
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.749194712873487e-06,
+      "loss": 0.0428,
+      "step": 12845
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.743641008552705e-06,
+      "loss": 0.0754,
+      "step": 12850
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 5.7380873042319226e-06,
+      "loss": 0.1655,
+      "step": 12855
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.732533599911141e-06,
+      "loss": 0.1624,
+      "step": 12860
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.726979895590359e-06,
+      "loss": 0.0503,
+      "step": 12865
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.7214261912695765e-06,
+      "loss": 0.0594,
+      "step": 12870
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.715872486948796e-06,
+      "loss": 0.0418,
+      "step": 12875
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.710318782628014e-06,
+      "loss": 0.0566,
+      "step": 12880
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.704765078307231e-06,
+      "loss": 0.0272,
+      "step": 12885
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.6992113739864496e-06,
+      "loss": 0.0386,
+      "step": 12890
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.693657669665668e-06,
+      "loss": 0.0576,
+      "step": 12895
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 5.688103965344885e-06,
+      "loss": 0.0503,
+      "step": 12900
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.6825502610241035e-06,
+      "loss": 0.1295,
+      "step": 12905
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.676996556703322e-06,
+      "loss": 0.1865,
+      "step": 12910
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.671442852382539e-06,
+      "loss": 0.0393,
+      "step": 12915
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.6658891480617574e-06,
+      "loss": 0.0599,
+      "step": 12920
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.660335443740975e-06,
+      "loss": 0.0625,
+      "step": 12925
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.654781739420194e-06,
+      "loss": 0.0595,
+      "step": 12930
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.649228035099412e-06,
+      "loss": 0.0206,
+      "step": 12935
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.6436743307786305e-06,
+      "loss": 0.035,
+      "step": 12940
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.638120626457848e-06,
+      "loss": 0.0583,
+      "step": 12945
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.632566922137066e-06,
+      "loss": 0.0746,
+      "step": 12950
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.627013217816284e-06,
+      "loss": 0.1286,
+      "step": 12955
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.621459513495502e-06,
+      "loss": 0.1657,
+      "step": 12960
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.61590580917472e-06,
+      "loss": 0.0569,
+      "step": 12965
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.6103521048539375e-06,
+      "loss": 0.041,
+      "step": 12970
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.604798400533156e-06,
+      "loss": 0.041,
+      "step": 12975
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.599244696212374e-06,
+      "loss": 0.0747,
+      "step": 12980
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.593690991891592e-06,
+      "loss": 0.0217,
+      "step": 12985
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.588137287570811e-06,
+      "loss": 0.0485,
+      "step": 12990
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.582583583250029e-06,
+      "loss": 0.0505,
+      "step": 12995
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.577029878929246e-06,
+      "loss": 0.0787,
+      "step": 13000
+    },
+    {
+      "epoch": 2.89,
+      "eval_loss": 0.024242974817752838,
+      "eval_runtime": 1599.3201,
+      "eval_samples_per_second": 10.016,
+      "eval_steps_per_second": 2.504,
+      "eval_wer": 0.04432023079697079,
+      "step": 13000
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.5714761746084645e-06,
+      "loss": 0.1112,
+      "step": 13005
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.565922470287683e-06,
+      "loss": 0.1968,
+      "step": 13010
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.5603687659669e-06,
+      "loss": 0.0424,
+      "step": 13015
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.5548150616461185e-06,
+      "loss": 0.0361,
+      "step": 13020
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.549261357325336e-06,
+      "loss": 0.0473,
+      "step": 13025
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.543707653004554e-06,
+      "loss": 0.0534,
+      "step": 13030
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.5381539486837724e-06,
+      "loss": 0.0201,
+      "step": 13035
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.53260024436299e-06,
+      "loss": 0.039,
+      "step": 13040
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.527046540042209e-06,
+      "loss": 0.067,
+      "step": 13045
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.521492835721427e-06,
+      "loss": 0.0974,
+      "step": 13050
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.515939131400645e-06,
+      "loss": 0.0977,
+      "step": 13055
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.510385427079863e-06,
+      "loss": 0.2235,
+      "step": 13060
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.504831722759081e-06,
+      "loss": 0.0406,
+      "step": 13065
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.499278018438299e-06,
+      "loss": 0.0517,
+      "step": 13070
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.493724314117517e-06,
+      "loss": 0.0481,
+      "step": 13075
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.488170609796734e-06,
+      "loss": 0.0484,
+      "step": 13080
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.4826169054759525e-06,
+      "loss": 0.0416,
+      "step": 13085
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.477063201155171e-06,
+      "loss": 0.0208,
+      "step": 13090
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.471509496834388e-06,
+      "loss": 0.0607,
+      "step": 13095
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.465955792513607e-06,
+      "loss": 0.0756,
+      "step": 13100
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.460402088192826e-06,
+      "loss": 0.153,
+      "step": 13105
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.454848383872043e-06,
+      "loss": 0.1992,
+      "step": 13110
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.449294679551261e-06,
+      "loss": 0.0535,
+      "step": 13115
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.4437409752304795e-06,
+      "loss": 0.0354,
+      "step": 13120
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 5.438187270909697e-06,
+      "loss": 0.0491,
+      "step": 13125
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.432633566588915e-06,
+      "loss": 0.0543,
+      "step": 13130
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.4270798622681335e-06,
+      "loss": 0.0306,
+      "step": 13135
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.421526157947351e-06,
+      "loss": 0.0458,
+      "step": 13140
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.415972453626569e-06,
+      "loss": 0.0569,
+      "step": 13145
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.4104187493057866e-06,
+      "loss": 0.0434,
+      "step": 13150
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.404865044985006e-06,
+      "loss": 0.1366,
+      "step": 13155
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.399311340664224e-06,
+      "loss": 0.2183,
+      "step": 13160
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.393757636343442e-06,
+      "loss": 0.0428,
+      "step": 13165
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.38820393202266e-06,
+      "loss": 0.0404,
+      "step": 13170
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 5.382650227701878e-06,
+      "loss": 0.047,
+      "step": 13175
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.377096523381095e-06,
+      "loss": 0.0688,
+      "step": 13180
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.3715428190603136e-06,
+      "loss": 0.0204,
+      "step": 13185
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.365989114739532e-06,
+      "loss": 0.0421,
+      "step": 13190
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.360435410418749e-06,
+      "loss": 0.0511,
+      "step": 13195
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.3548817060979675e-06,
+      "loss": 0.0686,
+      "step": 13200
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.349328001777186e-06,
+      "loss": 0.105,
+      "step": 13205
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.343774297456403e-06,
+      "loss": 0.1827,
+      "step": 13210
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.338220593135622e-06,
+      "loss": 0.0452,
+      "step": 13215
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 5.3326668888148406e-06,
+      "loss": 0.0538,
+      "step": 13220
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.327113184494058e-06,
+      "loss": 0.0409,
+      "step": 13225
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.321559480173276e-06,
+      "loss": 0.0534,
+      "step": 13230
