diff --git "a/last-checkpoint/trainer_state.json" "b/last-checkpoint/trainer_state.json"
--- "a/last-checkpoint/trainer_state.json"
+++ "b/last-checkpoint/trainer_state.json"
@@ -1,9 +1,9 @@
 {
   "best_metric": 0.2962397336959839,
   "best_model_checkpoint": "chickens-composite-201616161616-150-epochs-wo-transform/checkpoint-45000",
-  "epoch": 112.0,
+  "epoch": 150.0,
   "eval_steps": 500,
-  "global_step": 56000,
+  "global_step": 75000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -15981,6 +15981,5432 @@
       "eval_samples_per_second": 8.843,
       "eval_steps_per_second": 1.15,
       "step": 56000
+    },
+    {
+      "epoch": 112.02,
+      "grad_norm": 31.33954620361328,
+      "learning_rate": 1.5001872163025954e-06,
+      "loss": 0.199,
+      "step": 56010
+    },
+    {
+      "epoch": 112.08,
+      "grad_norm": 77.1370620727539,
+      "learning_rate": 1.4957026584154926e-06,
+      "loss": 0.2169,
+      "step": 56040
+    },
+    {
+      "epoch": 112.14,
+      "grad_norm": 57.998966217041016,
+      "learning_rate": 1.4912236342922143e-06,
+      "loss": 0.2132,
+      "step": 56070
+    },
+    {
+      "epoch": 112.2,
+      "grad_norm": 82.21308898925781,
+      "learning_rate": 1.4867501510057548e-06,
+      "loss": 0.2074,
+      "step": 56100
+    },
+    {
+      "epoch": 112.26,
+      "grad_norm": 52.17494201660156,
+      "learning_rate": 1.482282215620352e-06,
+      "loss": 0.201,
+      "step": 56130
+    },
+    {
+      "epoch": 112.32,
+      "grad_norm": 65.00828552246094,
+      "learning_rate": 1.4778198351914853e-06,
+      "loss": 0.2029,
+      "step": 56160
+    },
+    {
+      "epoch": 112.38,
+      "grad_norm": 208.12747192382812,
+      "learning_rate": 1.4733630167658652e-06,
+      "loss": 0.2259,
+      "step": 56190
+    },
+    {
+      "epoch": 112.44,
+      "grad_norm": 85.68213653564453,
+      "learning_rate": 1.4689117673814135e-06,
+      "loss": 0.2179,
+      "step": 56220
+    },
+    {
+      "epoch": 112.5,
+      "grad_norm": 45.39154052734375,
+      "learning_rate": 1.4644660940672628e-06,
+      "loss": 0.2035,
+      "step": 56250
+    },
+    {
+      "epoch": 112.56,
+      "grad_norm": 85.99728393554688,
+      "learning_rate": 1.4600260038437376e-06,
+      "loss": 0.2346,
+      "step": 56280
+    },
+    {
+      "epoch": 112.62,
+      "grad_norm": 56.224971771240234,
+      "learning_rate": 1.4555915037223438e-06,
+      "loss": 0.253,
+      "step": 56310
+    },
+    {
+      "epoch": 112.68,
+      "grad_norm": 40.9771728515625,
+      "learning_rate": 1.4511626007057667e-06,
+      "loss": 0.2058,
+      "step": 56340
+    },
+    {
+      "epoch": 112.74,
+      "grad_norm": 50.493865966796875,
+      "learning_rate": 1.4467393017878444e-06,
+      "loss": 0.2431,
+      "step": 56370
+    },
+    {
+      "epoch": 112.8,
+      "grad_norm": 63.593685150146484,
+      "learning_rate": 1.4423216139535735e-06,
+      "loss": 0.2349,
+      "step": 56400
+    },
+    {
+      "epoch": 112.86,
+      "grad_norm": 57.44195556640625,
+      "learning_rate": 1.4379095441790847e-06,
+      "loss": 0.2371,
+      "step": 56430
+    },
+    {
+      "epoch": 112.92,
+      "grad_norm": 40.976165771484375,
+      "learning_rate": 1.4335030994316357e-06,
+      "loss": 0.2195,
+      "step": 56460
+    },
+    {
+      "epoch": 112.98,
+      "grad_norm": 40.92074966430664,
+      "learning_rate": 1.4291022866696086e-06,
+      "loss": 0.2108,
+      "step": 56490
+    },
+    {
+      "epoch": 113.0,
+      "eval_loss": 0.31088173389434814,
+      "eval_map": 0.8054,
+      "eval_map_50": 0.9499,
+      "eval_map_75": 0.9015,
+      "eval_map_chicken": 0.8076,
+      "eval_map_duck": 0.7398,
+      "eval_map_large": 0.8011,
+      "eval_map_medium": 0.8084,
+      "eval_map_plant": 0.8688,
+      "eval_map_small": 0.3634,
+      "eval_mar_1": 0.3177,
+      "eval_mar_10": 0.8407,
+      "eval_mar_100": 0.8441,
+      "eval_mar_100_chicken": 0.8468,
+      "eval_mar_100_duck": 0.7887,
+      "eval_mar_100_plant": 0.8967,
+      "eval_mar_large": 0.8296,
+      "eval_mar_medium": 0.8485,
+      "eval_mar_small": 0.4129,
+      "eval_runtime": 12.34,
+      "eval_samples_per_second": 8.104,
+      "eval_steps_per_second": 1.053,
+      "step": 56500
+    },
+    {
+      "epoch": 113.04,
+      "grad_norm": 55.92573165893555,
+      "learning_rate": 1.4247071128424838e-06,
+      "loss": 0.2265,
+      "step": 56520
+    },
+    {
+      "epoch": 113.1,
+      "grad_norm": 60.161190032958984,
+      "learning_rate": 1.420317584890844e-06,
+      "loss": 0.2015,
+      "step": 56550
+    },
+    {
+      "epoch": 113.16,
+      "grad_norm": 49.9560661315918,
+      "learning_rate": 1.4159337097463515e-06,
+      "loss": 0.2438,
+      "step": 56580
+    },
+    {
+      "epoch": 113.22,
+      "grad_norm": 83.88067626953125,
+      "learning_rate": 1.4115554943317416e-06,
+      "loss": 0.2644,
+      "step": 56610
+    },
+    {
+      "epoch": 113.28,
+      "grad_norm": 42.70439529418945,
+      "learning_rate": 1.407182945560817e-06,
+      "loss": 0.2174,
+      "step": 56640
+    },
+    {
+      "epoch": 113.34,
+      "grad_norm": 77.9223861694336,
+      "learning_rate": 1.402816070338427e-06,
+      "loss": 0.2076,
+      "step": 56670
+    },
+    {
+      "epoch": 113.4,
+      "grad_norm": 45.262393951416016,
+      "learning_rate": 1.3984548755604655e-06,
+      "loss": 0.2369,
+      "step": 56700
+    },
+    {
+      "epoch": 113.46,
+      "grad_norm": 61.057743072509766,
+      "learning_rate": 1.3940993681138533e-06,
+      "loss": 0.1969,
+      "step": 56730
+    },
+    {
+      "epoch": 113.52,
+      "grad_norm": 54.26447677612305,
+      "learning_rate": 1.38974955487653e-06,
+      "loss": 0.2151,
+      "step": 56760
+    },
+    {
+      "epoch": 113.58,
+      "grad_norm": 77.25838470458984,
+      "learning_rate": 1.3854054427174468e-06,
+      "loss": 0.2445,
+      "step": 56790
+    },
+    {
+      "epoch": 113.64,
+      "grad_norm": 27.18859100341797,
+      "learning_rate": 1.3810670384965469e-06,
+      "loss": 0.2098,
+      "step": 56820
+    },
+    {
+      "epoch": 113.7,
+      "grad_norm": 66.12752532958984,
+      "learning_rate": 1.3767343490647668e-06,
+      "loss": 0.2193,
+      "step": 56850
+    },
+    {
+      "epoch": 113.76,
+      "grad_norm": 31.070798873901367,
+      "learning_rate": 1.372407381264011e-06,
+      "loss": 0.2348,
+      "step": 56880
+    },
+    {
+      "epoch": 113.82,
+      "grad_norm": 86.24375915527344,
+      "learning_rate": 1.368086141927154e-06,
+      "loss": 0.2188,
+      "step": 56910
+    },
+    {
+      "epoch": 113.88,
+      "grad_norm": 142.45108032226562,
+      "learning_rate": 1.3637706378780209e-06,
+      "loss": 0.2058,
+      "step": 56940
+    },
+    {
+      "epoch": 113.94,
+      "grad_norm": 113.668212890625,
+      "learning_rate": 1.3594608759313832e-06,
+      "loss": 0.2353,
+      "step": 56970
+    },
+    {
+      "epoch": 114.0,
+      "grad_norm": 81.07337951660156,
+      "learning_rate": 1.3551568628929434e-06,
+      "loss": 0.2148,
+      "step": 57000
+    },
+    {
+      "epoch": 114.0,
+      "eval_loss": 0.3065720796585083,
+      "eval_map": 0.803,
+      "eval_map_50": 0.9505,
+      "eval_map_75": 0.9011,
+      "eval_map_chicken": 0.8097,
+      "eval_map_duck": 0.7335,
+      "eval_map_large": 0.8033,
+      "eval_map_medium": 0.8068,
+      "eval_map_plant": 0.8657,
+      "eval_map_small": 0.3337,
+      "eval_mar_1": 0.3192,
+      "eval_mar_10": 0.8388,
+      "eval_mar_100": 0.8421,
+      "eval_mar_100_chicken": 0.8484,
+      "eval_mar_100_duck": 0.7845,
+      "eval_mar_100_plant": 0.8933,
+      "eval_mar_large": 0.8336,
+      "eval_mar_medium": 0.8466,
+      "eval_mar_small": 0.3914,
+      "eval_runtime": 13.3798,
+      "eval_samples_per_second": 7.474,
+      "eval_steps_per_second": 0.972,
+      "step": 57000
+    },
+    {
+      "epoch": 114.06,
+      "grad_norm": 59.886474609375,
+      "learning_rate": 1.350858605559323e-06,
+      "loss": 0.2309,
+      "step": 57030
+    },
+    {
+      "epoch": 114.12,
+      "grad_norm": 57.11491775512695,
+      "learning_rate": 1.346566110718061e-06,
+      "loss": 0.2288,
+      "step": 57060
+    },
+    {
+      "epoch": 114.18,
+      "grad_norm": 43.07657241821289,
+      "learning_rate": 1.3422793851475907e-06,
+      "loss": 0.2108,
+      "step": 57090
+    },
+    {
+      "epoch": 114.24,
+      "grad_norm": 52.78602981567383,
+      "learning_rate": 1.337998435617235e-06,
+      "loss": 0.2891,
+      "step": 57120
+    },
+    {
+      "epoch": 114.3,
+      "grad_norm": 49.36983108520508,
+      "learning_rate": 1.333723268887201e-06,
+      "loss": 0.2174,
+      "step": 57150
+    },
+    {
+      "epoch": 114.36,
+      "grad_norm": 55.36530685424805,
+      "learning_rate": 1.3294538917085586e-06,
+      "loss": 0.2449,
+      "step": 57180
+    },
+    {
+      "epoch": 114.42,
+      "grad_norm": 73.97029876708984,
+      "learning_rate": 1.3251903108232362e-06,
+      "loss": 0.1986,
+      "step": 57210
+    },
+    {
+      "epoch": 114.48,
+      "grad_norm": 65.40650939941406,
+      "learning_rate": 1.3209325329640126e-06,
+      "loss": 0.2228,
+      "step": 57240
+    },
+    {
+      "epoch": 114.54,
+      "grad_norm": 57.859867095947266,
+      "learning_rate": 1.316680564854499e-06,
+      "loss": 0.2143,
+      "step": 57270
+    },
+    {
+      "epoch": 114.6,
+      "grad_norm": 80.3431625366211,
+      "learning_rate": 1.312434413209131e-06,
+      "loss": 0.2277,
+      "step": 57300
+    },
+    {
+      "epoch": 114.66,
+      "grad_norm": 124.59980010986328,
+      "learning_rate": 1.3081940847331658e-06,
+      "loss": 0.1979,
+      "step": 57330
+    },
+    {
+      "epoch": 114.72,
+      "grad_norm": 42.293277740478516,
+      "learning_rate": 1.3039595861226579e-06,
+      "loss": 0.2265,
+      "step": 57360
+    },
+    {
+      "epoch": 114.78,
+      "grad_norm": 55.87450408935547,
+      "learning_rate": 1.2997309240644607e-06,
+      "loss": 0.1922,
+      "step": 57390
+    },
+    {
+      "epoch": 114.84,
+      "grad_norm": 40.74118423461914,
+      "learning_rate": 1.2955081052362072e-06,
+      "loss": 0.2295,
+      "step": 57420
+    },
+    {
+      "epoch": 114.9,
+      "grad_norm": 80.4583740234375,
+      "learning_rate": 1.2912911363063048e-06,
+      "loss": 0.1982,
+      "step": 57450
+    },
+    {
+      "epoch": 114.96,
+      "grad_norm": 70.01721954345703,
+      "learning_rate": 1.2870800239339237e-06,
+      "loss": 0.2107,
+      "step": 57480
+    },
+    {
+      "epoch": 115.0,
+      "eval_loss": 0.30899959802627563,
+      "eval_map": 0.8052,
+      "eval_map_50": 0.9505,
+      "eval_map_75": 0.9017,
+      "eval_map_chicken": 0.809,
+      "eval_map_duck": 0.7345,
+      "eval_map_large": 0.802,
+      "eval_map_medium": 0.8074,
+      "eval_map_plant": 0.8722,
+      "eval_map_small": 0.359,
+      "eval_mar_1": 0.3198,
+      "eval_mar_10": 0.8409,
+      "eval_mar_100": 0.8438,
+      "eval_mar_100_chicken": 0.8456,
+      "eval_mar_100_duck": 0.7866,
+      "eval_mar_100_plant": 0.8991,
+      "eval_mar_large": 0.8363,
+      "eval_mar_medium": 0.8487,
+      "eval_mar_small": 0.4076,
+      "eval_runtime": 11.3887,
+      "eval_samples_per_second": 8.781,
+      "eval_steps_per_second": 1.141,
+      "step": 57500
+    },
+    {
+      "epoch": 115.02,
+      "grad_norm": 75.13191986083984,
+      "learning_rate": 1.2828747747689846e-06,
+      "loss": 0.2267,
+      "step": 57510
+    },
+    {
+      "epoch": 115.08,
+      "grad_norm": 38.007808685302734,
+      "learning_rate": 1.2786753954521508e-06,
+      "loss": 0.23,
+      "step": 57540
+    },
+    {
+      "epoch": 115.14,
+      "grad_norm": 51.685386657714844,
+      "learning_rate": 1.2744818926148157e-06,
+      "loss": 0.2341,
+      "step": 57570
+    },
+    {
+      "epoch": 115.2,
+      "grad_norm": 49.958229064941406,
+      "learning_rate": 1.2702942728790897e-06,
+      "loss": 0.2442,
+      "step": 57600
+    },
+    {
+      "epoch": 115.26,
+      "grad_norm": 24.888076782226562,
+      "learning_rate": 1.2661125428577998e-06,
+      "loss": 0.2252,
+      "step": 57630
+    },
+    {
+      "epoch": 115.32,
+      "grad_norm": 51.06270217895508,
+      "learning_rate": 1.2619367091544654e-06,
+      "loss": 0.2564,
+      "step": 57660
+    },
+    {
+      "epoch": 115.38,
+      "grad_norm": 51.42506408691406,
+      "learning_rate": 1.2577667783633007e-06,
+      "loss": 0.2233,
+      "step": 57690
+    },
+    {
+      "epoch": 115.44,
+      "grad_norm": 33.248722076416016,
+      "learning_rate": 1.2536027570691938e-06,
+      "loss": 0.2404,
+      "step": 57720
+    },
+    {
+      "epoch": 115.5,
+      "grad_norm": 51.07853698730469,
+      "learning_rate": 1.2494446518477022e-06,
+      "loss": 0.2312,
+      "step": 57750
+    },
+    {
+      "epoch": 115.56,
+      "grad_norm": 73.65450286865234,
+      "learning_rate": 1.2452924692650443e-06,
+      "loss": 0.2013,
+      "step": 57780
+    },
+    {
+      "epoch": 115.62,
+      "grad_norm": 27.405363082885742,
+      "learning_rate": 1.2411462158780791e-06,
+      "loss": 0.216,
+      "step": 57810
+    },
+    {
+      "epoch": 115.68,
+      "grad_norm": 62.627403259277344,
+      "learning_rate": 1.2370058982343109e-06,
+      "loss": 0.23,
+      "step": 57840
+    },
+    {
+      "epoch": 115.74,
+      "grad_norm": 39.418758392333984,
+      "learning_rate": 1.232871522871864e-06,
+      "loss": 0.2142,
+      "step": 57870
+    },
+    {
+      "epoch": 115.8,
+      "grad_norm": 44.806175231933594,
+      "learning_rate": 1.2287430963194807e-06,
+      "loss": 0.2075,
+      "step": 57900
+    },
+    {
+      "epoch": 115.86,
+      "grad_norm": 56.002132415771484,
+      "learning_rate": 1.2246206250965127e-06,
+      "loss": 0.2078,
+      "step": 57930
+    },
+    {
+      "epoch": 115.92,
+      "grad_norm": 137.88645935058594,
+      "learning_rate": 1.2205041157129017e-06,