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.3160057758524945e-06,
+      "loss": 0.04,
+      "step": 13235
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.310452071531712e-06,
+      "loss": 0.0354,
+      "step": 13240
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.30489836721093e-06,
+      "loss": 0.0553,
+      "step": 13245
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.299344662890148e-06,
+      "loss": 0.0761,
+      "step": 13250
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.293790958569366e-06,
+      "loss": 0.1325,
+      "step": 13255
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.288237254248584e-06,
+      "loss": 0.1614,
+      "step": 13260
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 5.2826835499278015e-06,
+      "loss": 0.0289,
+      "step": 13265
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.277129845607021e-06,
+      "loss": 0.0462,
+      "step": 13270
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.271576141286239e-06,
+      "loss": 0.0587,
+      "step": 13275
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.266022436965456e-06,
+      "loss": 0.062,
+      "step": 13280
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.260468732644675e-06,
+      "loss": 0.0388,
+      "step": 13285
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.254915028323893e-06,
+      "loss": 0.0305,
+      "step": 13290
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.24936132400311e-06,
+      "loss": 0.0658,
+      "step": 13295
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.2438076196823285e-06,
+      "loss": 0.0981,
+      "step": 13300
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.238253915361546e-06,
+      "loss": 0.135,
+      "step": 13305
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 5.232700211040764e-06,
+      "loss": 0.153,
+      "step": 13310
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.2271465067199825e-06,
+      "loss": 0.0391,
+      "step": 13315
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.2215928023992e-06,
+      "loss": 0.0428,
+      "step": 13320
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.216039098078419e-06,
+      "loss": 0.054,
+      "step": 13325
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.210485393757637e-06,
+      "loss": 0.0645,
+      "step": 13330
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.204931689436855e-06,
+      "loss": 0.0226,
+      "step": 13335
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.199377985116073e-06,
+      "loss": 0.1013,
+      "step": 13340
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.193824280795291e-06,
+      "loss": 0.0699,
+      "step": 13345
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.188270576474509e-06,
+      "loss": 0.0887,
+      "step": 13350
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 5.182716872153727e-06,
+      "loss": 0.112,
+      "step": 13355
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.177163167832945e-06,
+      "loss": 0.1593,
+      "step": 13360
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.171609463512163e-06,
+      "loss": 0.0393,
+      "step": 13365
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.166055759191381e-06,
+      "loss": 0.0398,
+      "step": 13370
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.160502054870598e-06,
+      "loss": 0.0428,
+      "step": 13375
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.154948350549817e-06,
+      "loss": 0.0551,
+      "step": 13380
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.149394646229036e-06,
+      "loss": 0.0454,
+      "step": 13385
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.143840941908254e-06,
+      "loss": 0.0657,
+      "step": 13390
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.138287237587471e-06,
+      "loss": 0.0481,
+      "step": 13395
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.13273353326669e-06,
+      "loss": 0.0731,
+      "step": 13400
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.127179828945907e-06,
+      "loss": 0.1257,
+      "step": 13405
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.121626124625125e-06,
+      "loss": 0.2243,
+      "step": 13410
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.1160724203043435e-06,
+      "loss": 0.0385,
+      "step": 13415
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.110518715983561e-06,
+      "loss": 0.0414,
+      "step": 13420
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.104965011662779e-06,
+      "loss": 0.0564,
+      "step": 13425
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.0994113073419975e-06,
+      "loss": 0.0693,
+      "step": 13430
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.093857603021215e-06,
+      "loss": 0.0217,
+      "step": 13435
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.088303898700434e-06,
+      "loss": 0.0385,
+      "step": 13440
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 5.082750194379652e-06,
+      "loss": 0.0665,
+      "step": 13445
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.07719649005887e-06,
+      "loss": 0.0908,
+      "step": 13450
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.071642785738088e-06,
+      "loss": 0.1358,
+      "step": 13455
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.066089081417306e-06,
+      "loss": 0.1904,
+      "step": 13460
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.060535377096524e-06,
+      "loss": 0.0292,
+      "step": 13465
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.054981672775742e-06,
+      "loss": 0.042,
+      "step": 13470
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.049427968454959e-06,
+      "loss": 0.0439,
+      "step": 13475
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.0438742641341776e-06,
+      "loss": 0.1132,
+      "step": 13480
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.038320559813396e-06,
+      "loss": 0.0425,
+      "step": 13485
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.032766855492613e-06,
+      "loss": 0.0318,
+      "step": 13490
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.027213151171832e-06,
+      "loss": 0.0566,
+      "step": 13495
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.021659446851051e-06,
+      "loss": 0.0831,
+      "step": 13500
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.016105742530268e-06,
+      "loss": 0.1122,
+      "step": 13505
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.010552038209486e-06,
+      "loss": 0.1708,
+      "step": 13510
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.0049983338887046e-06,
+      "loss": 0.0811,
+      "step": 13515
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.999444629567922e-06,
+      "loss": 0.0386,
+      "step": 13520
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.99389092524714e-06,
+      "loss": 0.0481,
+      "step": 13525
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.988337220926358e-06,
+      "loss": 0.0455,
+      "step": 13530
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.982783516605577e-06,
+      "loss": 0.0541,
+      "step": 13535
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.977229812284794e-06,
+      "loss": 0.0234,
+      "step": 13540
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.9716761079640125e-06,
+      "loss": 0.0339,
+      "step": 13545
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.966122403643231e-06,
+      "loss": 0.0835,
+      "step": 13550
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.960568699322448e-06,
+      "loss": 0.0879,
+      "step": 13555
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.955014995001666e-06,
+      "loss": 0.1474,
+      "step": 13560
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.949461290680885e-06,
+      "loss": 0.1736,
+      "step": 13565
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.943907586360103e-06,
+      "loss": 0.0374,
+      "step": 13570
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.93835388203932e-06,
+      "loss": 0.0299,
+      "step": 13575
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.932800177718539e-06,
+      "loss": 0.0532,
+      "step": 13580
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.927246473397757e-06,
+      "loss": 0.0492,
+      "step": 13585
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.921692769076975e-06,
+      "loss": 0.018,
+      "step": 13590
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.9161390647561925e-06,
+      "loss": 0.0367,
+      "step": 13595
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.910585360435411e-06,
+      "loss": 0.0452,
+      "step": 13600
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.905031656114629e-06,
+      "loss": 0.0746,
+      "step": 13605
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.8994779517938465e-06,
+      "loss": 0.1374,
+      "step": 13610
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.893924247473065e-06,
+      "loss": 0.1607,
+      "step": 13615
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.888370543152283e-06,
+      "loss": 0.0426,
+      "step": 13620
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.882816838831501e-06,
+      "loss": 0.038,
+      "step": 13625
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.877263134510719e-06,
+      "loss": 0.049,
+      "step": 13630
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.871709430189937e-06,
+      "loss": 0.0741,
+      "step": 13635
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.866155725869155e-06,
+      "loss": 0.0196,
+      "step": 13640
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.860602021548373e-06,
+      "loss": 0.0364,
+      "step": 13645
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.855048317227592e-06,
+      "loss": 0.0537,
+      "step": 13650
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.849494612906809e-06,
+      "loss": 0.0612,
+      "step": 13655
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.8439409085860274e-06,
+      "loss": 0.1277,
+      "step": 13660
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.838387204265245e-06,
+      "loss": 0.1152,
+      "step": 13665
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.832833499944463e-06,
+      "loss": 0.0468,
+      "step": 13670
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.827279795623681e-06,
+      "loss": 0.0317,
+      "step": 13675
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.8217260913029e-06,
+      "loss": 0.0419,
+      "step": 13680
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.816172386982118e-06,
+      "loss": 0.0611,
+      "step": 13685
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.810618682661335e-06,
+      "loss": 0.0354,
+      "step": 13690
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.805064978340554e-06,
+      "loss": 0.0339,
+      "step": 13695
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.799511274019771e-06,
+      "loss": 0.0781,
+      "step": 13700
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.79395756969899e-06,
+      "loss": 0.0652,
+      "step": 13705
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.7884038653782075e-06,
+      "loss": 0.1137,
+      "step": 13710
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.783960901921582e-06,
+      "loss": 0.1953,
+      "step": 13715
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.7784071976008e-06,
+      "loss": 0.0447,
+      "step": 13720
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.772853493280018e-06,
+      "loss": 0.0438,
+      "step": 13725
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.767299788959236e-06,
+      "loss": 0.0416,
+      "step": 13730
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.761746084638455e-06,
+      "loss": 0.0701,
+      "step": 13735
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.756192380317672e-06,
+      "loss": 0.0204,
+      "step": 13740
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.7506386759968905e-06,
+      "loss": 0.0383,
+      "step": 13745
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.745084971676108e-06,
+      "loss": 0.0457,
+      "step": 13750
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.739531267355326e-06,
+      "loss": 0.0962,
+      "step": 13755
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.7339775630345445e-06,
+      "loss": 0.1407,
+      "step": 13760
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.728423858713763e-06,
+      "loss": 0.154,
+      "step": 13765
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.72287015439298e-06,
+      "loss": 0.0404,
+      "step": 13770
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.7173164500721984e-06,
+      "loss": 0.0435,
+      "step": 13775
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.711762745751417e-06,
+      "loss": 0.0602,
+      "step": 13780
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.706209041430634e-06,
+      "loss": 0.061,
+      "step": 13785
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.700655337109853e-06,
+      "loss": 0.036,
+      "step": 13790
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.695101632789071e-06,
+      "loss": 0.035,
+      "step": 13795
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.689547928468289e-06,
+      "loss": 0.0497,
+      "step": 13800
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.683994224147506e-06,
+      "loss": 0.0797,
+      "step": 13805
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.678440519826725e-06,
+      "loss": 0.1479,
+      "step": 13810
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.672886815505943e-06,
+      "loss": 0.1496,
+      "step": 13815
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.667333111185161e-06,
+      "loss": 0.0377,
+      "step": 13820
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.661779406864379e-06,
+      "loss": 0.058,
+      "step": 13825
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.656225702543597e-06,
+      "loss": 0.0498,
+      "step": 13830
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.650671998222815e-06,
+      "loss": 0.0627,
+      "step": 13835
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.6451182939020325e-06,
+      "loss": 0.0314,
+      "step": 13840
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.639564589581252e-06,
+      "loss": 0.0322,
+      "step": 13845
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.634010885260469e-06,
+      "loss": 0.0646,
+      "step": 13850
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.628457180939687e-06,
+      "loss": 0.0726,
+      "step": 13855
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.6229034766189055e-06,
+      "loss": 0.1476,
+      "step": 13860
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.617349772298123e-06,
+      "loss": 0.1621,
+      "step": 13865
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.611796067977341e-06,
+      "loss": 0.0399,
+      "step": 13870
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.6062423636565595e-06,
+      "loss": 0.0613,
+      "step": 13875
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.600688659335778e-06,
+      "loss": 0.0553,
+      "step": 13880
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.595134955014995e-06,
+      "loss": 0.045,
+      "step": 13885
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.589581250694213e-06,
+      "loss": 0.0274,
+      "step": 13890
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.584027546373432e-06,
+      "loss": 0.0534,
+      "step": 13895
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.578473842052649e-06,
+      "loss": 0.0448,
+      "step": 13900
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.572920137731867e-06,
+      "loss": 0.0856,
+      "step": 13905
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.567366433411086e-06,
+      "loss": 0.1374,
+      "step": 13910
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.561812729090304e-06,
+      "loss": 0.13,
+      "step": 13915
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.556259024769521e-06,
+      "loss": 0.0524,
+      "step": 13920
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.5507053204487396e-06,
+      "loss": 0.052,
+      "step": 13925
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.545151616127958e-06,
+      "loss": 0.0476,
+      "step": 13930
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.539597911807176e-06,
+      "loss": 0.0595,
+      "step": 13935
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.5340442074863935e-06,
+      "loss": 0.0231,
+      "step": 13940
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.528490503165612e-06,
+      "loss": 0.027,
+      "step": 13945
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.52293679884483e-06,
+      "loss": 0.0551,
+      "step": 13950
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.5173830945240474e-06,
+      "loss": 0.1033,
+      "step": 13955
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.5118293902032666e-06,
+      "loss": 0.13,
+      "step": 13960
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.506275685882484e-06,
+      "loss": 0.204,
+      "step": 13965
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.500721981561702e-06,
+      "loss": 0.0429,
+      "step": 13970
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.49516827724092e-06,
+      "loss": 0.0566,
+      "step": 13975
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.489614572920138e-06,
+      "loss": 0.0459,
+      "step": 13980
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.484060868599356e-06,
+      "loss": 0.0549,
+      "step": 13985
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.4785071642785744e-06,
+      "loss": 0.0285,
+      "step": 13990
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.472953459957793e-06,
+      "loss": 0.0533,
+      "step": 13995
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.46739975563701e-06,
+      "loss": 0.0672,
+      "step": 14000
+    },
+    {
+      "epoch": 3.11,
+      "eval_loss": 0.023140855133533478,
+      "eval_runtime": 1621.1022,
+      "eval_samples_per_second": 9.881,
+      "eval_steps_per_second": 2.471,
+      "eval_wer": 0.0426974395961053,
+      "step": 14000
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.461846051316228e-06,
+      "loss": 0.0839,
+      "step": 14005
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.456292346995446e-06,
+      "loss": 0.1067,
+      "step": 14010
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.450738642674665e-06,
+      "loss": 0.1429,
+      "step": 14015
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.445184938353882e-06,
+      "loss": 0.0417,
+      "step": 14020
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.439631234033101e-06,