+      "loss": 0.2048,
+      "step": 57960
+    },
+    {
+      "epoch": 115.98,
+      "grad_norm": 15.981607437133789,
+      "learning_rate": 1.2163935746691807e-06,
+      "loss": 0.2204,
+      "step": 57990
+    },
+    {
+      "epoch": 116.0,
+      "eval_loss": 0.31037890911102295,
+      "eval_map": 0.8033,
+      "eval_map_50": 0.9505,
+      "eval_map_75": 0.9073,
+      "eval_map_chicken": 0.8066,
+      "eval_map_duck": 0.7341,
+      "eval_map_large": 0.8085,
+      "eval_map_medium": 0.8036,
+      "eval_map_plant": 0.8693,
+      "eval_map_small": 0.3617,
+      "eval_mar_1": 0.3187,
+      "eval_mar_10": 0.8387,
+      "eval_mar_100": 0.8419,
+      "eval_mar_100_chicken": 0.8448,
+      "eval_mar_100_duck": 0.7845,
+      "eval_mar_100_plant": 0.8964,
+      "eval_mar_large": 0.8359,
+      "eval_mar_medium": 0.8446,
+      "eval_mar_small": 0.4143,
+      "eval_runtime": 12.9378,
+      "eval_samples_per_second": 7.729,
+      "eval_steps_per_second": 1.005,
+      "step": 58000
+    },
+    {
+      "epoch": 116.04,
+      "grad_norm": 202.0614471435547,
+      "learning_rate": 1.2122890084564542e-06,
+      "loss": 0.2063,
+      "step": 58020
+    },
+    {
+      "epoch": 116.1,
+      "grad_norm": 47.288394927978516,
+      "learning_rate": 1.2081904235563908e-06,
+      "loss": 0.24,
+      "step": 58050
+    },
+    {
+      "epoch": 116.16,
+      "grad_norm": 133.25660705566406,
+      "learning_rate": 1.204097826441218e-06,
+      "loss": 0.2194,
+      "step": 58080
+    },
+    {
+      "epoch": 116.22,
+      "grad_norm": 38.492435455322266,
+      "learning_rate": 1.200011223573702e-06,
+      "loss": 0.2196,
+      "step": 58110
+    },
+    {
+      "epoch": 116.28,
+      "grad_norm": 55.27732467651367,
+      "learning_rate": 1.1959306214071508e-06,
+      "loss": 0.2093,
+      "step": 58140
+    },
+    {
+      "epoch": 116.34,
+      "grad_norm": 22.916257858276367,
+      "learning_rate": 1.1918560263853902e-06,
+      "loss": 0.2325,
+      "step": 58170
+    },
+    {
+      "epoch": 116.4,
+      "grad_norm": 48.5063362121582,
+      "learning_rate": 1.18778744494276e-06,
+      "loss": 0.2011,
+      "step": 58200
+    },
+    {
+      "epoch": 116.46,
+      "grad_norm": 41.77861022949219,
+      "learning_rate": 1.1837248835041093e-06,
+      "loss": 0.2232,
+      "step": 58230
+    },
+    {
+      "epoch": 116.52,
+      "grad_norm": 63.43424987792969,
+      "learning_rate": 1.1796683484847731e-06,
+      "loss": 0.242,
+      "step": 58260
+    },
+    {
+      "epoch": 116.58,
+      "grad_norm": 28.93366813659668,
+      "learning_rate": 1.1756178462905782e-06,
+      "loss": 0.2271,
+      "step": 58290
+    },
+    {
+      "epoch": 116.64,
+      "grad_norm": 41.52293395996094,
+      "learning_rate": 1.1715733833178178e-06,
+      "loss": 0.2268,
+      "step": 58320
+    },
+    {
+      "epoch": 116.7,
+      "grad_norm": 51.30208969116211,
+      "learning_rate": 1.1675349659532514e-06,
+      "loss": 0.2221,
+      "step": 58350
+    },
+    {
+      "epoch": 116.76,
+      "grad_norm": 126.49063110351562,
+      "learning_rate": 1.1635026005740902e-06,
+      "loss": 0.2103,
+      "step": 58380
+    },
+    {
+      "epoch": 116.82,
+      "grad_norm": 24.768518447875977,
+      "learning_rate": 1.159476293547992e-06,
+      "loss": 0.2149,
+      "step": 58410
+    },
+    {
+      "epoch": 116.88,
+      "grad_norm": 36.70859909057617,
+      "learning_rate": 1.1554560512330437e-06,
+      "loss": 0.2211,
+      "step": 58440
+    },
+    {
+      "epoch": 116.94,
+      "grad_norm": 65.63562774658203,
+      "learning_rate": 1.1514418799777554e-06,
+      "loss": 0.1944,
+      "step": 58470
+    },
+    {
+      "epoch": 117.0,
+      "grad_norm": 48.93583297729492,
+      "learning_rate": 1.1474337861210543e-06,
+      "loss": 0.2192,
+      "step": 58500
+    },
+    {
+      "epoch": 117.0,
+      "eval_loss": 0.30984461307525635,
+      "eval_map": 0.8062,
+      "eval_map_50": 0.9504,
+      "eval_map_75": 0.8968,
+      "eval_map_chicken": 0.807,
+      "eval_map_duck": 0.7388,
+      "eval_map_large": 0.8052,
+      "eval_map_medium": 0.8092,
+      "eval_map_plant": 0.8726,
+      "eval_map_small": 0.3382,
+      "eval_mar_1": 0.321,
+      "eval_mar_10": 0.8416,
+      "eval_mar_100": 0.8454,
+      "eval_mar_100_chicken": 0.848,
+      "eval_mar_100_duck": 0.7897,
+      "eval_mar_100_plant": 0.8985,
+      "eval_mar_large": 0.8363,
+      "eval_mar_medium": 0.8511,
+      "eval_mar_small": 0.3948,
+      "eval_runtime": 12.8976,
+      "eval_samples_per_second": 7.753,
+      "eval_steps_per_second": 1.008,
+      "step": 58500
+    },
+    {
+      "epoch": 117.06,
+      "grad_norm": 64.96910858154297,
+      "learning_rate": 1.1434317759922664e-06,
+      "loss": 0.23,
+      "step": 58530
+    },
+    {
+      "epoch": 117.12,
+      "grad_norm": 57.87126922607422,
+      "learning_rate": 1.1394358559111101e-06,
+      "loss": 0.2053,
+      "step": 58560
+    },
+    {
+      "epoch": 117.18,
+      "grad_norm": 171.56883239746094,
+      "learning_rate": 1.135446032187692e-06,
+      "loss": 0.2277,
+      "step": 58590
+    },
+    {
+      "epoch": 117.24,
+      "grad_norm": 76.24279022216797,
+      "learning_rate": 1.1314623111224865e-06,
+      "loss": 0.2346,
+      "step": 58620
+    },
+    {
+      "epoch": 117.3,
+      "grad_norm": 43.50970458984375,
+      "learning_rate": 1.1274846990063314e-06,
+      "loss": 0.2009,
+      "step": 58650
+    },
+    {
+      "epoch": 117.36,
+      "grad_norm": 48.1057014465332,
+      "learning_rate": 1.1235132021204226e-06,
+      "loss": 0.2197,
+      "step": 58680
+    },
+    {
+      "epoch": 117.42,
+      "grad_norm": 34.25832748413086,
+      "learning_rate": 1.1195478267362924e-06,
+      "loss": 0.2065,
+      "step": 58710
+    },
+    {
+      "epoch": 117.48,
+      "grad_norm": 122.34854888916016,
+      "learning_rate": 1.1155885791158128e-06,
+      "loss": 0.2114,
+      "step": 58740
+    },
+    {
+      "epoch": 117.54,
+      "grad_norm": 57.05229187011719,
+      "learning_rate": 1.111635465511175e-06,
+      "loss": 0.218,
+      "step": 58770
+    },
+    {
+      "epoch": 117.6,
+      "grad_norm": 25.191370010375977,
+      "learning_rate": 1.1076884921648834e-06,
+      "loss": 0.2492,
+      "step": 58800
+    },
+    {
+      "epoch": 117.66,
+      "grad_norm": 87.51880645751953,
+      "learning_rate": 1.1037476653097501e-06,
+      "loss": 0.1996,
+      "step": 58830
+    },
+    {
+      "epoch": 117.72,
+      "grad_norm": 45.07328796386719,
+      "learning_rate": 1.0998129911688766e-06,
+      "loss": 0.2378,
+      "step": 58860
+    },
+    {
+      "epoch": 117.78,
+      "grad_norm": 29.409072875976562,
+      "learning_rate": 1.0958844759556525e-06,
+      "loss": 0.2098,
+      "step": 58890
+    },
+    {
+      "epoch": 117.84,
+      "grad_norm": 18.03131675720215,
+      "learning_rate": 1.0919621258737384e-06,
+      "loss": 0.2214,
+      "step": 58920
+    },
+    {
+      "epoch": 117.9,
+      "grad_norm": 47.66557693481445,
+      "learning_rate": 1.0880459471170597e-06,
+      "loss": 0.2387,
+      "step": 58950
+    },
+    {
+      "epoch": 117.96,
+      "grad_norm": 30.374011993408203,
+      "learning_rate": 1.0841359458697986e-06,
+      "loss": 0.2384,
+      "step": 58980
+    },
+    {
+      "epoch": 118.0,
+      "eval_loss": 0.3125670850276947,
+      "eval_map": 0.8051,
+      "eval_map_50": 0.9507,
+      "eval_map_75": 0.9077,
+      "eval_map_chicken": 0.8043,
+      "eval_map_duck": 0.7379,
+      "eval_map_large": 0.8069,
+      "eval_map_medium": 0.8085,
+      "eval_map_plant": 0.8732,
+      "eval_map_small": 0.3702,
+      "eval_mar_1": 0.3187,
+      "eval_mar_10": 0.8387,
+      "eval_mar_100": 0.8421,
+      "eval_mar_100_chicken": 0.8421,
+      "eval_mar_100_duck": 0.7845,
+      "eval_mar_100_plant": 0.8997,
+      "eval_mar_large": 0.8358,
+      "eval_mar_medium": 0.8476,
+      "eval_mar_small": 0.4219,
+      "eval_runtime": 13.1737,
+      "eval_samples_per_second": 7.591,
+      "eval_steps_per_second": 0.987,
+      "step": 59000
+    },
+    {
+      "epoch": 118.02,
+      "grad_norm": 57.89711380004883,
+      "learning_rate": 1.0802321283063794e-06,
+      "loss": 0.211,
+      "step": 59010
+    },
+    {
+      "epoch": 118.08,
+      "grad_norm": 89.98204803466797,
+      "learning_rate": 1.0763345005914649e-06,
+      "loss": 0.2281,
+      "step": 59040
+    },
+    {
+      "epoch": 118.14,
+      "grad_norm": 37.554039001464844,
+      "learning_rate": 1.0724430688799402e-06,
+      "loss": 0.199,
+      "step": 59070
+    },
+    {
+      "epoch": 118.2,
+      "grad_norm": 43.45093536376953,
+      "learning_rate": 1.0685578393169054e-06,
+      "loss": 0.2155,
+      "step": 59100
+    },
+    {
+      "epoch": 118.26,
+      "grad_norm": 43.8154411315918,
+      "learning_rate": 1.0646788180376716e-06,
+      "loss": 0.2171,
+      "step": 59130
+    },
+    {
+      "epoch": 118.32,
+      "grad_norm": 65.75592803955078,
+      "learning_rate": 1.0608060111677409e-06,
+      "loss": 0.2068,
+      "step": 59160
+    },
+    {
+      "epoch": 118.38,
+      "grad_norm": 30.5941104888916,
+      "learning_rate": 1.0569394248228026e-06,
+      "loss": 0.2257,
+      "step": 59190
+    },
+    {
+      "epoch": 118.44,
+      "grad_norm": 107.75626373291016,
+      "learning_rate": 1.053079065108728e-06,
+      "loss": 0.2102,
+      "step": 59220
+    },
+    {
+      "epoch": 118.5,
+      "grad_norm": 43.346866607666016,
+      "learning_rate": 1.049224938121548e-06,
+      "loss": 0.2367,
+      "step": 59250
+    },
+    {
+      "epoch": 118.56,
+      "grad_norm": 62.630615234375,
+      "learning_rate": 1.0453770499474585e-06,
+      "loss": 0.2341,
+      "step": 59280
+    },
+    {
+      "epoch": 118.62,
+      "grad_norm": 64.70538330078125,
+      "learning_rate": 1.0415354066627993e-06,
+      "loss": 0.2207,
+      "step": 59310
+    },
+    {
+      "epoch": 118.68,
+      "grad_norm": 35.14129638671875,
+      "learning_rate": 1.037700014334047e-06,
+      "loss": 0.2262,
+      "step": 59340
+    },
+    {
+      "epoch": 118.74,
+      "grad_norm": 77.49797821044922,
+      "learning_rate": 1.0338708790178136e-06,
+      "loss": 0.2323,
+      "step": 59370
+    },
+    {
+      "epoch": 118.8,
+      "grad_norm": 33.764774322509766,
+      "learning_rate": 1.0300480067608232e-06,
+      "loss": 0.2213,
+      "step": 59400
+    },
+    {
+      "epoch": 118.86,
+      "grad_norm": 67.69646453857422,
+      "learning_rate": 1.026231403599915e-06,
+      "loss": 0.206,
+      "step": 59430
+    },
+    {
+      "epoch": 118.92,
+      "grad_norm": 26.622072219848633,
+      "learning_rate": 1.0224210755620257e-06,
+      "loss": 0.2049,
+      "step": 59460
+    },
+    {
+      "epoch": 118.98,
+      "grad_norm": 28.23842430114746,
+      "learning_rate": 1.0186170286641816e-06,
+      "loss": 0.2001,
+      "step": 59490
+    },
+    {
+      "epoch": 119.0,
+      "eval_loss": 0.3057597875595093,
+      "eval_map": 0.8046,
+      "eval_map_50": 0.9506,
+      "eval_map_75": 0.9047,
+      "eval_map_chicken": 0.804,
+      "eval_map_duck": 0.7374,
+      "eval_map_large": 0.8057,
+      "eval_map_medium": 0.8089,
+      "eval_map_plant": 0.8725,
+      "eval_map_small": 0.3726,
+      "eval_mar_1": 0.3203,
+      "eval_mar_10": 0.8405,
+      "eval_mar_100": 0.8437,
+      "eval_mar_100_chicken": 0.8437,
+      "eval_mar_100_duck": 0.7887,
+      "eval_mar_100_plant": 0.8988,
+      "eval_mar_large": 0.8301,
+      "eval_mar_medium": 0.8507,
+      "eval_mar_small": 0.4205,
+      "eval_runtime": 12.7193,
+      "eval_samples_per_second": 7.862,
+      "eval_steps_per_second": 1.022,
+      "step": 59500
+    },
+    {
+      "epoch": 119.04,
+      "grad_norm": 43.927146911621094,
+      "learning_rate": 1.014819268913495e-06,
+      "loss": 0.2077,
+      "step": 59520
+    },
+    {
+      "epoch": 119.1,
+      "grad_norm": 49.40350341796875,
+      "learning_rate": 1.0110278023071445e-06,
+      "loss": 0.2033,
+      "step": 59550
+    },
+    {
+      "epoch": 119.16,
+      "grad_norm": 237.99090576171875,
+      "learning_rate": 1.0072426348323754e-06,
+      "loss": 0.246,
+      "step": 59580
+    },
+    {
+      "epoch": 119.22,
+      "grad_norm": 38.9443244934082,
+      "learning_rate": 1.0034637724664832e-06,
+      "loss": 0.2126,
+      "step": 59610
+    },
+    {
+      "epoch": 119.28,
+      "grad_norm": 67.23509979248047,
+      "learning_rate": 9.99691221176805e-07,
+      "loss": 0.1921,
+      "step": 59640
+    },
+    {
+      "epoch": 119.34,
+      "grad_norm": 47.9791259765625,
+      "learning_rate": 9.959249869207177e-07,
+      "loss": 0.2247,
+      "step": 59670
+    },
+    {
+      "epoch": 119.4,
+      "grad_norm": 98.65447235107422,
+      "learning_rate": 9.921650756456164e-07,
+      "loss": 0.2286,
+      "step": 59700
+    },
+    {
+      "epoch": 119.46,
+      "grad_norm": 50.2000846862793,
+      "learning_rate": 9.884114932889172e-07,
+      "loss": 0.2398,
+      "step": 59730
+    },
+    {
+      "epoch": 119.52,
+      "grad_norm": 32.53291702270508,
+      "learning_rate": 9.84664245778037e-07,
+      "loss": 0.2037,
+      "step": 59760
+    },
+    {
+      "epoch": 119.58,
+      "grad_norm": 50.67923355102539,
+      "learning_rate": 9.809233390303901e-07,
+      "loss": 0.2632,
+      "step": 59790
+    },
+    {
+      "epoch": 119.64,
+      "grad_norm": 63.43164825439453,
+      "learning_rate": 9.771887789533818e-07,
+      "loss": 0.2612,
+      "step": 59820
+    },
+    {
+      "epoch": 119.7,
+      "grad_norm": 34.49391174316406,
+      "learning_rate": 9.734605714443906e-07,
+      "loss": 0.2278,
+      "step": 59850
+    },
+    {
+      "epoch": 119.76,
+      "grad_norm": 69.05148315429688,
+      "learning_rate": 9.69738722390765e-07,
+      "loss": 0.1917,
+      "step": 59880
+    },
+    {
+      "epoch": 119.82,
+      "grad_norm": 138.90768432617188,