+      "loss": 0.0473,
+      "step": 14025
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.434077529712318e-06,
+      "loss": 0.0509,
+      "step": 14030
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.428523825391536e-06,
+      "loss": 0.0652,
+      "step": 14035
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.4229701210707545e-06,
+      "loss": 0.0266,
+      "step": 14040
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.417416416749973e-06,
+      "loss": 0.0261,
+      "step": 14045
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.411862712429191e-06,
+      "loss": 0.0735,
+      "step": 14050
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.4063090081084085e-06,
+      "loss": 0.0837,
+      "step": 14055
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.400755303787627e-06,
+      "loss": 0.1769,
+      "step": 14060
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.395201599466844e-06,
+      "loss": 0.1602,
+      "step": 14065
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.389647895146063e-06,
+      "loss": 0.0359,
+      "step": 14070
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.384094190825281e-06,
+      "loss": 0.0476,
+      "step": 14075
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.378540486504499e-06,
+      "loss": 0.0453,
+      "step": 14080
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.372986782183717e-06,
+      "loss": 0.0559,
+      "step": 14085
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.367433077862935e-06,
+      "loss": 0.0228,
+      "step": 14090
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.361879373542153e-06,
+      "loss": 0.0258,
+      "step": 14095
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.356325669221371e-06,
+      "loss": 0.0581,
+      "step": 14100
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.3507719649005894e-06,
+      "loss": 0.0716,
+      "step": 14105
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.345218260579807e-06,
+      "loss": 0.1667,
+      "step": 14110
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.339664556259025e-06,
+      "loss": 0.1279,
+      "step": 14115
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.334110851938243e-06,
+      "loss": 0.0533,
+      "step": 14120
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.328557147617461e-06,
+      "loss": 0.0438,
+      "step": 14125
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.323003443296679e-06,
+      "loss": 0.0474,
+      "step": 14130
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.317449738975897e-06,
+      "loss": 0.0404,
+      "step": 14135
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.311896034655116e-06,
+      "loss": 0.0242,
+      "step": 14140
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.306342330334333e-06,
+      "loss": 0.0391,
+      "step": 14145
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.300788626013551e-06,
+      "loss": 0.0575,
+      "step": 14150
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.2952349216927695e-06,
+      "loss": 0.0706,
+      "step": 14155
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.289681217371988e-06,
+      "loss": 0.132,
+      "step": 14160
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.284127513051205e-06,
+      "loss": 0.1496,
+      "step": 14165
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.2785738087304235e-06,
+      "loss": 0.0403,
+      "step": 14170
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.273020104409642e-06,
+      "loss": 0.0398,
+      "step": 14175
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.267466400088859e-06,
+      "loss": 0.0504,
+      "step": 14180
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.261912695768078e-06,
+      "loss": 0.049,
+      "step": 14185
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.256358991447296e-06,
+      "loss": 0.0142,
+      "step": 14190
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.250805287126514e-06,
+      "loss": 0.0519,
+      "step": 14195
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.245251582805731e-06,
+      "loss": 0.055,
+      "step": 14200
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.23969787848495e-06,
+      "loss": 0.0542,
+      "step": 14205
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.234144174164168e-06,
+      "loss": 0.1572,
+      "step": 14210
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.228590469843386e-06,
+      "loss": 0.1989,
+      "step": 14215
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.223036765522604e-06,
+      "loss": 0.0461,
+      "step": 14220
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.217483061201822e-06,
+      "loss": 0.0387,
+      "step": 14225
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.21192935688104e-06,
+      "loss": 0.0615,
+      "step": 14230
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.2063756525602575e-06,
+      "loss": 0.0505,
+      "step": 14235
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.200821948239477e-06,
+      "loss": 0.0261,
+      "step": 14240
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.195268243918694e-06,
+      "loss": 0.03,
+      "step": 14245
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.189714539597912e-06,
+      "loss": 0.0723,
+      "step": 14250
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.18416083527713e-06,
+      "loss": 0.0871,
+      "step": 14255
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.178607130956348e-06,
+      "loss": 0.1236,
+      "step": 14260
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.173053426635566e-06,
+      "loss": 0.1194,
+      "step": 14265
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.1674997223147845e-06,
+      "loss": 0.042,
+      "step": 14270
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.161946017994003e-06,
+      "loss": 0.0268,
+      "step": 14275
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.15639231367322e-06,
+      "loss": 0.0408,
+      "step": 14280
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.1508386093524385e-06,
+      "loss": 0.0475,
+      "step": 14285
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.145284905031656e-06,
+      "loss": 0.0187,
+      "step": 14290
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.139731200710875e-06,
+      "loss": 0.0293,
+      "step": 14295
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.134177496390092e-06,
+      "loss": 0.0459,
+      "step": 14300
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.128623792069311e-06,
+      "loss": 0.0705,
+      "step": 14305
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.123070087748529e-06,
+      "loss": 0.1853,
+      "step": 14310
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.117516383427746e-06,
+      "loss": 0.1518,
+      "step": 14315
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.111962679106965e-06,
+      "loss": 0.0486,
+      "step": 14320
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.106408974786183e-06,
+      "loss": 0.0374,
+      "step": 14325
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.100855270465401e-06,
+      "loss": 0.058,
+      "step": 14330
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.0953015661446185e-06,
+      "loss": 0.0532,
+      "step": 14335
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.089747861823837e-06,
+      "loss": 0.0313,
+      "step": 14340
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.084194157503055e-06,
+      "loss": 0.0372,
+      "step": 14345
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.0786404531822725e-06,
+      "loss": 0.0476,
+      "step": 14350
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.073086748861491e-06,
+      "loss": 0.0824,
+      "step": 14355
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.067533044540709e-06,
+      "loss": 0.1374,
+      "step": 14360
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.061979340219927e-06,
+      "loss": 0.1388,
+      "step": 14365
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.056425635899145e-06,
+      "loss": 0.0357,
+      "step": 14370
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.050871931578363e-06,
+      "loss": 0.0432,
+      "step": 14375
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.045318227257581e-06,
+      "loss": 0.0527,
+      "step": 14380
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.0397645229367995e-06,
+      "loss": 0.044,
+      "step": 14385
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.034210818616017e-06,
+      "loss": 0.0906,
+      "step": 14390
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.028657114295235e-06,
+      "loss": 0.0302,
+      "step": 14395
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.0231034099744534e-06,
+      "loss": 0.0681,
+      "step": 14400
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.017549705653671e-06,
+      "loss": 0.0746,
+      "step": 14405
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.01199600133289e-06,
+      "loss": 0.1498,
+      "step": 14410
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.006442297012107e-06,
+      "loss": 0.123,
+      "step": 14415
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.000888592691326e-06,
+      "loss": 0.0293,
+      "step": 14420
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.995334888370543e-06,
+      "loss": 0.0506,
+      "step": 14425
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.989781184049761e-06,
+      "loss": 0.0545,
+      "step": 14430
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.98422747972898e-06,
+      "loss": 0.0521,
+      "step": 14435
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.978673775408198e-06,
+      "loss": 0.0181,
+      "step": 14440
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.973120071087416e-06,
+      "loss": 0.0397,
+      "step": 14445
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.9675663667666335e-06,
+      "loss": 0.0394,
+      "step": 14450
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.962012662445852e-06,
+      "loss": 0.0786,
+      "step": 14455
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.956458958125069e-06,
+      "loss": 0.1221,
+      "step": 14460
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.950905253804288e-06,
+      "loss": 0.1732,
+      "step": 14465
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.945351549483506e-06,
+      "loss": 0.0398,
+      "step": 14470
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.939797845162724e-06,
+      "loss": 0.0411,
+      "step": 14475
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.934244140841941e-06,
+      "loss": 0.0598,
+      "step": 14480
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.92869043652116e-06,
+      "loss": 0.0812,
+      "step": 14485
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.923136732200378e-06,
+      "loss": 0.0276,
+      "step": 14490
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.917583027879596e-06,
+      "loss": 0.0421,
+      "step": 14495
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.9120293235588145e-06,
+      "loss": 0.0518,
+      "step": 14500
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.906475619238032e-06,
+      "loss": 0.0661,
+      "step": 14505
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.90092191491725e-06,
+      "loss": 0.1628,
+      "step": 14510
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.8953682105964676e-06,
+      "loss": 0.2081,
+      "step": 14515
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.889814506275687e-06,
+      "loss": 0.0419,
+      "step": 14520
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.884260801954904e-06,
+      "loss": 0.0415,
+      "step": 14525
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.878707097634122e-06,
+      "loss": 0.049,
+      "step": 14530
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.873153393313341e-06,
+      "loss": 0.0589,
+      "step": 14535
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.867599688992558e-06,
+      "loss": 0.026,
+      "step": 14540
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.862045984671776e-06,
+      "loss": 0.0416,
+      "step": 14545
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.8564922803509946e-06,
+      "loss": 0.0518,
+      "step": 14550
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.850938576030213e-06,
+      "loss": 0.1061,
+      "step": 14555
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.84538487170943e-06,
+      "loss": 0.1337,
+      "step": 14560
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.8398311673886485e-06,
+      "loss": 0.1199,
+      "step": 14565
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.834277463067867e-06,
+      "loss": 0.0513,
+      "step": 14570
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.828723758747084e-06,
+      "loss": 0.0494,
+      "step": 14575
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.8231700544263025e-06,
+      "loss": 0.048,
+      "step": 14580
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.817616350105521e-06,
+      "loss": 0.0586,
+      "step": 14585
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.8120626457847386e-06,
+      "loss": 0.0325,
+      "step": 14590
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.806508941463957e-06,
+      "loss": 0.0289,
+      "step": 14595
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.8009552371431747e-06,
+      "loss": 0.058,
+      "step": 14600
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.795401532822393e-06,
+      "loss": 0.0642,
+      "step": 14605
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.789847828501611e-06,
+      "loss": 0.1372,
+      "step": 14610
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.784294124180829e-06,
+      "loss": 0.1634,
+      "step": 14615
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.778740419860047e-06,
+      "loss": 0.0393,
+      "step": 14620
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.7731867155392647e-06,
+      "loss": 0.0367,
+      "step": 14625
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.767633011218483e-06,
+      "loss": 0.0376,
+      "step": 14630
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.7620793068977012e-06,
+      "loss": 0.0522,
+      "step": 14635
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.756525602576919e-06,
+      "loss": 0.0298,
+      "step": 14640
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.7509718982561373e-06,
+      "loss": 0.0396,
+      "step": 14645
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.745418193935355e-06,
+      "loss": 0.0638,
+      "step": 14650
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.739864489614573e-06,
+      "loss": 0.0948,
+      "step": 14655
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.7343107852937917e-06,
+      "loss": 0.1473,
+      "step": 14660
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.7287570809730096e-06,
+      "loss": 0.1715,
+      "step": 14665
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.7232033766522274e-06,
+      "loss": 0.0373,
+      "step": 14670
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.7176496723314452e-06,
+      "loss": 0.0499,
+      "step": 14675
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.7120959680106635e-06,
+      "loss": 0.0441,
+      "step": 14680
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.7065422636898813e-06,
+      "loss": 0.0573,
+      "step": 14685
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.7009885593690996e-06,
+      "loss": 0.0185,
+      "step": 14690
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.695434855048318e-06,
+      "loss": 0.0402,
+      "step": 14695
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.6898811507275357e-06,
+      "loss": 0.0626,
+      "step": 14700
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.6843274464067535e-06,
+      "loss": 0.0729,
+      "step": 14705
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6787737420859714e-06,
+      "loss": 0.1377,
+      "step": 14710
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6732200377651892e-06,
+      "loss": 0.1227,
+      "step": 14715
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.667666333444408e-06,
+      "loss": 0.0569,
+      "step": 14720
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6621126291236258e-06,
+      "loss": 0.043,
+      "step": 14725
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6565589248028436e-06,
+      "loss": 0.0419,
+      "step": 14730
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.651005220482062e-06,
+      "loss": 0.0462,
+      "step": 14735
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6454515161612797e-06,
+      "loss": 0.0234,
+      "step": 14740
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6398978118404984e-06,
+      "loss": 0.0374,
+      "step": 14745
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.6343441075197162e-06,
+      "loss": 0.0481,
+      "step": 14750
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.628790403198934e-06,
+      "loss": 0.0639,
+      "step": 14755
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.623236698878152e-06,
+      "loss": 0.1033,
+      "step": 14760
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.6176829945573697e-06,
+      "loss": 0.1499,
+      "step": 14765
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.612129290236588e-06,
+      "loss": 0.0413,
+      "step": 14770
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.6065755859158063e-06,
+      "loss": 0.0503,
+      "step": 14775
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.601021881595024e-06,
+      "loss": 0.0418,
+      "step": 14780
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.5954681772742424e-06,
+      "loss": 0.0468,
+      "step": 14785
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.5899144729534602e-06,
+      "loss": 0.0313,
+      "step": 14790
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.584360768632678e-06,
+      "loss": 0.0408,
+      "step": 14795
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.578807064311896e-06,
+      "loss": 0.0624,
+      "step": 14800
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5732533599911146e-06,
+      "loss": 0.0733,
+      "step": 14805
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5676996556703324e-06,
+      "loss": 0.1291,
+      "step": 14810
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5621459513495503e-06,
+      "loss": 0.1152,
+      "step": 14815
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5565922470287685e-06,
+      "loss": 0.0484,
+      "step": 14820
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5510385427079864e-06,
+      "loss": 0.0416,
+      "step": 14825