+      "learning_rate": 9.66023237669812e-07,
+      "loss": 0.2051,
+      "step": 59910
+    },
+    {
+      "epoch": 119.88,
+      "grad_norm": 67.18104553222656,
+      "learning_rate": 9.623141231487904e-07,
+      "loss": 0.187,
+      "step": 59940
+    },
+    {
+      "epoch": 119.94,
+      "grad_norm": 34.69009017944336,
+      "learning_rate": 9.586113846848982e-07,
+      "loss": 0.2267,
+      "step": 59970
+    },
+    {
+      "epoch": 120.0,
+      "grad_norm": 55.136592864990234,
+      "learning_rate": 9.549150281252633e-07,
+      "loss": 0.205,
+      "step": 60000
+    },
+    {
+      "epoch": 120.0,
+      "eval_loss": 0.3106890022754669,
+      "eval_map": 0.803,
+      "eval_map_50": 0.9504,
+      "eval_map_75": 0.9039,
+      "eval_map_chicken": 0.8033,
+      "eval_map_duck": 0.732,
+      "eval_map_large": 0.8106,
+      "eval_map_medium": 0.8072,
+      "eval_map_plant": 0.8737,
+      "eval_map_small": 0.3506,
+      "eval_mar_1": 0.3193,
+      "eval_mar_10": 0.8387,
+      "eval_mar_100": 0.8423,
+      "eval_mar_100_chicken": 0.8437,
+      "eval_mar_100_duck": 0.7845,
+      "eval_mar_100_plant": 0.8988,
+      "eval_mar_large": 0.8393,
+      "eval_mar_medium": 0.8474,
+      "eval_mar_small": 0.4043,
+      "eval_runtime": 13.2445,
+      "eval_samples_per_second": 7.55,
+      "eval_steps_per_second": 0.982,
+      "step": 60000
+    },
+    {
+      "epoch": 120.06,
+      "grad_norm": 64.85227966308594,
+      "learning_rate": 9.512250593069394e-07,
+      "loss": 0.2098,
+      "step": 60030
+    },
+    {
+      "epoch": 120.12,
+      "grad_norm": 90.1717529296875,
+      "learning_rate": 9.475414840568903e-07,
+      "loss": 0.2312,
+      "step": 60060
+    },
+    {
+      "epoch": 120.18,
+      "grad_norm": 56.66279602050781,
+      "learning_rate": 9.438643081919818e-07,
+      "loss": 0.2083,
+      "step": 60090
+    },
+    {
+      "epoch": 120.24,
+      "grad_norm": 77.5468521118164,
+      "learning_rate": 9.401935375189802e-07,
+      "loss": 0.208,
+      "step": 60120
+    },
+    {
+      "epoch": 120.3,
+      "grad_norm": 0.0028794463723897934,
+      "learning_rate": 9.365291778345303e-07,
+      "loss": 0.1874,
+      "step": 60150
+    },
+    {
+      "epoch": 120.36,
+      "grad_norm": 75.45403289794922,
+      "learning_rate": 9.32871234925159e-07,
+      "loss": 0.2362,
+      "step": 60180
+    },
+    {
+      "epoch": 120.42,
+      "grad_norm": 52.26958084106445,
+      "learning_rate": 9.29219714567256e-07,
+      "loss": 0.2354,
+      "step": 60210
+    },
+    {
+      "epoch": 120.48,
+      "grad_norm": 49.53770446777344,
+      "learning_rate": 9.255746225270689e-07,
+      "loss": 0.2693,
+      "step": 60240
+    },
+    {
+      "epoch": 120.54,
+      "grad_norm": 45.97019577026367,
+      "learning_rate": 9.21935964560699e-07,
+      "loss": 0.2156,
+      "step": 60270
+    },
+    {
+      "epoch": 120.6,
+      "grad_norm": 38.9113883972168,
+      "learning_rate": 9.183037464140804e-07,
+      "loss": 0.2108,
+      "step": 60300
+    },
+    {
+      "epoch": 120.66,
+      "grad_norm": 48.45543670654297,
+      "learning_rate": 9.146779738229838e-07,
+      "loss": 0.202,
+      "step": 60330
+    },
+    {
+      "epoch": 120.72,
+      "grad_norm": 109.9886703491211,
+      "learning_rate": 9.110586525129988e-07,
+      "loss": 0.2332,
+      "step": 60360
+    },
+    {
+      "epoch": 120.78,
+      "grad_norm": 36.915462493896484,
+      "learning_rate": 9.074457881995252e-07,
+      "loss": 0.2075,
+      "step": 60390
+    },
+    {
+      "epoch": 120.84,
+      "grad_norm": 52.093040466308594,
+      "learning_rate": 9.038393865877725e-07,
+      "loss": 0.2154,
+      "step": 60420
+    },
+    {
+      "epoch": 120.9,
+      "grad_norm": 119.32299041748047,
+      "learning_rate": 9.002394533727382e-07,
+      "loss": 0.2397,
+      "step": 60450
+    },
+    {
+      "epoch": 120.96,
+      "grad_norm": 52.55674743652344,
+      "learning_rate": 8.966459942392108e-07,
+      "loss": 0.2247,
+      "step": 60480
+    },
+    {
+      "epoch": 121.0,
+      "eval_loss": 0.3103560507297516,
+      "eval_map": 0.8078,
+      "eval_map_50": 0.9503,
+      "eval_map_75": 0.9034,
+      "eval_map_chicken": 0.8084,
+      "eval_map_duck": 0.7372,
+      "eval_map_large": 0.806,
+      "eval_map_medium": 0.8085,
+      "eval_map_plant": 0.8779,
+      "eval_map_small": 0.4008,
+      "eval_mar_1": 0.321,
+      "eval_mar_10": 0.8437,
+      "eval_mar_100": 0.847,
+      "eval_mar_100_chicken": 0.8464,
+      "eval_mar_100_duck": 0.7928,
+      "eval_mar_100_plant": 0.9018,
+      "eval_mar_large": 0.8377,
+      "eval_mar_medium": 0.851,
+      "eval_mar_small": 0.4462,
+      "eval_runtime": 11.5428,
+      "eval_samples_per_second": 8.663,
+      "eval_steps_per_second": 1.126,
+      "step": 60500
+    },
+    {
+      "epoch": 121.02,
+      "grad_norm": 42.727779388427734,
+      "learning_rate": 8.930590148617513e-07,
+      "loss": 0.1956,
+      "step": 60510
+    },
+    {
+      "epoch": 121.08,
+      "grad_norm": 57.22825622558594,
+      "learning_rate": 8.894785209046886e-07,
+      "loss": 0.2028,
+      "step": 60540
+    },
+    {
+      "epoch": 121.14,
+      "grad_norm": 90.83641052246094,
+      "learning_rate": 8.859045180221137e-07,
+      "loss": 0.2213,
+      "step": 60570
+    },
+    {
+      "epoch": 121.2,
+      "grad_norm": 31.503341674804688,
+      "learning_rate": 8.823370118578628e-07,
+      "loss": 0.2004,
+      "step": 60600
+    },
+    {
+      "epoch": 121.26,
+      "grad_norm": 70.24781799316406,
+      "learning_rate": 8.787760080455171e-07,
+      "loss": 0.2337,
+      "step": 60630
+    },
+    {
+      "epoch": 121.32,
+      "grad_norm": 39.60617446899414,
+      "learning_rate": 8.752215122083874e-07,
+      "loss": 0.2097,
+      "step": 60660
+    },
+    {
+      "epoch": 121.38,
+      "grad_norm": 53.7212028503418,
+      "learning_rate": 8.716735299595059e-07,
+      "loss": 0.2256,
+      "step": 60690
+    },
+    {
+      "epoch": 121.44,
+      "grad_norm": 45.7958984375,
+      "learning_rate": 8.68132066901623e-07,
+      "loss": 0.2443,
+      "step": 60720
+    },
+    {
+      "epoch": 121.5,
+      "grad_norm": 79.56392669677734,
+      "learning_rate": 8.645971286271903e-07,
+      "loss": 0.2151,
+      "step": 60750
+    },
+    {
+      "epoch": 121.56,
+      "grad_norm": 32.804012298583984,
+      "learning_rate": 8.610687207183604e-07,
+      "loss": 0.1883,
+      "step": 60780
+    },
+    {
+      "epoch": 121.62,
+      "grad_norm": 30.23906707763672,
+      "learning_rate": 8.575468487469696e-07,
+      "loss": 0.2344,
+      "step": 60810
+    },
+    {
+      "epoch": 121.68,
+      "grad_norm": 62.960548400878906,
+      "learning_rate": 8.540315182745329e-07,
+      "loss": 0.2359,
+      "step": 60840
+    },
+    {
+      "epoch": 121.74,
+      "grad_norm": 61.267372131347656,
+      "learning_rate": 8.505227348522404e-07,
+      "loss": 0.2399,
+      "step": 60870
+    },
+    {
+      "epoch": 121.8,
+      "grad_norm": 87.58113098144531,
+      "learning_rate": 8.470205040209362e-07,
+      "loss": 0.1891,
+      "step": 60900
+    },
+    {
+      "epoch": 121.86,
+      "grad_norm": 52.148704528808594,
+      "learning_rate": 8.435248313111244e-07,
+      "loss": 0.2217,
+      "step": 60930
+    },
+    {
+      "epoch": 121.92,
+      "grad_norm": 40.73577880859375,
+      "learning_rate": 8.400357222429473e-07,
+      "loss": 0.2098,
+      "step": 60960
+    },
+    {
+      "epoch": 121.98,
+      "grad_norm": 74.6595687866211,
+      "learning_rate": 8.365531823261841e-07,
+      "loss": 0.2042,
+      "step": 60990
+    },
+    {
+      "epoch": 122.0,
+      "eval_loss": 0.31579113006591797,
+      "eval_map": 0.8016,
+      "eval_map_50": 0.9466,
+      "eval_map_75": 0.9042,
+      "eval_map_chicken": 0.8045,
+      "eval_map_duck": 0.7261,
+      "eval_map_large": 0.7924,
+      "eval_map_medium": 0.8066,
+      "eval_map_plant": 0.8743,
+      "eval_map_small": 0.3568,
+      "eval_mar_1": 0.3195,
+      "eval_mar_10": 0.8391,
+      "eval_mar_100": 0.8429,
+      "eval_mar_100_chicken": 0.846,
+      "eval_mar_100_duck": 0.7825,
+      "eval_mar_100_plant": 0.9003,
+      "eval_mar_large": 0.8266,
+      "eval_mar_medium": 0.8485,
+      "eval_mar_small": 0.4157,
+      "eval_runtime": 12.4336,
+      "eval_samples_per_second": 8.043,
+      "eval_steps_per_second": 1.046,
+      "step": 61000
+    },
+    {
+      "epoch": 122.04,
+      "grad_norm": 38.343040466308594,
+      "learning_rate": 8.330772170602424e-07,
+      "loss": 0.2252,
+      "step": 61020
+    },
+    {
+      "epoch": 122.1,
+      "grad_norm": 74.18119812011719,
+      "learning_rate": 8.296078319341444e-07,
+      "loss": 0.2137,
+      "step": 61050
+    },
+    {
+      "epoch": 122.16,
+      "grad_norm": 65.24456787109375,
+      "learning_rate": 8.261450324265225e-07,
+      "loss": 0.2098,
+      "step": 61080
+    },
+    {
+      "epoch": 122.22,
+      "grad_norm": 45.015743255615234,
+      "learning_rate": 8.226888240056114e-07,
+      "loss": 0.1982,
+      "step": 61110
+    },
+    {
+      "epoch": 122.28,
+      "grad_norm": 67.18856811523438,
+      "learning_rate": 8.192392121292336e-07,
+      "loss": 0.1895,
+      "step": 61140
+    },
+    {
+      "epoch": 122.34,
+      "grad_norm": 70.43605041503906,
+      "learning_rate": 8.157962022448001e-07,
+      "loss": 0.2065,
+      "step": 61170
+    },
+    {
+      "epoch": 122.4,
+      "grad_norm": 62.40924072265625,
+      "learning_rate": 8.123597997892918e-07,
+      "loss": 0.2141,
+      "step": 61200
+    },
+    {
+      "epoch": 122.46,
+      "grad_norm": 38.64645767211914,
+      "learning_rate": 8.089300101892561e-07,
+      "loss": 0.2097,
+      "step": 61230
+    },
+    {
+      "epoch": 122.52,
+      "grad_norm": 59.256107330322266,
+      "learning_rate": 8.055068388608011e-07,
+      "loss": 0.2103,
+      "step": 61260
+    },
+    {
+      "epoch": 122.58,
+      "grad_norm": 28.657575607299805,
+      "learning_rate": 8.020902912095807e-07,
+      "loss": 0.2334,
+      "step": 61290
+    },
+    {
+      "epoch": 122.64,
+      "grad_norm": 28.822608947753906,
+      "learning_rate": 7.986803726307901e-07,
+      "loss": 0.2003,
+      "step": 61320
+    },
+    {
+      "epoch": 122.7,
+      "grad_norm": 64.58875274658203,
+      "learning_rate": 7.952770885091548e-07,
+      "loss": 0.2058,
+      "step": 61350
+    },
+    {
+      "epoch": 122.76,
+      "grad_norm": 56.4006233215332,
+      "learning_rate": 7.918804442189271e-07,
+      "loss": 0.2472,
+      "step": 61380
+    },
+    {
+      "epoch": 122.82,
+      "grad_norm": 35.49840545654297,
+      "learning_rate": 7.884904451238712e-07,
+      "loss": 0.2192,
+      "step": 61410
+    },
+    {
+      "epoch": 122.88,
+      "grad_norm": 45.37355422973633,
+      "learning_rate": 7.851070965772572e-07,
+      "loss": 0.2269,
+      "step": 61440
+    },
+    {
+      "epoch": 122.94,
+      "grad_norm": 60.03096389770508,
+      "learning_rate": 7.81730403921856e-07,
+      "loss": 0.2392,
+      "step": 61470
+    },
+    {
+      "epoch": 123.0,
+      "grad_norm": 34.28696823120117,
+      "learning_rate": 7.783603724899258e-07,
+      "loss": 0.2029,
+      "step": 61500
+    },
+    {
+      "epoch": 123.0,
+      "eval_loss": 0.3060416281223297,
+      "eval_map": 0.8059,
+      "eval_map_50": 0.9503,
+      "eval_map_75": 0.8976,
+      "eval_map_chicken": 0.802,
+      "eval_map_duck": 0.7403,
+      "eval_map_large": 0.8096,
+      "eval_map_medium": 0.808,
+      "eval_map_plant": 0.8753,
+      "eval_map_small": 0.336,
+      "eval_mar_1": 0.3202,
+      "eval_mar_10": 0.8431,
+      "eval_mar_100": 0.8466,
+      "eval_mar_100_chicken": 0.846,
+      "eval_mar_100_duck": 0.7938,
+      "eval_mar_100_plant": 0.9,
+      "eval_mar_large": 0.8403,
+      "eval_mar_medium": 0.8509,
+      "eval_mar_small": 0.39,
+      "eval_runtime": 12.9518,
+      "eval_samples_per_second": 7.721,
+      "eval_steps_per_second": 1.004,
+      "step": 61500
+    },
+    {
+      "epoch": 123.06,
+      "grad_norm": 72.92243194580078,
+      "learning_rate": 7.749970076032048e-07,
+      "loss": 0.2276,
+      "step": 61530
+    },
+    {
+      "epoch": 123.12,
+      "grad_norm": 64.05213928222656,
+      "learning_rate": 7.716403145729073e-07,
+      "loss": 0.2084,
+      "step": 61560
+    },
+    {
+      "epoch": 123.18,
+      "grad_norm": 90.82645416259766,
+      "learning_rate": 7.682902986997076e-07,
+      "loss": 0.2284,
+      "step": 61590
+    },
+    {
+      "epoch": 123.24,
+      "grad_norm": 70.66219329833984,
+      "learning_rate": 7.649469652737407e-07,
+      "loss": 0.2002,
+      "step": 61620
+    },
+    {
+      "epoch": 123.3,
+      "grad_norm": 46.48184585571289,
+      "learning_rate": 7.61610319574585e-07,
+      "loss": 0.2054,
+      "step": 61650
+    },
+    {
+      "epoch": 123.36,
+      "grad_norm": 49.623992919921875,
+      "learning_rate": 7.582803668712579e-07,
+      "loss": 0.2303,
+      "step": 61680
+    },
+    {
+      "epoch": 123.42,
+      "grad_norm": 54.12312316894531,
+      "learning_rate": 7.549571124222127e-07,
+      "loss": 0.2509,
+      "step": 61710
+    },
+    {
+      "epoch": 123.48,
+      "grad_norm": 39.39469528198242,
+      "learning_rate": 7.51640561475318e-07,
+      "loss": 0.2076,
+      "step": 61740
+    },
+    {
+      "epoch": 123.54,
+      "grad_norm": 27.189287185668945,
+      "learning_rate": 7.48330719267864e-07,
+      "loss": 0.2316,
+      "step": 61770
+    },
+    {
+      "epoch": 123.6,
+      "grad_norm": 88.34009552001953,
+      "learning_rate": 7.450275910265415e-07,
+      "loss": 0.2575,
+      "step": 61800
+    },
+    {
+      "epoch": 123.66,
+      "grad_norm": 53.184043884277344,
+      "learning_rate": 7.4173118196744e-07,
+      "loss": 0.1986,
+      "step": 61830
+    },
+    {
+      "epoch": 123.72,
+      "grad_norm": 48.865074157714844,