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5454848383872046e-06,
+      "loss": 0.0438,
+      "step": 14830
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.539931134066423e-06,
+      "loss": 0.0427,
+      "step": 14835
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5343774297456407e-06,
+      "loss": 0.0135,
+      "step": 14840
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.5288237254248586e-06,
+      "loss": 0.0292,
+      "step": 14845
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.5232700211040764e-06,
+      "loss": 0.0529,
+      "step": 14850
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.5177163167832947e-06,
+      "loss": 0.0711,
+      "step": 14855
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.512162612462513e-06,
+      "loss": 0.1422,
+      "step": 14860
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.5066089081417308e-06,
+      "loss": 0.153,
+      "step": 14865
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.501055203820949e-06,
+      "loss": 0.0325,
+      "step": 14870
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.495501499500167e-06,
+      "loss": 0.0389,
+      "step": 14875
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.4899477951793847e-06,
+      "loss": 0.0482,
+      "step": 14880
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.4843940908586026e-06,
+      "loss": 0.0467,
+      "step": 14885
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.4788403865378213e-06,
+      "loss": 0.0241,
+      "step": 14890
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.473286682217039e-06,
+      "loss": 0.0541,
+      "step": 14895
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.467732977896257e-06,
+      "loss": 0.0458,
+      "step": 14900
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.462179273575475e-06,
+      "loss": 0.0819,
+      "step": 14905
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.456625569254693e-06,
+      "loss": 0.0887,
+      "step": 14910
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.4510718649339113e-06,
+      "loss": 0.208,
+      "step": 14915
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.4455181606131296e-06,
+      "loss": 0.035,
+      "step": 14920
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.4399644562923474e-06,
+      "loss": 0.0442,
+      "step": 14925
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.4344107519715652e-06,
+      "loss": 0.0495,
+      "step": 14930
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.428857047650783e-06,
+      "loss": 0.0478,
+      "step": 14935
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.4233033433300013e-06,
+      "loss": 0.0222,
+      "step": 14940
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.4177496390092196e-06,
+      "loss": 0.0478,
+      "step": 14945
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.4121959346884375e-06,
+      "loss": 0.0628,
+      "step": 14950
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.4066422303676557e-06,
+      "loss": 0.08,
+      "step": 14955
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.4010885260468736e-06,
+      "loss": 0.1492,
+      "step": 14960
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.3955348217260914e-06,
+      "loss": 0.1534,
+      "step": 14965
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.38998111740531e-06,
+      "loss": 0.0436,
+      "step": 14970
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.384427413084528e-06,
+      "loss": 0.0488,
+      "step": 14975
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.3788737087637458e-06,
+      "loss": 0.0452,
+      "step": 14980
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.3733200044429636e-06,
+      "loss": 0.0485,
+      "step": 14985
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.3677663001221814e-06,
+      "loss": 0.0355,
+      "step": 14990
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.3622125958013997e-06,
+      "loss": 0.0304,
+      "step": 14995
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.356658891480618e-06,
+      "loss": 0.0545,
+      "step": 15000
+    },
+    {
+      "epoch": 3.33,
+      "eval_loss": 0.024257851764559746,
+      "eval_runtime": 2136.2659,
+      "eval_samples_per_second": 7.498,
+      "eval_steps_per_second": 1.875,
+      "eval_wer": 0.0430580598629643,
+      "step": 15000
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.351105187159836e-06,
+      "loss": 0.0769,
+      "step": 15005
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.345551482839054e-06,
+      "loss": 0.1333,
+      "step": 15010
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.339997778518272e-06,
+      "loss": 0.1119,
+      "step": 15015
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.3344440741974898e-06,
+      "loss": 0.0408,
+      "step": 15020
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.3288903698767076e-06,
+      "loss": 0.046,
+      "step": 15025
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.3233366655559263e-06,
+      "loss": 0.0466,
+      "step": 15030
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.317782961235144e-06,
+      "loss": 0.0577,
+      "step": 15035
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.312229256914362e-06,
+      "loss": 0.0242,
+      "step": 15040
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.3066755525935802e-06,
+      "loss": 0.0568,
+      "step": 15045
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.301121848272798e-06,
+      "loss": 0.0489,
+      "step": 15050
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.2955681439520163e-06,
+      "loss": 0.0895,
+      "step": 15055
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.2900144396312346e-06,
+      "loss": 0.1436,
+      "step": 15060
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.2844607353104524e-06,
+      "loss": 0.1437,
+      "step": 15065
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2789070309896703e-06,
+      "loss": 0.0361,
+      "step": 15070
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.273353326668888e-06,
+      "loss": 0.0483,
+      "step": 15075
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2677996223481064e-06,
+      "loss": 0.0587,
+      "step": 15080
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2622459180273246e-06,
+      "loss": 0.0571,
+      "step": 15085
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2566922137065425e-06,
+      "loss": 0.016,
+      "step": 15090
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2511385093857607e-06,
+      "loss": 0.0392,
+      "step": 15095
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2455848050649786e-06,
+      "loss": 0.0549,
+      "step": 15100
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2400311007441964e-06,
+      "loss": 0.0697,
+      "step": 15105
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.2344773964234143e-06,
+      "loss": 0.1307,
+      "step": 15110
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.228923692102633e-06,
+      "loss": 0.1407,
+      "step": 15115
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.223369987781851e-06,
+      "loss": 0.0304,
+      "step": 15120
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.2178162834610686e-06,
+      "loss": 0.0439,
+      "step": 15125
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.212262579140287e-06,
+      "loss": 0.0434,
+      "step": 15130
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.2067088748195047e-06,
+      "loss": 0.046,
+      "step": 15135
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.201155170498723e-06,
+      "loss": 0.0138,
+      "step": 15140
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.1956014661779413e-06,
+      "loss": 0.0624,
+      "step": 15145
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.190047761857159e-06,
+      "loss": 0.055,
+      "step": 15150
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.184494057536377e-06,
+      "loss": 0.0807,
+      "step": 15155
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.1789403532155948e-06,
+      "loss": 0.1188,
+      "step": 15160
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.173386648894813e-06,
+      "loss": 0.1414,
+      "step": 15165
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.1678329445740313e-06,
+      "loss": 0.0365,
+      "step": 15170
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.162279240253249e-06,
+      "loss": 0.0407,
+      "step": 15175
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.1567255359324674e-06,
+      "loss": 0.0472,
+      "step": 15180
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.1511718316116853e-06,
+      "loss": 0.051,
+      "step": 15185
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.145618127290903e-06,
+      "loss": 0.029,
+      "step": 15190
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.1400644229701218e-06,
+      "loss": 0.0342,
+      "step": 15195
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.1345107186493396e-06,
+      "loss": 0.0621,
+      "step": 15200
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.1289570143285575e-06,
+      "loss": 0.1165,
+      "step": 15205
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.1234033100077753e-06,
+      "loss": 0.1623,
+      "step": 15210
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.117849605686993e-06,
+      "loss": 0.1588,
+      "step": 15215