+      "learning_rate": 7.384414972960419e-07,
+      "loss": 0.1832,
+      "step": 61860
+    },
+    {
+      "epoch": 123.78,
+      "grad_norm": 44.44317626953125,
+      "learning_rate": 7.351585422072049e-07,
+      "loss": 0.247,
+      "step": 61890
+    },
+    {
+      "epoch": 123.84,
+      "grad_norm": 73.54085540771484,
+      "learning_rate": 7.318823218851668e-07,
+      "loss": 0.1979,
+      "step": 61920
+    },
+    {
+      "epoch": 123.9,
+      "grad_norm": 90.45979309082031,
+      "learning_rate": 7.286128415035249e-07,
+      "loss": 0.2072,
+      "step": 61950
+    },
+    {
+      "epoch": 123.96,
+      "grad_norm": 74.4012680053711,
+      "learning_rate": 7.253501062252338e-07,
+      "loss": 0.2077,
+      "step": 61980
+    },
+    {
+      "epoch": 124.0,
+      "eval_loss": 0.31227219104766846,
+      "eval_map": 0.8028,
+      "eval_map_50": 0.9499,
+      "eval_map_75": 0.8959,
+      "eval_map_chicken": 0.8023,
+      "eval_map_duck": 0.7324,
+      "eval_map_large": 0.802,
+      "eval_map_medium": 0.8068,
+      "eval_map_plant": 0.8737,
+      "eval_map_small": 0.3735,
+      "eval_mar_1": 0.3185,
+      "eval_mar_10": 0.8399,
+      "eval_mar_100": 0.8435,
+      "eval_mar_100_chicken": 0.8437,
+      "eval_mar_100_duck": 0.7866,
+      "eval_mar_100_plant": 0.9003,
+      "eval_mar_large": 0.8301,
+      "eval_mar_medium": 0.85,
+      "eval_mar_small": 0.4319,
+      "eval_runtime": 13.0439,
+      "eval_samples_per_second": 7.666,
+      "eval_steps_per_second": 0.997,
+      "step": 62000
+    },
+    {
+      "epoch": 124.02,
+      "grad_norm": 56.22016525268555,
+      "learning_rate": 7.220941212026005e-07,
+      "loss": 0.1965,
+      "step": 62010
+    },
+    {
+      "epoch": 124.08,
+      "grad_norm": 50.742515563964844,
+      "learning_rate": 7.188448915772673e-07,
+      "loss": 0.2208,
+      "step": 62040
+    },
+    {
+      "epoch": 124.14,
+      "grad_norm": 27.431562423706055,
+      "learning_rate": 7.156024224802139e-07,
+      "loss": 0.1939,
+      "step": 62070
+    },
+    {
+      "epoch": 124.2,
+      "grad_norm": 33.407440185546875,
+      "learning_rate": 7.123667190317396e-07,
+      "loss": 0.2363,
+      "step": 62100
+    },
+    {
+      "epoch": 124.26,
+      "grad_norm": 60.38604736328125,
+      "learning_rate": 7.091377863414611e-07,
+      "loss": 0.2159,
+      "step": 62130
+    },
+    {
+      "epoch": 124.32,
+      "grad_norm": 28.370864868164062,
+      "learning_rate": 7.059156295083064e-07,
+      "loss": 0.2346,
+      "step": 62160
+    },
+    {
+      "epoch": 124.38,
+      "grad_norm": 32.50497817993164,
+      "learning_rate": 7.027002536204986e-07,
+      "loss": 0.1986,
+      "step": 62190
+    },
+    {
+      "epoch": 124.44,
+      "grad_norm": 66.82977294921875,
+      "learning_rate": 6.994916637555571e-07,
+      "loss": 0.2031,
+      "step": 62220
+    },
+    {
+      "epoch": 124.5,
+      "grad_norm": 85.54017639160156,
+      "learning_rate": 6.962898649802824e-07,
+      "loss": 0.1977,
+      "step": 62250
+    },
+    {
+      "epoch": 124.56,
+      "grad_norm": 66.37556457519531,
+      "learning_rate": 6.930948623507505e-07,
+      "loss": 0.2266,
+      "step": 62280
+    },
+    {
+      "epoch": 124.62,
+      "grad_norm": 55.63164138793945,
+      "learning_rate": 6.89906660912309e-07,
+      "loss": 0.2168,
+      "step": 62310
+    },
+    {
+      "epoch": 124.68,
+      "grad_norm": 42.36758804321289,
+      "learning_rate": 6.86725265699561e-07,
+      "loss": 0.2227,
+      "step": 62340
+    },
+    {
+      "epoch": 124.74,
+      "grad_norm": 220.3474578857422,
+      "learning_rate": 6.835506817363657e-07,
+      "loss": 0.1851,
+      "step": 62370
+    },
+    {
+      "epoch": 124.8,
+      "grad_norm": 40.383907318115234,
+      "learning_rate": 6.803829140358237e-07,
+      "loss": 0.2162,
+      "step": 62400
+    },
+    {
+      "epoch": 124.86,
+      "grad_norm": 64.54944610595703,
+      "learning_rate": 6.772219676002717e-07,
+      "loss": 0.2291,
+      "step": 62430
+    },
+    {
+      "epoch": 124.92,
+      "grad_norm": 80.85274505615234,
+      "learning_rate": 6.74067847421277e-07,
+      "loss": 0.2275,
+      "step": 62460
+    },
+    {
+      "epoch": 124.98,
+      "grad_norm": 103.29847717285156,
+      "learning_rate": 6.709205584796241e-07,
+      "loss": 0.2262,
+      "step": 62490
+    },
+    {
+      "epoch": 125.0,
+      "eval_loss": 0.3111739158630371,
+      "eval_map": 0.8043,
+      "eval_map_50": 0.9499,
+      "eval_map_75": 0.9056,
+      "eval_map_chicken": 0.8,
+      "eval_map_duck": 0.739,
+      "eval_map_large": 0.8116,
+      "eval_map_medium": 0.8063,
+      "eval_map_plant": 0.8741,
+      "eval_map_small": 0.3527,
+      "eval_mar_1": 0.3196,
+      "eval_mar_10": 0.8401,
+      "eval_mar_100": 0.8434,
+      "eval_mar_100_chicken": 0.8401,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.8994,
+      "eval_mar_large": 0.8383,
+      "eval_mar_medium": 0.8486,
+      "eval_mar_small": 0.3957,
+      "eval_runtime": 12.7565,
+      "eval_samples_per_second": 7.839,
+      "eval_steps_per_second": 1.019,
+      "step": 62500
+    },
+    {
+      "epoch": 125.04,
+      "grad_norm": 106.2325210571289,
+      "learning_rate": 6.677801057453143e-07,
+      "loss": 0.2064,
+      "step": 62520
+    },
+    {
+      "epoch": 125.1,
+      "grad_norm": 64.54126739501953,
+      "learning_rate": 6.646464941775499e-07,
+      "loss": 0.2138,
+      "step": 62550
+    },
+    {
+      "epoch": 125.16,
+      "grad_norm": 62.765010833740234,
+      "learning_rate": 6.615197287247299e-07,
+      "loss": 0.1982,
+      "step": 62580
+    },
+    {
+      "epoch": 125.22,
+      "grad_norm": 80.47004699707031,
+      "learning_rate": 6.583998143244463e-07,
+      "loss": 0.2141,
+      "step": 62610
+    },
+    {
+      "epoch": 125.28,
+      "grad_norm": 74.80974578857422,
+      "learning_rate": 6.552867559034687e-07,
+      "loss": 0.2398,
+      "step": 62640
+    },
+    {
+      "epoch": 125.34,
+      "grad_norm": 34.195377349853516,
+      "learning_rate": 6.521805583777396e-07,
+      "loss": 0.2437,
+      "step": 62670
+    },
+    {
+      "epoch": 125.4,
+      "grad_norm": 27.298751831054688,
+      "learning_rate": 6.490812266523716e-07,
+      "loss": 0.2042,
+      "step": 62700
+    },
+    {
+      "epoch": 125.46,
+      "grad_norm": 35.28337097167969,
+      "learning_rate": 6.459887656216318e-07,
+      "loss": 0.2129,
+      "step": 62730
+    },
+    {
+      "epoch": 125.52,
+      "grad_norm": 52.30793380737305,
+      "learning_rate": 6.429031801689362e-07,
+      "loss": 0.2216,
+      "step": 62760
+    },
+    {
+      "epoch": 125.58,
+      "grad_norm": 51.34511184692383,
+      "learning_rate": 6.398244751668481e-07,
+      "loss": 0.1996,
+      "step": 62790
+    },
+    {
+      "epoch": 125.64,
+      "grad_norm": 43.16114807128906,
+      "learning_rate": 6.36752655477062e-07,
+      "loss": 0.2037,
+      "step": 62820
+    },
+    {
+      "epoch": 125.7,
+      "grad_norm": 51.460853576660156,
+      "learning_rate": 6.336877259504004e-07,
+      "loss": 0.2291,
+      "step": 62850
+    },
+    {
+      "epoch": 125.76,
+      "grad_norm": 65.08202362060547,
+      "learning_rate": 6.30629691426804e-07,
+      "loss": 0.2223,
+      "step": 62880
+    },
+    {
+      "epoch": 125.82,
+      "grad_norm": 73.04798889160156,
+      "learning_rate": 6.275785567353293e-07,
+      "loss": 0.2204,
+      "step": 62910
+    },
+    {
+      "epoch": 125.88,
+      "grad_norm": 55.04792404174805,
+      "learning_rate": 6.245343266941328e-07,
+      "loss": 0.2026,
+      "step": 62940
+    },
+    {
+      "epoch": 125.94,
+      "grad_norm": 70.68724060058594,
+      "learning_rate": 6.214970061104686e-07,
+      "loss": 0.224,
+      "step": 62970
+    },
+    {
+      "epoch": 126.0,
+      "grad_norm": 51.036346435546875,
+      "learning_rate": 6.184665997806832e-07,
+      "loss": 0.2136,
+      "step": 63000
+    },
+    {
+      "epoch": 126.0,
+      "eval_loss": 0.3078014850616455,
+      "eval_map": 0.8045,
+      "eval_map_50": 0.9492,
+      "eval_map_75": 0.9035,
+      "eval_map_chicken": 0.8033,
+      "eval_map_duck": 0.7368,
+      "eval_map_large": 0.8059,
+      "eval_map_medium": 0.8086,
+      "eval_map_plant": 0.8735,
+      "eval_map_small": 0.3766,
+      "eval_mar_1": 0.3189,
+      "eval_mar_10": 0.8406,
+      "eval_mar_100": 0.8439,
+      "eval_mar_100_chicken": 0.844,
+      "eval_mar_100_duck": 0.7887,
+      "eval_mar_100_plant": 0.8991,
+      "eval_mar_large": 0.8334,
+      "eval_mar_medium": 0.8497,
+      "eval_mar_small": 0.4271,
+      "eval_runtime": 11.2037,
+      "eval_samples_per_second": 8.926,
+      "eval_steps_per_second": 1.16,
+      "step": 63000
+    },
+    {
+      "epoch": 126.06,
+      "grad_norm": 64.0252685546875,
+      "learning_rate": 6.154431124901983e-07,
+      "loss": 0.2113,
+      "step": 63030
+    },
+    {
+      "epoch": 126.12,
+      "grad_norm": 84.33251190185547,
+      "learning_rate": 6.124265490135161e-07,
+      "loss": 0.2276,
+      "step": 63060
+    },
+    {
+      "epoch": 126.18,
+      "grad_norm": 55.78408432006836,
+      "learning_rate": 6.094169141142014e-07,
+      "loss": 0.19,
+      "step": 63090
+    },
+    {
+      "epoch": 126.24,
+      "grad_norm": 39.02850341796875,
+      "learning_rate": 6.064142125448763e-07,
+      "loss": 0.2128,
+      "step": 63120
+    },
+    {
+      "epoch": 126.3,
+      "grad_norm": 38.10506057739258,
+      "learning_rate": 6.034184490472195e-07,
+      "loss": 0.203,
+      "step": 63150
+    },
+    {
+      "epoch": 126.36,
+      "grad_norm": 75.8965835571289,
+      "learning_rate": 6.004296283519478e-07,
+      "loss": 0.2176,
+      "step": 63180
+    },
+    {
+      "epoch": 126.42,
+      "grad_norm": 49.18169403076172,
+      "learning_rate": 5.974477551788194e-07,
+      "loss": 0.2212,
+      "step": 63210
+    },
+    {
+      "epoch": 126.48,
+      "grad_norm": 192.56015014648438,
+      "learning_rate": 5.944728342366179e-07,
+      "loss": 0.2368,
+      "step": 63240
+    },
+    {
+      "epoch": 126.54,
+      "grad_norm": 100.00894165039062,
+      "learning_rate": 5.915048702231491e-07,
+      "loss": 0.2315,
+      "step": 63270
+    },
+    {
+      "epoch": 126.6,
+      "grad_norm": 64.26509094238281,
+      "learning_rate": 5.885438678252342e-07,
+      "loss": 0.1936,
+      "step": 63300
+    },
+    {
+      "epoch": 126.66,
+      "grad_norm": 47.66757583618164,
+      "learning_rate": 5.855898317186992e-07,
+      "loss": 0.2149,
+      "step": 63330
+    },
+    {
+      "epoch": 126.72,
+      "grad_norm": 23.463951110839844,
+      "learning_rate": 5.826427665683715e-07,
+      "loss": 0.1777,
+      "step": 63360
+    },
+    {
+      "epoch": 126.78,
+      "grad_norm": 36.043025970458984,
+      "learning_rate": 5.797026770280683e-07,
+      "loss": 0.2227,
+      "step": 63390
+    },
+    {
+      "epoch": 126.84,
+      "grad_norm": 72.77320861816406,
+      "learning_rate": 5.767695677405921e-07,
+      "loss": 0.2175,
+      "step": 63420
+    },
+    {
+      "epoch": 126.9,
+      "grad_norm": 75.97449493408203,
+      "learning_rate": 5.738434433377244e-07,
+      "loss": 0.1979,
+      "step": 63450
+    },
+    {
+      "epoch": 126.96,
+      "grad_norm": 45.13546371459961,
+      "learning_rate": 5.709243084402128e-07,
+      "loss": 0.2446,
+      "step": 63480
+    },
+    {
+      "epoch": 127.0,
+      "eval_loss": 0.3091703951358795,
+      "eval_map": 0.8063,
+      "eval_map_50": 0.9502,
+      "eval_map_75": 0.9039,
+      "eval_map_chicken": 0.8081,
+      "eval_map_duck": 0.7388,
+      "eval_map_large": 0.7998,
+      "eval_map_medium": 0.8106,
+      "eval_map_plant": 0.872,
+      "eval_map_small": 0.3843,
+      "eval_mar_1": 0.3193,
+      "eval_mar_10": 0.8417,
+      "eval_mar_100": 0.8447,
+      "eval_mar_100_chicken": 0.8464,
+      "eval_mar_100_duck": 0.7897,
+      "eval_mar_100_plant": 0.8979,
+      "eval_mar_large": 0.8321,
+      "eval_mar_medium": 0.85,
+      "eval_mar_small": 0.4267,
+      "eval_runtime": 13.5522,
+      "eval_samples_per_second": 7.379,
+      "eval_steps_per_second": 0.959,
+      "step": 63500
+    },
+    {
+      "epoch": 127.02,
+      "grad_norm": 22.453636169433594,
+      "learning_rate": 5.680121676577721e-07,
+      "loss": 0.2081,
+      "step": 63510
+    },
+    {
+      "epoch": 127.08,
+      "grad_norm": 66.31897735595703,
+      "learning_rate": 5.651070255890689e-07,
+      "loss": 0.2167,
+      "step": 63540
+    },
+    {
+      "epoch": 127.14,
+      "grad_norm": 107.52672576904297,
+      "learning_rate": 5.622088868217179e-07,
+      "loss": 0.2255,
+      "step": 63570
+    },
+    {
+      "epoch": 127.2,
+      "grad_norm": 49.077579498291016,
+      "learning_rate": 5.593177559322776e-07,
+      "loss": 0.2447,
+      "step": 63600
+    },
+    {
+      "epoch": 127.26,
+      "grad_norm": 560.881591796875,
+      "learning_rate": 5.564336374862373e-07,
+      "loss": 0.193,
+      "step": 63630
+    },
+    {
+      "epoch": 127.32,
+      "grad_norm": 79.59185791015625,
+      "learning_rate": 5.535565360380146e-07,
+      "loss": 0.2024,
+      "step": 63660
+    },
+    {
+      "epoch": 127.38,
+      "grad_norm": 86.64059448242188,
+      "learning_rate": 5.506864561309455e-07,
+      "loss": 0.2036,
+      "step": 63690
+    },
+    {
+      "epoch": 127.44,
+      "grad_norm": 47.96025848388672,
+      "learning_rate": 5.478234022972756e-07,
+      "loss": 0.2284,
+      "step": 63720
+    },
+    {
+      "epoch": 127.5,
+      "grad_norm": 43.72831344604492,
+      "learning_rate": 5.449673790581611e-07,
+      "loss": 0.2129,
+      "step": 63750
+    },
+    {
+      "epoch": 127.56,
+      "grad_norm": 38.098045349121094,
+      "learning_rate": 5.421183909236494e-07,
+      "loss": 0.2136,
+      "step": 63780
+    },
+    {
+      "epoch": 127.62,
+      "grad_norm": 44.35323715209961,
+      "learning_rate": 5.392764423926844e-07,
+      "loss": 0.2376,
+      "step": 63810
+    },
+    {
+      "epoch": 127.68,
+      "grad_norm": 48.356201171875,
+      "learning_rate": 5.364415379530891e-07,
+      "loss": 0.2212,
+      "step": 63840
+    },
+    {
+      "epoch": 127.74,
+      "grad_norm": 33.646507263183594,
+      "learning_rate": 5.33613682081564e-07,
+      "loss": 0.185,
+      "step": 63870
+    },
+    {
+      "epoch": 127.8,
+      "grad_norm": 52.02384567260742,
+      "learning_rate": 5.307928792436812e-07,
+      "loss": 0.2146,
+      "step": 63900
+    },
+    {
+      "epoch": 127.86,
+      "grad_norm": 38.83525085449219,
+      "learning_rate": 5.279791338938717e-07,
+      "loss": 0.2158,
+      "step": 63930
+    },
+    {
+      "epoch": 127.92,
+      "grad_norm": 32.53917694091797,
+      "learning_rate": 5.251724504754258e-07,
+      "loss": 0.202,
+      "step": 63960
+    },
+    {
+      "epoch": 127.98,
+      "grad_norm": 40.70000457763672,
+      "learning_rate": 5.22372833420478e-07,
+      "loss": 0.2116,
+      "step": 63990
+    },
+    {
+      "epoch": 128.0,
+      "eval_loss": 0.30406132340431213,
+      "eval_map": 0.809,
+      "eval_map_50": 0.9479,
+      "eval_map_75": 0.9041,
+      "eval_map_chicken": 0.807,
+      "eval_map_duck": 0.7429,
+      "eval_map_large": 0.8049,
+      "eval_map_medium": 0.8146,
+      "eval_map_plant": 0.8771,
+      "eval_map_small": 0.3855,
+      "eval_mar_1": 0.3241,
+      "eval_mar_10": 0.844,
+      "eval_mar_100": 0.8471,
+      "eval_mar_100_chicken": 0.8476,
+      "eval_mar_100_duck": 0.7918,
+      "eval_mar_100_plant": 0.9018,
+      "eval_mar_large": 0.832,
+      "eval_mar_medium": 0.8535,
+      "eval_mar_small": 0.4233,
+      "eval_runtime": 14.4971,
+      "eval_samples_per_second": 6.898,
+      "eval_steps_per_second": 0.897,
+      "step": 64000
+    },
+    {
+      "epoch": 128.04,
+      "grad_norm": 49.22426223754883,
+      "learning_rate": 5.19580287150005e-07,
+      "loss": 0.2364,
+      "step": 64020
+    },
+    {
+      "epoch": 128.1,
+      "grad_norm": 30.956523895263672,
+      "learning_rate": 5.167948160738206e-07,
+      "loss": 0.2098,
+      "step": 64050
+    },
+    {
+      "epoch": 128.16,
+      "grad_norm": 46.6041145324707,
+      "learning_rate": 5.140164245905633e-07,
+      "loss": 0.1986,
+      "step": 64080
+    },
+    {
+      "epoch": 128.22,
+      "grad_norm": 50.73973846435547,
+      "learning_rate": 5.112451170876903e-07,
+      "loss": 0.2013,
+      "step": 64110
+    },
+    {
+      "epoch": 128.28,
+      "grad_norm": 53.47914505004883,
+      "learning_rate": 5.084808979414779e-07,
+      "loss": 0.2164,
+      "step": 64140
+    },
+    {
+      "epoch": 128.34,
+      "grad_norm": 53.10034942626953,
+      "learning_rate": 5.057237715170032e-07,
+      "loss": 0.2212,
+      "step": 64170
+    },
+    {
+      "epoch": 128.4,
+      "grad_norm": 39.50797653198242,
+      "learning_rate": 5.029737421681446e-07,
+      "loss": 0.2196,
+      "step": 64200
+    },
+    {
+      "epoch": 128.46,
+      "grad_norm": 125.84817504882812,
+      "learning_rate": 5.002308142375762e-07,
+      "loss": 0.2078,
+      "step": 64230
+    },
+    {
+      "epoch": 128.52,
+      "grad_norm": 66.1211929321289,
+      "learning_rate": 4.97494992056754e-07,
+      "loss": 0.2078,
+      "step": 64260
+    },
+    {
+      "epoch": 128.58,
+      "grad_norm": 77.76156616210938,
+      "learning_rate": 4.947662799459152e-07,
+      "loss": 0.2247,
+      "step": 64290
+    },
+    {
+      "epoch": 128.64,
+      "grad_norm": 41.21811294555664,
+      "learning_rate": 4.920446822140673e-07,
+      "loss": 0.2013,
+      "step": 64320
+    },
+    {
+      "epoch": 128.7,
+      "grad_norm": 48.723506927490234,
+      "learning_rate": 4.893302031589864e-07,
+      "loss": 0.2264,
+      "step": 64350
+    },
+    {
+      "epoch": 128.76,
+      "grad_norm": 48.01104736328125,
+      "learning_rate": 4.866228470672041e-07,
+      "loss": 0.2086,
+      "step": 64380
+    },
+    {
+      "epoch": 128.82,
+      "grad_norm": 52.20193862915039,
+      "learning_rate": 4.839226182140072e-07,
+      "loss": 0.2079,
+      "step": 64410
+    },
+    {
+      "epoch": 128.88,
+      "grad_norm": 30.023849487304688,
+      "learning_rate": 4.812295208634238e-07,
+      "loss": 0.2095,
+      "step": 64440
+    },
+    {
+      "epoch": 128.94,
+      "grad_norm": 44.33883285522461,
+      "learning_rate": 4.78543559268222e-07,
+      "loss": 0.227,
+      "step": 64470
+    },
+    {
+      "epoch": 129.0,
+      "grad_norm": 45.76134490966797,
+      "learning_rate": 4.758647376699033e-07,
+      "loss": 0.2194,
+      "step": 64500
+    },
+    {
+      "epoch": 129.0,
+      "eval_loss": 0.3098631203174591,
+      "eval_map": 0.809,
+      "eval_map_50": 0.9511,
+      "eval_map_75": 0.9049,
+      "eval_map_chicken": 0.8081,
+      "eval_map_duck": 0.7421,
+      "eval_map_large": 0.8153,
+      "eval_map_medium": 0.8112,
+      "eval_map_plant": 0.8767,
+      "eval_map_small": 0.3936,
+      "eval_mar_1": 0.3202,
+      "eval_mar_10": 0.8442,
+      "eval_mar_100": 0.8471,
+      "eval_mar_100_chicken": 0.8468,
+      "eval_mar_100_duck": 0.7928,
+      "eval_mar_100_plant": 0.9018,
+      "eval_mar_large": 0.8389,
+      "eval_mar_medium": 0.8508,
+      "eval_mar_small": 0.4448,
+      "eval_runtime": 13.9885,
+      "eval_samples_per_second": 7.149,
+      "eval_steps_per_second": 0.929,
+      "step": 64500
+    },
+    {
+      "epoch": 129.06,
+      "grad_norm": 78.0044174194336,
+      "learning_rate": 4.731930602986906e-07,
+      "loss": 0.2242,
+      "step": 64530
+    },
+    {
+      "epoch": 129.12,
+      "grad_norm": 99.87544250488281,
+      "learning_rate": 4.705285313735297e-07,
+      "loss": 0.2007,
+      "step": 64560
+    },
+    {
+      "epoch": 129.18,
+      "grad_norm": 43.78367233276367,
+      "learning_rate": 4.678711551020743e-07,
+      "loss": 0.1949,
+      "step": 64590
+    },
+    {
+      "epoch": 129.24,
+      "grad_norm": 41.33878707885742,
+      "learning_rate": 4.6522093568068307e-07,
+      "loss": 0.1895,
+      "step": 64620
+    },
+    {
+      "epoch": 129.3,
+      "grad_norm": 112.94983673095703,
+      "learning_rate": 4.625778772944156e-07,
+      "loss": 0.1941,
+      "step": 64650
+    },
+    {
+      "epoch": 129.36,
+      "grad_norm": 39.820884704589844,
+      "learning_rate": 4.599419841170216e-07,
+      "loss": 0.1935,
+      "step": 64680
+    },
+    {
+      "epoch": 129.42,
+      "grad_norm": 52.91804885864258,
+      "learning_rate": 4.5731326031093645e-07,
+      "loss": 0.2079,
+      "step": 64710
+    },
+    {
+      "epoch": 129.48,
+      "grad_norm": 65.22171020507812,
+      "learning_rate": 4.546917100272735e-07,
+      "loss": 0.2195,
+      "step": 64740
+    },
+    {
+      "epoch": 129.54,
+      "grad_norm": 39.4925651550293,
+      "learning_rate": 4.520773374058179e-07,
+      "loss": 0.2156,
+      "step": 64770
+    },
+    {
+      "epoch": 129.6,
+      "grad_norm": 55.07014846801758,
+      "learning_rate": 4.494701465750217e-07,
+      "loss": 0.2638,
+      "step": 64800
+    },
+    {
+      "epoch": 129.66,
+      "grad_norm": 45.676753997802734,
+      "learning_rate": 4.4687014165199547e-07,
+      "loss": 0.2218,
+      "step": 64830
+    },
+    {
+      "epoch": 129.72,
+      "grad_norm": 90.54066467285156,
+      "learning_rate": 4.4427732674250045e-07,
+      "loss": 0.1989,
+      "step": 64860
+    },
+    {
+      "epoch": 129.78,
+      "grad_norm": 101.72560119628906,
+      "learning_rate": 4.416917059409465e-07,
+      "loss": 0.2263,
+      "step": 64890
+    },
+    {
+      "epoch": 129.84,
+      "grad_norm": 36.34654998779297,
+      "learning_rate": 4.391132833303807e-07,
+      "loss": 0.2257,
+      "step": 64920
+    },
+    {
+      "epoch": 129.9,
+      "grad_norm": 62.203880310058594,
+      "learning_rate": 4.3654206298248625e-07,
+      "loss": 0.2127,
+      "step": 64950
+    },
+    {
+      "epoch": 129.96,
+      "grad_norm": 44.24311065673828,
+      "learning_rate": 4.3397804895756957e-07,
+      "loss": 0.2259,
+      "step": 64980
+    },
+    {
+      "epoch": 130.0,
+      "eval_loss": 0.30975058674812317,
+      "eval_map": 0.8074,
+      "eval_map_50": 0.951,
+      "eval_map_75": 0.9035,
+      "eval_map_chicken": 0.8069,
+      "eval_map_duck": 0.7433,
+      "eval_map_large": 0.8103,
+      "eval_map_medium": 0.808,
+      "eval_map_plant": 0.872,
+      "eval_map_small": 0.385,
+      "eval_mar_1": 0.323,
+      "eval_mar_10": 0.8424,
+      "eval_mar_100": 0.8457,
+      "eval_mar_100_chicken": 0.846,
+      "eval_mar_100_duck": 0.7928,
+      "eval_mar_100_plant": 0.8982,
+      "eval_mar_large": 0.8372,
+      "eval_mar_medium": 0.8483,
+      "eval_mar_small": 0.4367,
+      "eval_runtime": 12.5307,
+      "eval_samples_per_second": 7.98,
+      "eval_steps_per_second": 1.037,
+      "step": 65000
+    },
+    {
+      "epoch": 130.02,
+      "grad_norm": 26.39763069152832,
+      "learning_rate": 4.31421245304558e-07,
+      "loss": 0.2253,
+      "step": 65010
+    },
+    {
+      "epoch": 130.08,
+      "grad_norm": 103.76341247558594,
+      "learning_rate": 4.2887165606099513e-07,
+      "loss": 0.1987,
+      "step": 65040
+    },
+    {
+      "epoch": 130.14,
+      "grad_norm": 63.88904571533203,
+      "learning_rate": 4.263292852530293e-07,
+      "loss": 0.1979,
+      "step": 65070
+    },
+    {
+      "epoch": 130.2,
+      "grad_norm": 83.46735382080078,
+      "learning_rate": 4.237941368954124e-07,
+      "loss": 0.2374,
+      "step": 65100
+    },
+    {
+      "epoch": 130.26,
+      "grad_norm": 55.29493713378906,
+      "learning_rate": 4.212662149914887e-07,
+      "loss": 0.2058,
+      "step": 65130
+    },
+    {
+      "epoch": 130.32,
+      "grad_norm": 43.588871002197266,
+      "learning_rate": 4.1874552353319107e-07,
+      "loss": 0.2035,
+      "step": 65160
+    },
+    {
+      "epoch": 130.38,
+      "grad_norm": 36.1904296875,
+      "learning_rate": 4.162320665010372e-07,
+      "loss": 0.1872,
+      "step": 65190
+    },
+    {
+      "epoch": 130.44,
+      "grad_norm": 39.65531921386719,
+      "learning_rate": 4.137258478641176e-07,
+      "loss": 0.1862,
+      "step": 65220
+    },
+    {
+      "epoch": 130.5,
+      "grad_norm": 88.49795532226562,
+      "learning_rate": 4.112268715800943e-07,
+      "loss": 0.2218,
+      "step": 65250
+    },
+    {
+      "epoch": 130.56,
+      "grad_norm": 31.362567901611328,
+      "learning_rate": 4.087351415951918e-07,
+      "loss": 0.2277,
+      "step": 65280
+    },
+    {
+      "epoch": 130.62,
+      "grad_norm": 25.125764846801758,
+      "learning_rate": 4.062506618441908e-07,
+      "loss": 0.1873,
+      "step": 65310
+    },
+    {
+      "epoch": 130.68,
+      "grad_norm": 32.94488525390625,
+      "learning_rate": 4.0377343625042587e-07,
+      "loss": 0.2049,
+      "step": 65340
+    },
+    {
+      "epoch": 130.74,
+      "grad_norm": 79.44256591796875,
+      "learning_rate": 4.013034687257727e-07,
+      "loss": 0.221,
+      "step": 65370
+    },
+    {
+      "epoch": 130.8,
+      "grad_norm": 57.485897064208984,
+      "learning_rate": 3.9884076317064813e-07,
+      "loss": 0.2389,
+      "step": 65400
+    },
+    {
+      "epoch": 130.86,
+      "grad_norm": 22.714303970336914,
+      "learning_rate": 3.963853234740006e-07,
+      "loss": 0.2628,
+      "step": 65430
+    },
+    {
+      "epoch": 130.92,
+      "grad_norm": 83.66651916503906,
+      "learning_rate": 3.9393715351330243e-07,
+      "loss": 0.2366,
+      "step": 65460
+    },
+    {
+      "epoch": 130.98,
+      "grad_norm": 34.74546813964844,
+      "learning_rate": 3.9149625715455107e-07,
+      "loss": 0.1932,
+      "step": 65490
+    },
+    {
+      "epoch": 131.0,
+      "eval_loss": 0.30847281217575073,
+      "eval_map": 0.8098,
+      "eval_map_50": 0.9506,
+      "eval_map_75": 0.9102,
+      "eval_map_chicken": 0.8093,
+      "eval_map_duck": 0.7464,
+      "eval_map_large": 0.8153,
+      "eval_map_medium": 0.8121,
+      "eval_map_plant": 0.8738,
+      "eval_map_small": 0.373,
+      "eval_mar_1": 0.3205,
+      "eval_mar_10": 0.8444,
+      "eval_mar_100": 0.8473,
+      "eval_mar_100_chicken": 0.8476,
+      "eval_mar_100_duck": 0.7948,
+      "eval_mar_100_plant": 0.8994,
+      "eval_mar_large": 0.8406,
+      "eval_mar_medium": 0.8522,
+      "eval_mar_small": 0.4119,
+      "eval_runtime": 11.4515,
+      "eval_samples_per_second": 8.733,
+      "eval_steps_per_second": 1.135,
+      "step": 65500
+    },
+    {
+      "epoch": 131.04,
+      "grad_norm": 79.6670913696289,
+      "learning_rate": 3.890626382522539e-07,
+      "loss": 0.1955,
+      "step": 65520
+    },
+    {
+      "epoch": 131.1,
+      "grad_norm": 31.21834945678711,
+      "learning_rate": 3.866363006494256e-07,
+      "loss": 0.2079,
+      "step": 65550
+    },
+    {
+      "epoch": 131.16,
+      "grad_norm": 108.16787719726562,
+      "learning_rate": 3.8421724817758745e-07,
+      "loss": 0.1955,
+      "step": 65580
+    },
+    {
+      "epoch": 131.22,
+      "grad_norm": 51.9688606262207,
+      "learning_rate": 3.818054846567515e-07,
+      "loss": 0.2034,
+      "step": 65610
+    },
+    {
+      "epoch": 131.28,
+      "grad_norm": 32.75484848022461,
+      "learning_rate": 3.794010138954213e-07,
+      "loss": 0.231,
+      "step": 65640
+    },
+    {
+      "epoch": 131.34,
+      "grad_norm": 41.51599884033203,
+      "learning_rate": 3.770038396905862e-07,
+      "loss": 0.2235,
+      "step": 65670
+    },
+    {
+      "epoch": 131.4,
+      "grad_norm": 59.717350006103516,
+      "learning_rate": 3.7461396582771035e-07,
+      "loss": 0.2181,
+      "step": 65700
+    },
+    {
+      "epoch": 131.46,
+      "grad_norm": 99.7775650024414,
+      "learning_rate": 3.7223139608073e-07,
+      "loss": 0.2233,
+      "step": 65730
+    },
+    {
+      "epoch": 131.52,
+      "grad_norm": 50.78632354736328,