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.1122959013662114e-06,
+      "loss": 0.0395,
+      "step": 15220
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.1067421970454297e-06,
+      "loss": 0.0398,
+      "step": 15225
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.101188492724648e-06,
+      "loss": 0.06,
+      "step": 15230
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.0956347884038658e-06,
+      "loss": 0.0458,
+      "step": 15235
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.0900810840830836e-06,
+      "loss": 0.0143,
+      "step": 15240
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.0845273797623015e-06,
+      "loss": 0.0424,
+      "step": 15245
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.0789736754415193e-06,
+      "loss": 0.0632,
+      "step": 15250
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.073419971120738e-06,
+      "loss": 0.0695,
+      "step": 15255
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.067866266799956e-06,
+      "loss": 0.1518,
+      "step": 15260
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.0623125624791737e-06,
+      "loss": 0.1443,
+      "step": 15265
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.056758858158392e-06,
+      "loss": 0.0546,
+      "step": 15270
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.0512051538376098e-06,
+      "loss": 0.0533,
+      "step": 15275
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.045651449516828e-06,
+      "loss": 0.0491,
+      "step": 15280
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.0400977451960463e-06,
+      "loss": 0.0452,
+      "step": 15285
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.034544040875264e-06,
+      "loss": 0.0283,
+      "step": 15290
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.028990336554482e-06,
+      "loss": 0.0377,
+      "step": 15295
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.0234366322337e-06,
+      "loss": 0.0565,
+      "step": 15300
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.017882927912918e-06,
+      "loss": 0.0712,
+      "step": 15305
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.0123292235921363e-06,
+      "loss": 0.0967,
+      "step": 15310
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.006775519271354e-06,
+      "loss": 0.13,
+      "step": 15315
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.0012218149505724e-06,
+      "loss": 0.0286,
+      "step": 15320
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.9956681106297903e-06,
+      "loss": 0.0391,
+      "step": 15325
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.990114406309008e-06,
+      "loss": 0.0393,
+      "step": 15330
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.984560701988226e-06,
+      "loss": 0.0521,
+      "step": 15335
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9790069976674447e-06,
+      "loss": 0.0231,
+      "step": 15340
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9734532933466625e-06,
+      "loss": 0.0303,
+      "step": 15345
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9678995890258803e-06,
+      "loss": 0.067,
+      "step": 15350
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9623458847050986e-06,
+      "loss": 0.0995,
+      "step": 15355
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9567921803843164e-06,
+      "loss": 0.125,
+      "step": 15360
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9512384760635347e-06,
+      "loss": 0.1428,
+      "step": 15365
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.945684771742753e-06,
+      "loss": 0.0454,
+      "step": 15370
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.940131067421971e-06,
+      "loss": 0.0411,
+      "step": 15375
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.9345773631011886e-06,
+      "loss": 0.0505,
+      "step": 15380
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.9290236587804065e-06,
+      "loss": 0.066,
+      "step": 15385
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.9234699544596248e-06,
+      "loss": 0.0352,
+      "step": 15390
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.917916250138843e-06,
+      "loss": 0.0354,
+      "step": 15395
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.912362545818061e-06,
+      "loss": 0.0351,
+      "step": 15400
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.906808841497279e-06,
+      "loss": 0.0844,
+      "step": 15405
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.901255137176497e-06,
+      "loss": 0.1356,
+      "step": 15410
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.895701432855715e-06,
+      "loss": 0.1789,
+      "step": 15415
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.8901477285349335e-06,
+      "loss": 0.0331,
+      "step": 15420
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.8845940242141513e-06,
+      "loss": 0.0262,
+      "step": 15425
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.879040319893369e-06,
+      "loss": 0.0411,
+      "step": 15430
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.873486615572587e-06,
+      "loss": 0.0496,
+      "step": 15435
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.8679329112518053e-06,
+      "loss": 0.0167,
+      "step": 15440
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.862379206931023e-06,
+      "loss": 0.0326,
+      "step": 15445
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.8568255026102414e-06,
+      "loss": 0.0557,
+      "step": 15450
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.8512717982894596e-06,
+      "loss": 0.0848,
+      "step": 15455
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.8457180939686775e-06,
+      "loss": 0.1751,
+      "step": 15460
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.8401643896478953e-06,
+      "loss": 0.1241,
+      "step": 15465
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.834610685327113e-06,
+      "loss": 0.0357,
+      "step": 15470
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.829056981006331e-06,
+      "loss": 0.0448,
+      "step": 15475
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.8235032766855497e-06,
+      "loss": 0.0438,
+      "step": 15480
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.8179495723647675e-06,
+      "loss": 0.0379,
+      "step": 15485
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.8123958680439854e-06,
+      "loss": 0.0168,
+      "step": 15490
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.8068421637232036e-06,
+      "loss": 0.0434,
+      "step": 15495
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.8012884594024215e-06,
+      "loss": 0.0476,
+      "step": 15500
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.7957347550816397e-06,
+      "loss": 0.0803,
+      "step": 15505
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.790181050760858e-06,
+      "loss": 0.1372,
+      "step": 15510
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.784627346440076e-06,
+      "loss": 0.1518,
+      "step": 15515
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.7790736421192937e-06,
+      "loss": 0.0386,
+      "step": 15520
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.7735199377985115e-06,
+      "loss": 0.0464,
+      "step": 15525
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.7679662334777298e-06,
+      "loss": 0.0519,
+      "step": 15530
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.762412529156948e-06,
+      "loss": 0.0514,
+      "step": 15535
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.756858824836166e-06,
+      "loss": 0.0301,
+      "step": 15540
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.751305120515384e-06,
+      "loss": 0.0627,
+      "step": 15545
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.745751416194602e-06,
+      "loss": 0.0464,
+      "step": 15550
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.74019771187382e-06,
+      "loss": 0.0729,
+      "step": 15555
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.7346440075530377e-06,
+      "loss": 0.1235,
+      "step": 15560
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.7290903032322564e-06,
+      "loss": 0.1336,
+      "step": 15565
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.723536598911474e-06,
+      "loss": 0.0393,
+      "step": 15570
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.717982894590692e-06,
+      "loss": 0.0449,
+      "step": 15575
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.7124291902699103e-06,
+      "loss": 0.0384,
+      "step": 15580
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.706875485949128e-06,
+      "loss": 0.0753,
+      "step": 15585
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.7013217816283464e-06,
+      "loss": 0.0265,
+      "step": 15590
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.6957680773075647e-06,
+      "loss": 0.0364,
+      "step": 15595
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.6902143729867825e-06,
+      "loss": 0.052,
+      "step": 15600
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.6846606686660003e-06,
+      "loss": 0.0789,
+      "step": 15605
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.679106964345218e-06,
+      "loss": 0.1592,
+      "step": 15610
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.6735532600244365e-06,
+      "loss": 0.1796,
+      "step": 15615