+      "learning_rate": 3.698561342120499e-07,
+      "loss": 0.2195,
+      "step": 65760
+    },
+    {
+      "epoch": 131.58,
+      "grad_norm": 73.87471771240234,
+      "learning_rate": 3.674881839725314e-07,
+      "loss": 0.2186,
+      "step": 65790
+    },
+    {
+      "epoch": 131.64,
+      "grad_norm": 63.35624313354492,
+      "learning_rate": 3.651275491014905e-07,
+      "loss": 0.2371,
+      "step": 65820
+    },
+    {
+      "epoch": 131.7,
+      "grad_norm": 111.2394027709961,
+      "learning_rate": 3.627742333266937e-07,
+      "loss": 0.2038,
+      "step": 65850
+    },
+    {
+      "epoch": 131.76,
+      "grad_norm": 46.248443603515625,
+      "learning_rate": 3.604282403643472e-07,
+      "loss": 0.1969,
+      "step": 65880
+    },
+    {
+      "epoch": 131.82,
+      "grad_norm": 109.54549407958984,
+      "learning_rate": 3.5808957391909315e-07,
+      "loss": 0.2108,
+      "step": 65910
+    },
+    {
+      "epoch": 131.88,
+      "grad_norm": 77.7681655883789,
+      "learning_rate": 3.557582376840063e-07,
+      "loss": 0.2123,
+      "step": 65940
+    },
+    {
+      "epoch": 131.94,
+      "grad_norm": 37.92878723144531,
+      "learning_rate": 3.534342353405834e-07,
+      "loss": 0.2221,
+      "step": 65970
+    },
+    {
+      "epoch": 132.0,
+      "grad_norm": 75.26317596435547,
+      "learning_rate": 3.511175705587433e-07,
+      "loss": 0.1961,
+      "step": 66000
+    },
+    {
+      "epoch": 132.0,
+      "eval_loss": 0.30847275257110596,
+      "eval_map": 0.8095,
+      "eval_map_50": 0.9505,
+      "eval_map_75": 0.9027,
+      "eval_map_chicken": 0.8096,
+      "eval_map_duck": 0.7432,
+      "eval_map_large": 0.8094,
+      "eval_map_medium": 0.8132,
+      "eval_map_plant": 0.8755,
+      "eval_map_small": 0.3761,
+      "eval_mar_1": 0.3203,
+      "eval_mar_10": 0.8439,
+      "eval_mar_100": 0.8473,
+      "eval_mar_100_chicken": 0.8484,
+      "eval_mar_100_duck": 0.7928,
+      "eval_mar_100_plant": 0.9006,
+      "eval_mar_large": 0.8352,
+      "eval_mar_medium": 0.8526,
+      "eval_mar_small": 0.4219,
+      "eval_runtime": 12.145,
+      "eval_samples_per_second": 8.234,
+      "eval_steps_per_second": 1.07,
+      "step": 66000
+    },
+    {
+      "epoch": 132.06,
+      "grad_norm": 29.321550369262695,
+      "learning_rate": 3.488082469968146e-07,
+      "loss": 0.2137,
+      "step": 66030
+    },
+    {
+      "epoch": 132.12,
+      "grad_norm": 54.83131408691406,
+      "learning_rate": 3.465062683015341e-07,
+      "loss": 0.2166,
+      "step": 66060
+    },
+    {
+      "epoch": 132.18,
+      "grad_norm": 50.95697784423828,
+      "learning_rate": 3.442116381080418e-07,
+      "loss": 0.2172,
+      "step": 66090
+    },
+    {
+      "epoch": 132.24,
+      "grad_norm": 44.23823928833008,
+      "learning_rate": 3.419243600398703e-07,
+      "loss": 0.2118,
+      "step": 66120
+    },
+    {
+      "epoch": 132.3,
+      "grad_norm": 40.69063186645508,
+      "learning_rate": 3.396444377089453e-07,
+      "loss": 0.2206,
+      "step": 66150
+    },
+    {
+      "epoch": 132.36,
+      "grad_norm": 49.35795974731445,
+      "learning_rate": 3.373718747155752e-07,
+      "loss": 0.2042,
+      "step": 66180
+    },
+    {
+      "epoch": 132.42,
+      "grad_norm": 21.74803924560547,
+      "learning_rate": 3.351066746484455e-07,
+      "loss": 0.2228,
+      "step": 66210
+    },
+    {
+      "epoch": 132.48,
+      "grad_norm": 58.39532470703125,
+      "learning_rate": 3.328488410846187e-07,
+      "loss": 0.2136,
+      "step": 66240
+    },
+    {
+      "epoch": 132.54,
+      "grad_norm": 41.151676177978516,
+      "learning_rate": 3.3059837758951995e-07,
+      "loss": 0.1966,
+      "step": 66270
+    },
+    {
+      "epoch": 132.6,
+      "grad_norm": 40.82617950439453,
+      "learning_rate": 3.283552877169399e-07,
+      "loss": 0.1999,
+      "step": 66300
+    },
+    {
+      "epoch": 132.66,
+      "grad_norm": 35.09123229980469,
+      "learning_rate": 3.2611957500902345e-07,
+      "loss": 0.2105,
+      "step": 66330
+    },
+    {
+      "epoch": 132.72,
+      "grad_norm": 45.79312515258789,
+      "learning_rate": 3.2389124299626483e-07,
+      "loss": 0.2106,
+      "step": 66360
+    },
+    {
+      "epoch": 132.78,
+      "grad_norm": 71.86212921142578,
+      "learning_rate": 3.216702951975059e-07,
+      "loss": 0.2331,
+      "step": 66390
+    },
+    {
+      "epoch": 132.84,
+      "grad_norm": 65.11103057861328,
+      "learning_rate": 3.194567351199257e-07,
+      "loss": 0.2018,
+      "step": 66420
+    },
+    {
+      "epoch": 132.9,
+      "grad_norm": 55.746700286865234,
+      "learning_rate": 3.172505662590386e-07,
+      "loss": 0.2141,
+      "step": 66450
+    },
+    {
+      "epoch": 132.96,
+      "grad_norm": 50.38323211669922,
+      "learning_rate": 3.150517920986851e-07,
+      "loss": 0.2426,
+      "step": 66480
+    },
+    {
+      "epoch": 133.0,
+      "eval_loss": 0.3091796040534973,
+      "eval_map": 0.8092,
+      "eval_map_50": 0.9509,
+      "eval_map_75": 0.9023,
+      "eval_map_chicken": 0.811,
+      "eval_map_duck": 0.7433,
+      "eval_map_large": 0.8172,
+      "eval_map_medium": 0.8094,
+      "eval_map_plant": 0.8734,
+      "eval_map_small": 0.3815,
+      "eval_mar_1": 0.3209,
+      "eval_mar_10": 0.8446,
+      "eval_mar_100": 0.8473,
+      "eval_mar_100_chicken": 0.8504,
+      "eval_mar_100_duck": 0.7928,
+      "eval_mar_100_plant": 0.8988,
+      "eval_mar_large": 0.8398,
+      "eval_mar_medium": 0.8504,
+      "eval_mar_small": 0.4271,
+      "eval_runtime": 14.3657,
+      "eval_samples_per_second": 6.961,
+      "eval_steps_per_second": 0.905,
+      "step": 66500
+    },
+    {
+      "epoch": 133.02,
+      "grad_norm": 41.360443115234375,
+      "learning_rate": 3.128604161110299e-07,
+      "loss": 0.2051,
+      "step": 66510
+    },
+    {
+      "epoch": 133.08,
+      "grad_norm": 207.290283203125,
+      "learning_rate": 3.106764417565561e-07,
+      "loss": 0.1918,
+      "step": 66540
+    },
+    {
+      "epoch": 133.14,
+      "grad_norm": 49.47697448730469,
+      "learning_rate": 3.08499872484056e-07,
+      "loss": 0.1988,
+      "step": 66570
+    },
+    {
+      "epoch": 133.2,
+      "grad_norm": 101.23628234863281,
+      "learning_rate": 3.0633071173062966e-07,
+      "loss": 0.2284,
+      "step": 66600
+    },
+    {
+      "epoch": 133.26,
+      "grad_norm": 62.6041374206543,
+      "learning_rate": 3.0416896292167873e-07,
+      "loss": 0.2599,
+      "step": 66630
+    },
+    {
+      "epoch": 133.32,
+      "grad_norm": 59.834163665771484,
+      "learning_rate": 3.0201462947089865e-07,
+      "loss": 0.2035,
+      "step": 66660
+    },
+    {
+      "epoch": 133.38,
+      "grad_norm": 60.794857025146484,
+      "learning_rate": 2.99867714780277e-07,
+      "loss": 0.2439,
+      "step": 66690
+    },
+    {
+      "epoch": 133.44,
+      "grad_norm": 48.02446746826172,
+      "learning_rate": 2.9772822224008515e-07,
+      "loss": 0.1922,
+      "step": 66720
+    },
+    {
+      "epoch": 133.5,
+      "grad_norm": 29.021291732788086,
+      "learning_rate": 2.9559615522887275e-07,
+      "loss": 0.1886,
+      "step": 66750
+    },
+    {
+      "epoch": 133.56,
+      "grad_norm": 84.00584411621094,
+      "learning_rate": 2.9347151711346556e-07,
+      "loss": 0.2242,
+      "step": 66780
+    },
+    {
+      "epoch": 133.62,
+      "grad_norm": 35.25551223754883,
+      "learning_rate": 2.913543112489564e-07,
+      "loss": 0.2064,
+      "step": 66810
+    },
+    {
+      "epoch": 133.68,
+      "grad_norm": 47.507972717285156,
+      "learning_rate": 2.892445409787037e-07,
+      "loss": 0.1964,
+      "step": 66840
+    },
+    {
+      "epoch": 133.74,
+      "grad_norm": 44.02022171020508,
+      "learning_rate": 2.8714220963432125e-07,
+      "loss": 0.2238,
+      "step": 66870
+    },
+    {
+      "epoch": 133.8,
+      "grad_norm": 34.1710319519043,
+      "learning_rate": 2.850473205356774e-07,
+      "loss": 0.1996,
+      "step": 66900
+    },
+    {
+      "epoch": 133.86,
+      "grad_norm": 117.07316589355469,
+      "learning_rate": 2.829598769908892e-07,
+      "loss": 0.1996,
+      "step": 66930
+    },
+    {
+      "epoch": 133.92,
+      "grad_norm": 47.652687072753906,
+      "learning_rate": 2.8087988229631325e-07,
+      "loss": 0.2035,
+      "step": 66960
+    },
+    {
+      "epoch": 133.98,
+      "grad_norm": 82.08881378173828,
+      "learning_rate": 2.788073397365465e-07,
+      "loss": 0.2011,
+      "step": 66990
+    },
+    {
+      "epoch": 134.0,
+      "eval_loss": 0.3087918162345886,
+      "eval_map": 0.8087,
+      "eval_map_50": 0.9501,
+      "eval_map_75": 0.9024,
+      "eval_map_chicken": 0.8105,
+      "eval_map_duck": 0.7434,
+      "eval_map_large": 0.8148,
+      "eval_map_medium": 0.8093,
+      "eval_map_plant": 0.8724,
+      "eval_map_small": 0.3701,
+      "eval_mar_1": 0.3204,
+      "eval_mar_10": 0.8441,
+      "eval_mar_100": 0.847,
+      "eval_mar_100_chicken": 0.8496,
+      "eval_mar_100_duck": 0.7928,
+      "eval_mar_100_plant": 0.8985,
+      "eval_mar_large": 0.8407,
+      "eval_mar_medium": 0.8503,
+      "eval_mar_small": 0.4205,
+      "eval_runtime": 12.2877,
+      "eval_samples_per_second": 8.138,
+      "eval_steps_per_second": 1.058,
+      "step": 67000
+    },
+    {
+      "epoch": 134.04,
+      "grad_norm": 50.52214050292969,
+      "learning_rate": 2.76742252584416e-07,
+      "loss": 0.2204,
+      "step": 67020
+    },
+    {
+      "epoch": 134.1,
+      "grad_norm": 78.1152114868164,
+      "learning_rate": 2.746846241009765e-07,
+      "loss": 0.216,
+      "step": 67050
+    },
+    {
+      "epoch": 134.16,
+      "grad_norm": 77.4903793334961,
+      "learning_rate": 2.7263445753550275e-07,
+      "loss": 0.2467,
+      "step": 67080
+    },
+    {
+      "epoch": 134.22,
+      "grad_norm": 73.16485595703125,
+      "learning_rate": 2.7059175612548947e-07,
+      "loss": 0.216,
+      "step": 67110
+    },
+    {
+      "epoch": 134.28,
+      "grad_norm": 68.31117248535156,
+      "learning_rate": 2.685565230966408e-07,
+      "loss": 0.2242,
+      "step": 67140
+    },
+    {
+      "epoch": 134.34,
+      "grad_norm": 36.76439666748047,
+      "learning_rate": 2.665287616628659e-07,
+      "loss": 0.198,
+      "step": 67170
+    },
+    {
+      "epoch": 134.4,
+      "grad_norm": 59.230377197265625,
+      "learning_rate": 2.6450847502627883e-07,
+      "loss": 0.1889,
+      "step": 67200
+    },
+    {
+      "epoch": 134.46,
+      "grad_norm": 44.514556884765625,
+      "learning_rate": 2.6249566637718714e-07,
+      "loss": 0.2342,
+      "step": 67230
+    },
+    {
+      "epoch": 134.52,
+      "grad_norm": 50.98478317260742,
+      "learning_rate": 2.604903388940899e-07,
+      "loss": 0.2152,
+      "step": 67260
+    },
+    {
+      "epoch": 134.58,
+      "grad_norm": 51.002811431884766,
+      "learning_rate": 2.584924957436735e-07,
+      "loss": 0.2133,
+      "step": 67290
+    },
+    {
+      "epoch": 134.64,
+      "grad_norm": 54.643314361572266,
+      "learning_rate": 2.5650214008080544e-07,
+      "loss": 0.2243,
+      "step": 67320
+    },
+    {
+      "epoch": 134.7,
+      "grad_norm": 87.97196960449219,
+      "learning_rate": 2.5451927504852757e-07,
+      "loss": 0.2116,
+      "step": 67350
+    },
+    {
+      "epoch": 134.76,
+      "grad_norm": 36.109962463378906,
+      "learning_rate": 2.525439037780558e-07,
+      "loss": 0.1959,
+      "step": 67380
+    },
+    {
+      "epoch": 134.82,
+      "grad_norm": 55.99542236328125,
+      "learning_rate": 2.505760293887699e-07,
+      "loss": 0.2036,
+      "step": 67410
+    },
+    {
+      "epoch": 134.88,
+      "grad_norm": 61.448246002197266,
+      "learning_rate": 2.486156549882135e-07,
+      "loss": 0.2016,
+      "step": 67440
+    },
+    {
+      "epoch": 134.94,
+      "grad_norm": 64.0799789428711,
+      "learning_rate": 2.4666278367208417e-07,
+      "loss": 0.2162,
+      "step": 67470
+    },
+    {
+      "epoch": 135.0,
+      "grad_norm": 79.16766357421875,
+      "learning_rate": 2.447174185242324e-07,
+      "loss": 0.2408,
+      "step": 67500
+    },
+    {
+      "epoch": 135.0,
+      "eval_loss": 0.3119555413722992,
+      "eval_map": 0.8034,
+      "eval_map_50": 0.9497,
+      "eval_map_75": 0.9032,
+      "eval_map_chicken": 0.8028,
+      "eval_map_duck": 0.7355,
+      "eval_map_large": 0.8049,
+      "eval_map_medium": 0.8042,
+      "eval_map_plant": 0.8718,
+      "eval_map_small": 0.3635,
+      "eval_mar_1": 0.3183,
+      "eval_mar_10": 0.8406,
+      "eval_mar_100": 0.8434,
+      "eval_mar_100_chicken": 0.8437,
+      "eval_mar_100_duck": 0.7876,
+      "eval_mar_100_plant": 0.8988,
+      "eval_mar_large": 0.832,
+      "eval_mar_medium": 0.848,
+      "eval_mar_small": 0.409,
+      "eval_runtime": 13.7574,
+      "eval_samples_per_second": 7.269,
+      "eval_steps_per_second": 0.945,
+      "step": 67500
+    },
+    {
+      "epoch": 135.06,
+      "grad_norm": 65.69921875,
+      "learning_rate": 2.4277956261665624e-07,
+      "loss": 0.2107,
+      "step": 67530
+    },
+    {
+      "epoch": 135.12,
+      "grad_norm": 64.71975708007812,
+      "learning_rate": 2.40849219009493e-07,
+      "loss": 0.228,
+      "step": 67560
+    },
+    {
+      "epoch": 135.18,
+      "grad_norm": 59.862911224365234,
+      "learning_rate": 2.389263907510209e-07,
+      "loss": 0.2089,
+      "step": 67590
+    },
+    {
+      "epoch": 135.24,
+      "grad_norm": 32.341583251953125,
+      "learning_rate": 2.3701108087764657e-07,
+      "loss": 0.1987,
+      "step": 67620
+    },
+    {
+      "epoch": 135.3,
+      "grad_norm": 109.37493133544922,
+      "learning_rate": 2.351032924139063e-07,
+      "loss": 0.2295,
+      "step": 67650
+    },
+    {
+      "epoch": 135.36,
+      "grad_norm": 45.07475662231445,
+      "learning_rate": 2.3320302837245846e-07,
+      "loss": 0.1956,
+      "step": 67680
+    },
+    {
+      "epoch": 135.42,
+      "grad_norm": 54.63572692871094,
+      "learning_rate": 2.3131029175407883e-07,
+      "loss": 0.1937,
+      "step": 67710
+    },
+    {
+      "epoch": 135.48,
+      "grad_norm": 79.57504272460938,
+      "learning_rate": 2.2942508554765764e-07,
+      "loss": 0.1992,
+      "step": 67740
+    },
+    {
+      "epoch": 135.54,
+      "grad_norm": 74.65667724609375,
+      "learning_rate": 2.27547412730193e-07,
+      "loss": 0.2102,
+      "step": 67770
+    },
+    {
+      "epoch": 135.6,
+      "grad_norm": 22.859420776367188,
+      "learning_rate": 2.2567727626678527e-07,
+      "loss": 0.2013,
+      "step": 67800
+    },
+    {
+      "epoch": 135.66,
+      "grad_norm": 52.90283966064453,
+      "learning_rate": 2.2381467911063658e-07,
+      "loss": 0.2175,
+      "step": 67830
+    },
+    {
+      "epoch": 135.72,
+      "grad_norm": 93.8233871459961,
+      "learning_rate": 2.2195962420304083e-07,
+      "loss": 0.2423,
+      "step": 67860
+    },
+    {
+      "epoch": 135.78,
+      "grad_norm": 54.931026458740234,
+      "learning_rate": 2.2011211447338477e-07,
+      "loss": 0.1994,
+      "step": 67890
+    },
+    {
+      "epoch": 135.84,
+      "grad_norm": 80.65608978271484,
+      "learning_rate": 2.1827215283913683e-07,
+      "loss": 0.2231,
+      "step": 67920
+    },
+    {
+      "epoch": 135.9,
+      "grad_norm": 249.62669372558594,
+      "learning_rate": 2.1643974220584729e-07,
+      "loss": 0.1899,
+      "step": 67950
+    },
+    {
+      "epoch": 135.96,
+      "grad_norm": 103.24887084960938,
+      "learning_rate": 2.1461488546714425e-07,
+      "loss": 0.2169,
+      "step": 67980
+    },
+    {
+      "epoch": 136.0,
+      "eval_loss": 0.3096560537815094,
+      "eval_map": 0.8066,
+      "eval_map_50": 0.9498,
+      "eval_map_75": 0.9037,
+      "eval_map_chicken": 0.8083,
+      "eval_map_duck": 0.7384,
+      "eval_map_large": 0.8147,
+      "eval_map_medium": 0.8068,
+      "eval_map_plant": 0.8732,
+      "eval_map_small": 0.3689,
+      "eval_mar_1": 0.3194,
+      "eval_mar_10": 0.8436,
+      "eval_mar_100": 0.8464,
+      "eval_mar_100_chicken": 0.8476,
+      "eval_mar_100_duck": 0.7918,
+      "eval_mar_100_plant": 0.8997,
+      "eval_mar_large": 0.8411,
+      "eval_mar_medium": 0.8503,
+      "eval_mar_small": 0.4219,
+      "eval_runtime": 11.2635,
+      "eval_samples_per_second": 8.878,
+      "eval_steps_per_second": 1.154,
+      "step": 68000
+    },
+    {
+      "epoch": 136.02,
+      "grad_norm": 66.1946792602539,
+      "learning_rate": 2.127975855047243e-07,
+      "loss": 0.2674,
+      "step": 68010
+    },
+    {
+      "epoch": 136.08,
+      "grad_norm": 45.892452239990234,
+      "learning_rate": 2.1098784518835292e-07,
+      "loss": 0.2066,
+      "step": 68040
+    },
+    {
+      "epoch": 136.14,
+      "grad_norm": 48.29930114746094,
+      "learning_rate": 2.0918566737585688e-07,
+      "loss": 0.2193,
+      "step": 68070
+    },
+    {
+      "epoch": 136.2,
+      "grad_norm": 48.946964263916016,
+      "learning_rate": 2.0739105491312028e-07,
+      "loss": 0.2229,
+      "step": 68100
+    },
+    {
+      "epoch": 136.26,
+      "grad_norm": 52.66734313964844,
+      "learning_rate": 2.056040106340823e-07,
+      "loss": 0.2114,
+      "step": 68130
+    },
+    {
+      "epoch": 136.32,
+      "grad_norm": 61.51396942138672,
+      "learning_rate": 2.0382453736072838e-07,
+      "loss": 0.2044,
+      "step": 68160
+    },
+    {
+      "epoch": 136.38,
+      "grad_norm": 64.57730102539062,
+      "learning_rate": 2.0205263790309125e-07,
+      "loss": 0.195,
+      "step": 68190
+    },
+    {
+      "epoch": 136.44,
+      "grad_norm": 45.15483856201172,
+      "learning_rate": 2.0028831505924162e-07,
+      "loss": 0.2264,
+      "step": 68220
+    },
+    {
+      "epoch": 136.5,
+      "grad_norm": 42.67230987548828,
+      "learning_rate": 1.9853157161528468e-07,
+      "loss": 0.2093,
+      "step": 68250
+    },
+    {
+      "epoch": 136.56,
+      "grad_norm": 17.29395294189453,
+      "learning_rate": 1.967824103453597e-07,
+      "loss": 0.2212,
+      "step": 68280
+    },
+    {
+      "epoch": 136.62,
+      "grad_norm": 55.87034225463867,
+      "learning_rate": 1.9504083401162999e-07,
+      "loss": 0.2144,
+      "step": 68310
+    },
+    {
+      "epoch": 136.68,
+      "grad_norm": 57.51459884643555,
+      "learning_rate": 1.9330684536428335e-07,
+      "loss": 0.1862,
+      "step": 68340
+    },
+    {
+      "epoch": 136.74,
+      "grad_norm": 70.5501937866211,
+      "learning_rate": 1.9158044714152447e-07,
+      "loss": 0.2088,
+      "step": 68370
+    },
+    {
+      "epoch": 136.8,
+      "grad_norm": 26.095773696899414,
+      "learning_rate": 1.8986164206957037e-07,
+      "loss": 0.2091,
+      "step": 68400
+    },
+    {
+      "epoch": 136.86,
+      "grad_norm": 28.720632553100586,
+      "learning_rate": 1.8815043286265044e-07,
+      "loss": 0.194,
+      "step": 68430
+    },
+    {
+      "epoch": 136.92,
+      "grad_norm": 46.130802154541016,
+      "learning_rate": 1.8644682222299703e-07,
+      "loss": 0.191,
+      "step": 68460
+    },
+    {
+      "epoch": 136.98,
+      "grad_norm": 32.829002380371094,
+      "learning_rate": 1.8475081284084428e-07,
+      "loss": 0.2292,
+      "step": 68490
+    },
+    {
+      "epoch": 137.0,
+      "eval_loss": 0.30863094329833984,
+      "eval_map": 0.8046,
+      "eval_map_50": 0.9498,
+      "eval_map_75": 0.8996,
+      "eval_map_chicken": 0.8058,
+      "eval_map_duck": 0.7339,
+      "eval_map_large": 0.8031,
+      "eval_map_medium": 0.8068,
+      "eval_map_plant": 0.874,
+      "eval_map_small": 0.3796,
+      "eval_mar_1": 0.3184,
+      "eval_mar_10": 0.8407,
+      "eval_mar_100": 0.844,
+      "eval_mar_100_chicken": 0.8456,
+      "eval_mar_100_duck": 0.7866,
+      "eval_mar_100_plant": 0.8997,
+      "eval_mar_large": 0.8316,
+      "eval_mar_medium": 0.849,
+      "eval_mar_small": 0.4314,
+      "eval_runtime": 12.1831,
+      "eval_samples_per_second": 8.208,
+      "eval_steps_per_second": 1.067,
+      "step": 68500
+    },
+    {
+      "epoch": 137.04,
+      "grad_norm": 31.091623306274414,
+      "learning_rate": 1.8306240739442094e-07,
+      "loss": 0.2179,
+      "step": 68520
+    },
+    {
+      "epoch": 137.1,
+      "grad_norm": 40.29054641723633,
+      "learning_rate": 1.8138160854995145e-07,
+      "loss": 0.2034,
+      "step": 68550
+    },
+    {
+      "epoch": 137.16,
+      "grad_norm": 35.621116638183594,
+      "learning_rate": 1.7970841896164658e-07,
+      "loss": 0.2096,
+      "step": 68580
+    },
+    {
+      "epoch": 137.22,
+      "grad_norm": 106.7773666381836,
+      "learning_rate": 1.7804284127169946e-07,
+      "loss": 0.1982,
+      "step": 68610
+    },
+    {
+      "epoch": 137.28,
+      "grad_norm": 39.857696533203125,
+      "learning_rate": 1.7638487811028616e-07,
+      "loss": 0.1941,
+      "step": 68640
+    },
+    {
+      "epoch": 137.34,
+      "grad_norm": 57.87089157104492,
+      "learning_rate": 1.7473453209555625e-07,
+      "loss": 0.2281,
+      "step": 68670
+    },
+    {
+      "epoch": 137.4,
+      "grad_norm": 76.89950561523438,
+      "learning_rate": 1.7309180583363062e-07,
+      "loss": 0.2403,
+      "step": 68700
+    },
+    {
+      "epoch": 137.46,
+      "grad_norm": 58.04665756225586,
+      "learning_rate": 1.7145670191859977e-07,
+      "loss": 0.2296,
+      "step": 68730
+    },
+    {
+      "epoch": 137.52,
+      "grad_norm": 28.78618049621582,
+      "learning_rate": 1.6982922293251548e-07,
+      "loss": 0.2145,
+      "step": 68760
+    },
+    {
+      "epoch": 137.58,
+      "grad_norm": 45.55547332763672,
+      "learning_rate": 1.6820937144538807e-07,
+      "loss": 0.1947,
+      "step": 68790
+    },
+    {
+      "epoch": 137.64,
+      "grad_norm": 42.586551666259766,
+      "learning_rate": 1.6659715001518583e-07,
+      "loss": 0.2322,
+      "step": 68820
+    },
+    {
+      "epoch": 137.7,
+      "grad_norm": 25.68801498413086,
+      "learning_rate": 1.6499256118782503e-07,
+      "loss": 0.1923,
+      "step": 68850
+    },
+    {
+      "epoch": 137.76,
+      "grad_norm": 54.242618560791016,
+      "learning_rate": 1.6339560749717154e-07,
+      "loss": 0.2217,
+      "step": 68880
+    },
+    {
+      "epoch": 137.82,
+      "grad_norm": 59.135337829589844,
+      "learning_rate": 1.6180629146503256e-07,
+      "loss": 0.2069,
+      "step": 68910
+    },
+    {
+      "epoch": 137.88,
+      "grad_norm": 37.869014739990234,
+      "learning_rate": 1.6022461560115498e-07,
+      "loss": 0.207,
+      "step": 68940
+    },
+    {
+      "epoch": 137.94,
+      "grad_norm": 48.881080627441406,
+      "learning_rate": 1.586505824032214e-07,
+      "loss": 0.1919,
+      "step": 68970
+    },
+    {
+      "epoch": 138.0,
+      "grad_norm": 206.4425048828125,
+      "learning_rate": 1.5708419435684463e-07,
+      "loss": 0.2231,
+      "step": 69000
+    },
+    {
+      "epoch": 138.0,
+      "eval_loss": 0.30831366777420044,
+      "eval_map": 0.8073,
+      "eval_map_50": 0.9477,
+      "eval_map_75": 0.905,
+      "eval_map_chicken": 0.8075,
+      "eval_map_duck": 0.74,
+      "eval_map_large": 0.8085,
+      "eval_map_medium": 0.8098,
+      "eval_map_plant": 0.8742,
+      "eval_map_small": 0.3733,
+      "eval_mar_1": 0.3192,
+      "eval_mar_10": 0.8433,
+      "eval_mar_100": 0.8461,
+      "eval_mar_100_chicken": 0.848,
+      "eval_mar_100_duck": 0.7897,
+      "eval_mar_100_plant": 0.9006,
+      "eval_mar_large": 0.8356,
+      "eval_mar_medium": 0.8516,
+      "eval_mar_small": 0.4119,
+      "eval_runtime": 11.0571,
+      "eval_samples_per_second": 9.044,
+      "eval_steps_per_second": 1.176,
+      "step": 69000
+    },
+    {
+      "epoch": 138.06,
+      "grad_norm": 47.209869384765625,
+      "learning_rate": 1.555254539355655e-07,
+      "loss": 0.2146,
+      "step": 69030
+    },
+    {
+      "epoch": 138.12,
+      "grad_norm": 53.7562255859375,
+      "learning_rate": 1.5397436360084784e-07,
+      "loss": 0.2006,
+      "step": 69060
+    },
+    {
+      "epoch": 138.18,
+      "grad_norm": 82.0704345703125,
+      "learning_rate": 1.5243092580207507e-07,
+      "loss": 0.2285,
+      "step": 69090
+    },
+    {
+      "epoch": 138.24,
+      "grad_norm": 31.887880325317383,
+      "learning_rate": 1.5089514297654594e-07,
+      "loss": 0.2017,
+      "step": 69120
+    },
+    {
+      "epoch": 138.3,
+      "grad_norm": 94.9820327758789,
+      "learning_rate": 1.4936701754947104e-07,
+      "loss": 0.2195,
+      "step": 69150
+    },
+    {
+      "epoch": 138.36,
+      "grad_norm": 75.08333587646484,
+      "learning_rate": 1.4784655193396947e-07,
+      "loss": 0.2127,
+      "step": 69180
+    },
+    {
+      "epoch": 138.42,
+      "grad_norm": 40.21970748901367,
+      "learning_rate": 1.463337485310634e-07,
+      "loss": 0.1941,
+      "step": 69210
+    },
+    {
+      "epoch": 138.48,
+      "grad_norm": 50.572811126708984,
+      "learning_rate": 1.448286097296764e-07,
+      "loss": 0.2082,
+      "step": 69240
+    },
+    {
+      "epoch": 138.54,
+      "grad_norm": 503.557861328125,
+      "learning_rate": 1.4333113790662822e-07,
+      "loss": 0.2164,
+      "step": 69270
+    },
+    {
+      "epoch": 138.6,
+      "grad_norm": 46.357208251953125,
+      "learning_rate": 1.4184133542663014e-07,
+      "loss": 0.1926,
+      "step": 69300
+    },
+    {
+      "epoch": 138.66,
+      "grad_norm": 77.93207550048828,
+      "learning_rate": 1.4035920464228525e-07,
+      "loss": 0.2165,
+      "step": 69330
+    },
+    {
+      "epoch": 138.72,
+      "grad_norm": 32.061344146728516,
+      "learning_rate": 1.388847478940797e-07,
+      "loss": 0.1961,
+      "step": 69360
+    },
+    {
+      "epoch": 138.78,
+      "grad_norm": 45.14935302734375,
+      "learning_rate": 1.3741796751038095e-07,
+      "loss": 0.2195,
+      "step": 69390
+    },
+    {
+      "epoch": 138.84,
+      "grad_norm": 18.540773391723633,
+      "learning_rate": 1.3595886580743677e-07,
+      "loss": 0.2463,
+      "step": 69420
+    },
+    {
+      "epoch": 138.9,
+      "grad_norm": 61.1831169128418,
+      "learning_rate": 1.3450744508936687e-07,
+      "loss": 0.2171,
+      "step": 69450
+    },
+    {
+      "epoch": 138.96,
+      "grad_norm": 48.19976806640625,
+      "learning_rate": 1.330637076481639e-07,
+      "loss": 0.1987,
+      "step": 69480
+    },
+    {
+      "epoch": 139.0,
+      "eval_loss": 0.3095991015434265,
+      "eval_map": 0.8068,
+      "eval_map_50": 0.9492,
+      "eval_map_75": 0.8997,
+      "eval_map_chicken": 0.8089,
+      "eval_map_duck": 0.7366,
+      "eval_map_large": 0.8041,
+      "eval_map_medium": 0.8094,
+      "eval_map_plant": 0.8747,
+      "eval_map_small": 0.3756,
+      "eval_mar_1": 0.319,
+      "eval_mar_10": 0.8427,
+      "eval_mar_100": 0.8454,
+      "eval_mar_100_chicken": 0.8488,
+      "eval_mar_100_duck": 0.7876,
+      "eval_mar_100_plant": 0.8997,
+      "eval_mar_large": 0.8318,
+      "eval_mar_medium": 0.851,
+      "eval_mar_small": 0.4252,
+      "eval_runtime": 12.2386,
+      "eval_samples_per_second": 8.171,
+      "eval_steps_per_second": 1.062,
+      "step": 69500
+    },
+    {
+      "epoch": 139.02,
+      "grad_norm": 33.00763702392578,
+      "learning_rate": 1.3162765576368587e-07,
+      "loss": 0.2106,
+      "step": 69510
+    },
+    {
+      "epoch": 139.08,
+      "grad_norm": 71.31500244140625,
+      "learning_rate": 1.3019929170365376e-07,
+      "loss": 0.209,
+      "step": 69540
+    },
+    {
+      "epoch": 139.14,
+      "grad_norm": 61.765052795410156,
+      "learning_rate": 1.287786177236511e-07,
+      "loss": 0.2267,
+      "step": 69570
+    },
+    {
+      "epoch": 139.2,
+      "grad_norm": 25.743423461914062,
+      "learning_rate": 1.2736563606711384e-07,
+      "loss": 0.2014,
+      "step": 69600
+    },
+    {
+      "epoch": 139.26,
+      "grad_norm": 48.05965805053711,
+      "learning_rate": 1.259603489653355e-07,
+      "loss": 0.255,
+      "step": 69630