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.6679995557036547e-06,
+      "loss": 0.05,
+      "step": 15620
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.6624458513828726e-06,
+      "loss": 0.034,
+      "step": 15625
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.656892147062091e-06,
+      "loss": 0.0372,
+      "step": 15630
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.6513384427413087e-06,
+      "loss": 0.0522,
+      "step": 15635
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.6457847384205265e-06,
+      "loss": 0.0226,
+      "step": 15640
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.640231034099745e-06,
+      "loss": 0.0293,
+      "step": 15645
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.634677329778963e-06,
+      "loss": 0.0673,
+      "step": 15650
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.629123625458181e-06,
+      "loss": 0.0738,
+      "step": 15655
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.6235699211373987e-06,
+      "loss": 0.1373,
+      "step": 15660
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.618016216816617e-06,
+      "loss": 0.1167,
+      "step": 15665
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.612462512495835e-06,
+      "loss": 0.0361,
+      "step": 15670
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.606908808175053e-06,
+      "loss": 0.0439,
+      "step": 15675
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.6013551038542713e-06,
+      "loss": 0.0447,
+      "step": 15680
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.595801399533489e-06,
+      "loss": 0.0563,
+      "step": 15685
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.590247695212707e-06,
+      "loss": 0.0173,
+      "step": 15690
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.584693990891925e-06,
+      "loss": 0.0265,
+      "step": 15695
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.5791402865711427e-06,
+      "loss": 0.0462,
+      "step": 15700
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.5735865822503614e-06,
+      "loss": 0.0799,
+      "step": 15705
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.5680328779295792e-06,
+      "loss": 0.1394,
+      "step": 15710
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.562479173608797e-06,
+      "loss": 0.1238,
+      "step": 15715
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.5569254692880153e-06,
+      "loss": 0.04,
+      "step": 15720
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.551371764967233e-06,
+      "loss": 0.0533,
+      "step": 15725
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.545818060646452e-06,
+      "loss": 0.0479,
+      "step": 15730
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.5402643563256697e-06,
+      "loss": 0.0505,
+      "step": 15735
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.5347106520048875e-06,
+      "loss": 0.0159,
+      "step": 15740
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.5291569476841054e-06,
+      "loss": 0.0715,
+      "step": 15745
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.5236032433633232e-06,
+      "loss": 0.0635,
+      "step": 15750
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.5180495390425415e-06,
+      "loss": 0.0959,
+      "step": 15755
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.5124958347217597e-06,
+      "loss": 0.1654,
+      "step": 15760
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.5069421304009776e-06,
+      "loss": 0.1383,
+      "step": 15765
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.501388426080196e-06,
+      "loss": 0.0436,
+      "step": 15770
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.4958347217594137e-06,
+      "loss": 0.0521,
+      "step": 15775
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.490281017438632e-06,
+      "loss": 0.0456,
+      "step": 15780
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.48472731311785e-06,
+      "loss": 0.047,
+      "step": 15785
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.4791736087970676e-06,
+      "loss": 0.0474,
+      "step": 15790
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.473619904476286e-06,
+      "loss": 0.0398,
+      "step": 15795
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.4680662001555037e-06,
+      "loss": 0.0525,
+      "step": 15800
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.462512495834722e-06,
+      "loss": 0.08,
+      "step": 15805
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.4569587915139403e-06,
+      "loss": 0.1447,
+      "step": 15810
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.451405087193158e-06,
+      "loss": 0.153,
+      "step": 15815
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.445851382872376e-06,
+      "loss": 0.0412,
+      "step": 15820
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.440297678551594e-06,
+      "loss": 0.0402,
+      "step": 15825
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.434743974230812e-06,
+      "loss": 0.0418,
+      "step": 15830
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.4291902699100303e-06,
+      "loss": 0.0662,
+      "step": 15835
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.423636565589248e-06,
+      "loss": 0.026,
+      "step": 15840
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.418082861268466e-06,
+      "loss": 0.0388,
+      "step": 15845
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.4125291569476843e-06,
+      "loss": 0.0374,
+      "step": 15850
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.4069754526269025e-06,
+      "loss": 0.0729,
+      "step": 15855
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.4014217483061204e-06,
+      "loss": 0.1568,
+      "step": 15860
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.3958680439853386e-06,
+      "loss": 0.1509,
+      "step": 15865
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.3903143396645565e-06,
+      "loss": 0.0437,
+      "step": 15870
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.3847606353437743e-06,
+      "loss": 0.0565,
+      "step": 15875
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.3792069310229926e-06,
+      "loss": 0.0328,
+      "step": 15880
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.3736532267022104e-06,
+      "loss": 0.0605,
+      "step": 15885
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.3680995223814287e-06,
+      "loss": 0.0254,
+      "step": 15890
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.3625458180606465e-06,
+      "loss": 0.0334,
+      "step": 15895
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.3569921137398648e-06,
+      "loss": 0.0632,
+      "step": 15900
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.351438409419083e-06,
+      "loss": 0.0843,
+      "step": 15905
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.345884705098301e-06,
+      "loss": 0.1359,
+      "step": 15910
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.3403310007775187e-06,
+      "loss": 0.1622,
+      "step": 15915
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.334777296456737e-06,
+      "loss": 0.0342,
+      "step": 15920
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.329223592135955e-06,
+      "loss": 0.0438,
+      "step": 15925
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.3236698878151727e-06,
+      "loss": 0.0463,
+      "step": 15930
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.318116183494391e-06,
+      "loss": 0.0425,
+      "step": 15935
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.312562479173609e-06,
+      "loss": 0.0385,
+      "step": 15940
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.307008774852827e-06,
+      "loss": 0.0203,
+      "step": 15945
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.3014550705320453e-06,
+      "loss": 0.0516,
+      "step": 15950
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.295901366211263e-06,
+      "loss": 0.0605,
+      "step": 15955
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.290347661890481e-06,
+      "loss": 0.1229,
+      "step": 15960
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.2847939575696992e-06,
+      "loss": 0.129,
+      "step": 15965
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.279240253248917e-06,
+      "loss": 0.0423,
+      "step": 15970
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.273686548928135e-06,
+      "loss": 0.0397,
+      "step": 15975
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.268132844607353e-06,
+      "loss": 0.0507,
+      "step": 15980
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.2625791402865714e-06,
+      "loss": 0.0677,
+      "step": 15985
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.2570254359657893e-06,
+      "loss": 0.0126,
+      "step": 15990
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.2514717316450076e-06,
+      "loss": 0.0488,
+      "step": 15995
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.2459180273242254e-06,
+      "loss": 0.0316,
+      "step": 16000
+    },
+    {
+      "epoch": 3.55,
+      "eval_loss": 0.023078329861164093,
+      "eval_runtime": 1769.0465,
+      "eval_samples_per_second": 9.055,
+      "eval_steps_per_second": 2.264,
+      "eval_wer": 0.04107464839523981,
+      "step": 16000
+    }
+  ],
+  "max_steps": 18016,
+  "num_train_epochs": 4,
+  "total_flos": 1.580537608226449e+19,
+  "trial_name": null,
+  "trial_params": null
+}