+    },
+    {
+      "epoch": 139.32,
+      "grad_norm": 52.13318634033203,
+      "learning_rate": 1.2456275863745426e-07,
+      "loss": 0.1985,
+      "step": 69660
+    },
+    {
+      "epoch": 139.38,
+      "grad_norm": 67.66744995117188,
+      "learning_rate": 1.2317286729045586e-07,
+      "loss": 0.2397,
+      "step": 69690
+    },
+    {
+      "epoch": 139.44,
+      "grad_norm": 135.82752990722656,
+      "learning_rate": 1.2179067711917015e-07,
+      "loss": 0.1931,
+      "step": 69720
+    },
+    {
+      "epoch": 139.5,
+      "grad_norm": 28.21087074279785,
+      "learning_rate": 1.2041619030626283e-07,
+      "loss": 0.1983,
+      "step": 69750
+    },
+    {
+      "epoch": 139.56,
+      "grad_norm": 49.653682708740234,
+      "learning_rate": 1.1904940902223661e-07,
+      "loss": 0.2175,
+      "step": 69780
+    },
+    {
+      "epoch": 139.62,
+      "grad_norm": 158.8368682861328,
+      "learning_rate": 1.1769033542542552e-07,
+      "loss": 0.2223,
+      "step": 69810
+    },
+    {
+      "epoch": 139.68,
+      "grad_norm": 31.01651382446289,
+      "learning_rate": 1.1633897166199227e-07,
+      "loss": 0.192,
+      "step": 69840
+    },
+    {
+      "epoch": 139.74,
+      "grad_norm": 54.803897857666016,
+      "learning_rate": 1.1499531986592482e-07,
+      "loss": 0.1917,
+      "step": 69870
+    },
+    {
+      "epoch": 139.8,
+      "grad_norm": 41.02241897583008,
+      "learning_rate": 1.136593821590326e-07,
+      "loss": 0.2329,
+      "step": 69900
+    },
+    {
+      "epoch": 139.86,
+      "grad_norm": 39.60552978515625,
+      "learning_rate": 1.1233116065094363e-07,
+      "loss": 0.2045,
+      "step": 69930
+    },
+    {
+      "epoch": 139.92,
+      "grad_norm": 24.14561653137207,
+      "learning_rate": 1.1101065743910122e-07,
+      "loss": 0.2088,
+      "step": 69960
+    },
+    {
+      "epoch": 139.98,
+      "grad_norm": 36.25834655761719,
+      "learning_rate": 1.0969787460876013e-07,
+      "loss": 0.1917,
+      "step": 69990
+    },
+    {
+      "epoch": 140.0,
+      "eval_loss": 0.3100622296333313,
+      "eval_map": 0.8051,
+      "eval_map_50": 0.9476,
+      "eval_map_75": 0.9028,
+      "eval_map_chicken": 0.8057,
+      "eval_map_duck": 0.735,
+      "eval_map_large": 0.812,
+      "eval_map_medium": 0.8076,
+      "eval_map_plant": 0.8746,
+      "eval_map_small": 0.3669,
+      "eval_mar_1": 0.3184,
+      "eval_mar_10": 0.8411,
+      "eval_mar_100": 0.8441,
+      "eval_mar_100_chicken": 0.8456,
+      "eval_mar_100_duck": 0.7866,
+      "eval_mar_100_plant": 0.9,
+      "eval_mar_large": 0.8377,
+      "eval_mar_medium": 0.8494,
+      "eval_mar_small": 0.4038,
+      "eval_runtime": 13.1165,
+      "eval_samples_per_second": 7.624,
+      "eval_steps_per_second": 0.991,
+      "step": 70000
+    },
+    {
+      "epoch": 140.04,
+      "grad_norm": 77.3727035522461,
+      "learning_rate": 1.0839281423298375e-07,
+      "loss": 0.2235,
+      "step": 70020
+    },
+    {
+      "epoch": 140.1,
+      "grad_norm": 148.96051025390625,
+      "learning_rate": 1.0709547837263967e-07,
+      "loss": 0.1946,
+      "step": 70050
+    },
+    {
+      "epoch": 140.16,
+      "grad_norm": 37.42975997924805,
+      "learning_rate": 1.0580586907639912e-07,
+      "loss": 0.206,
+      "step": 70080
+    },
+    {
+      "epoch": 140.22,
+      "grad_norm": 39.80850601196289,
+      "learning_rate": 1.0452398838073141e-07,
+      "loss": 0.2201,
+      "step": 70110
+    },
+    {
+      "epoch": 140.28,
+      "grad_norm": 44.86832809448242,
+      "learning_rate": 1.032498383099001e-07,
+      "loss": 0.2143,
+      "step": 70140
+    },
+    {
+      "epoch": 140.34,
+      "grad_norm": 88.26593017578125,
+      "learning_rate": 1.0198342087596292e-07,
+      "loss": 0.2273,
+      "step": 70170
+    },
+    {
+      "epoch": 140.4,
+      "grad_norm": 36.841163635253906,
+      "learning_rate": 1.007247380787657e-07,
+      "loss": 0.221,
+      "step": 70200
+    },
+    {
+      "epoch": 140.46,
+      "grad_norm": 76.04727172851562,
+      "learning_rate": 9.947379190594076e-08,
+      "loss": 0.2166,
+      "step": 70230
+    },
+    {
+      "epoch": 140.52,
+      "grad_norm": 111.13554382324219,
+      "learning_rate": 9.823058433290178e-08,
+      "loss": 0.2146,
+      "step": 70260
+    },
+    {
+      "epoch": 140.58,
+      "grad_norm": 75.74388885498047,
+      "learning_rate": 9.699511732284395e-08,
+      "loss": 0.2137,
+      "step": 70290
+    },
+    {
+      "epoch": 140.64,
+      "grad_norm": 57.81622314453125,
+      "learning_rate": 9.576739282673886e-08,
+      "loss": 0.1948,
+      "step": 70320
+    },
+    {
+      "epoch": 140.7,
+      "grad_norm": 125.7251205444336,
+      "learning_rate": 9.454741278333013e-08,
+      "loss": 0.2356,
+      "step": 70350
+    },
+    {
+      "epoch": 140.76,
+      "grad_norm": 178.1024932861328,
+      "learning_rate": 9.333517911913281e-08,
+      "loss": 0.2086,
+      "step": 70380
+    },
+    {
+      "epoch": 140.82,
+      "grad_norm": 51.754905700683594,
+      "learning_rate": 9.213069374842953e-08,
+      "loss": 0.199,
+      "step": 70410
+    },
+    {
+      "epoch": 140.88,
+      "grad_norm": 65.02310180664062,
+      "learning_rate": 9.093395857326714e-08,
+      "loss": 0.213,
+      "step": 70440
+    },
+    {
+      "epoch": 140.94,
+      "grad_norm": 59.75958251953125,
+      "learning_rate": 8.974497548345396e-08,
+      "loss": 0.2072,
+      "step": 70470
+    },
+    {
+      "epoch": 141.0,
+      "grad_norm": 60.88467788696289,
+      "learning_rate": 8.856374635655696e-08,
+      "loss": 0.2281,
+      "step": 70500
+    },
+    {
+      "epoch": 141.0,
+      "eval_loss": 0.3109195828437805,
+      "eval_map": 0.8049,
+      "eval_map_50": 0.9476,
+      "eval_map_75": 0.9091,
+      "eval_map_chicken": 0.8048,
+      "eval_map_duck": 0.7368,
+      "eval_map_large": 0.8106,
+      "eval_map_medium": 0.8076,
+      "eval_map_plant": 0.8731,
+      "eval_map_small": 0.3608,
+      "eval_mar_1": 0.3194,
+      "eval_mar_10": 0.8415,
+      "eval_mar_100": 0.8444,
+      "eval_mar_100_chicken": 0.846,
+      "eval_mar_100_duck": 0.7887,
+      "eval_mar_100_plant": 0.8985,
+      "eval_mar_large": 0.8359,
+      "eval_mar_medium": 0.8499,
+      "eval_mar_small": 0.4038,
+      "eval_runtime": 12.3332,
+      "eval_samples_per_second": 8.108,
+      "eval_steps_per_second": 1.054,
+      "step": 70500
+    },
+    {
+      "epoch": 141.06,
+      "grad_norm": 60.512489318847656,
+      "learning_rate": 8.739027305789682e-08,
+      "loss": 0.2013,
+      "step": 70530
+    },
+    {
+      "epoch": 141.12,
+      "grad_norm": 54.249263763427734,
+      "learning_rate": 8.622455744054958e-08,
+      "loss": 0.2301,
+      "step": 70560
+    },
+    {
+      "epoch": 141.18,
+      "grad_norm": 494.5781555175781,
+      "learning_rate": 8.506660134533828e-08,
+      "loss": 0.1916,
+      "step": 70590
+    },
+    {
+      "epoch": 141.24,
+      "grad_norm": 42.008140563964844,
+      "learning_rate": 8.391640660083411e-08,
+      "loss": 0.2372,
+      "step": 70620
+    },
+    {
+      "epoch": 141.3,
+      "grad_norm": 107.05094146728516,
+      "learning_rate": 8.277397502335194e-08,
+      "loss": 0.2048,
+      "step": 70650
+    },
+    {
+      "epoch": 141.36,
+      "grad_norm": 33.05570602416992,
+      "learning_rate": 8.163930841694589e-08,
+      "loss": 0.2199,
+      "step": 70680
+    },
+    {
+      "epoch": 141.42,
+      "grad_norm": 60.28254699707031,
+      "learning_rate": 8.051240857341102e-08,
+      "loss": 0.2432,
+      "step": 70710
+    },
+    {
+      "epoch": 141.48,
+      "grad_norm": 70.86982727050781,
+      "learning_rate": 7.939327727227441e-08,
+      "loss": 0.22,
+      "step": 70740
+    },
+    {
+      "epoch": 141.54,
+      "grad_norm": 86.37015533447266,
+      "learning_rate": 7.828191628079851e-08,
+      "loss": 0.1896,
+      "step": 70770
+    },
+    {
+      "epoch": 141.6,
+      "grad_norm": 35.2357177734375,
+      "learning_rate": 7.717832735397335e-08,
+      "loss": 0.2088,
+      "step": 70800
+    },
+    {
+      "epoch": 141.66,
+      "grad_norm": 1039.4613037109375,
+      "learning_rate": 7.608251223451601e-08,
+      "loss": 0.2039,
+      "step": 70830
+    },
+    {
+      "epoch": 141.72,
+      "grad_norm": 24.711950302124023,
+      "learning_rate": 7.499447265286952e-08,
+      "loss": 0.1896,
+      "step": 70860
+    },
+    {
+      "epoch": 141.78,
+      "grad_norm": 30.11922264099121,
+      "learning_rate": 7.39142103271956e-08,
+      "loss": 0.2105,
+      "step": 70890
+    },
+    {
+      "epoch": 141.84,
+      "grad_norm": 37.19144058227539,
+      "learning_rate": 7.284172696337688e-08,
+      "loss": 0.2011,
+      "step": 70920
+    },
+    {
+      "epoch": 141.9,
+      "grad_norm": 99.82898712158203,
+      "learning_rate": 7.177702425500977e-08,
+      "loss": 0.2473,
+      "step": 70950
+    },
+    {
+      "epoch": 141.96,
+      "grad_norm": 33.77702713012695,
+      "learning_rate": 7.072010388340656e-08,
+      "loss": 0.2052,
+      "step": 70980
+    },
+    {
+      "epoch": 142.0,
+      "eval_loss": 0.308868944644928,
+      "eval_map": 0.8064,
+      "eval_map_50": 0.9478,
+      "eval_map_75": 0.9063,
+      "eval_map_chicken": 0.8072,
+      "eval_map_duck": 0.7376,
+      "eval_map_large": 0.8141,
+      "eval_map_medium": 0.8096,
+      "eval_map_plant": 0.8744,
+      "eval_map_small": 0.3715,
+      "eval_mar_1": 0.3192,
+      "eval_mar_10": 0.8427,
+      "eval_mar_100": 0.8454,
+      "eval_mar_100_chicken": 0.8468,
+      "eval_mar_100_duck": 0.7897,
+      "eval_mar_100_plant": 0.8997,
+      "eval_mar_large": 0.8407,
+      "eval_mar_medium": 0.8512,
+      "eval_mar_small": 0.4052,
+      "eval_runtime": 12.1023,
+      "eval_samples_per_second": 8.263,
+      "eval_steps_per_second": 1.074,
+      "step": 71000
+    },
+    {
+      "epoch": 142.02,
+      "grad_norm": 31.064472198486328,
+      "learning_rate": 6.967096751758773e-08,
+      "loss": 0.2284,
+      "step": 71010
+    },
+    {
+      "epoch": 142.08,
+      "grad_norm": 49.0639533996582,
+      "learning_rate": 6.862961681428304e-08,
+      "loss": 0.2014,
+      "step": 71040
+    },
+    {
+      "epoch": 142.14,
+      "grad_norm": 48.217796325683594,
+      "learning_rate": 6.759605341792819e-08,
+      "loss": 0.2459,
+      "step": 71070
+    },
+    {
+      "epoch": 142.2,
+      "grad_norm": 74.9462890625,
+      "learning_rate": 6.657027896065982e-08,
+      "loss": 0.2196,
+      "step": 71100
+    },
+    {
+      "epoch": 142.26,
+      "grad_norm": 31.848787307739258,
+      "learning_rate": 6.555229506231608e-08,
+      "loss": 0.2387,
+      "step": 71130
+    },
+    {
+      "epoch": 142.32,
+      "grad_norm": 45.96524429321289,
+      "learning_rate": 6.454210333043275e-08,
+      "loss": 0.1939,
+      "step": 71160
+    },
+    {
+      "epoch": 142.38,
+      "grad_norm": 51.15266036987305,
+      "learning_rate": 6.353970536024045e-08,
+      "loss": 0.1843,
+      "step": 71190
+    },
+    {
+      "epoch": 142.44,
+      "grad_norm": 57.012203216552734,
+      "learning_rate": 6.254510273466186e-08,
+      "loss": 0.2211,
+      "step": 71220
+    },
+    {
+      "epoch": 142.5,
+      "grad_norm": 40.6700439453125,
+      "learning_rate": 6.15582970243117e-08,
+      "loss": 0.1903,
+      "step": 71250
+    },
+    {
+      "epoch": 142.56,
+      "grad_norm": 36.734195709228516,
+      "learning_rate": 6.057928978748906e-08,
+      "loss": 0.2031,
+      "step": 71280
+    },
+    {
+      "epoch": 142.62,
+      "grad_norm": 122.18500518798828,
+      "learning_rate": 5.960808257018113e-08,
+      "loss": 0.1927,
+      "step": 71310
+    },
+    {
+      "epoch": 142.68,
+      "grad_norm": 37.88673400878906,
+      "learning_rate": 5.864467690605613e-08,
+      "loss": 0.2047,
+      "step": 71340
+    },
+    {
+      "epoch": 142.74,
+      "grad_norm": 69.3516616821289,
+      "learning_rate": 5.76890743164632e-08,
+      "loss": 0.2085,
+      "step": 71370
+    },
+    {
+      "epoch": 142.8,
+      "grad_norm": 98.10469818115234,
+      "learning_rate": 5.674127631043025e-08,
+      "loss": 0.2504,
+      "step": 71400
+    },
+    {
+      "epoch": 142.86,
+      "grad_norm": 47.47639083862305,
+      "learning_rate": 5.580128438465837e-08,
+      "loss": 0.198,
+      "step": 71430
+    },
+    {
+      "epoch": 142.92,
+      "grad_norm": 44.47444152832031,
+      "learning_rate": 5.4869100023523526e-08,
+      "loss": 0.2059,
+      "step": 71460
+    },
+    {
+      "epoch": 142.98,
+      "grad_norm": 75.74762725830078,
+      "learning_rate": 5.394472469907208e-08,
+      "loss": 0.215,
+      "step": 71490
+    },
+    {
+      "epoch": 143.0,
+      "eval_loss": 0.3087313175201416,
+      "eval_map": 0.8073,
+      "eval_map_50": 0.9476,
+      "eval_map_75": 0.9062,
+      "eval_map_chicken": 0.8082,
+      "eval_map_duck": 0.7384,
+      "eval_map_large": 0.815,
+      "eval_map_medium": 0.8087,
+      "eval_map_plant": 0.8752,
+      "eval_map_small": 0.3781,
+      "eval_mar_1": 0.3193,
+      "eval_mar_10": 0.8434,
+      "eval_mar_100": 0.8463,
+      "eval_mar_100_chicken": 0.8484,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.8997,
+      "eval_mar_large": 0.8418,
+      "eval_mar_medium": 0.8508,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 11.6975,
+      "eval_samples_per_second": 8.549,
+      "eval_steps_per_second": 1.111,
+      "step": 71500
+    },
+    {
+      "epoch": 143.04,
+      "grad_norm": 46.21229553222656,
+      "learning_rate": 5.302815987101917e-08,
+      "loss": 0.2404,
+      "step": 71520
+    },
+    {
+      "epoch": 143.1,
+      "grad_norm": 61.36137008666992,
+      "learning_rate": 5.2119406986745336e-08,
+      "loss": 0.2064,
+      "step": 71550
+    },
+    {
+      "epoch": 143.16,
+      "grad_norm": 51.88548278808594,
+      "learning_rate": 5.121846748129544e-08,
+      "loss": 0.2488,
+      "step": 71580
+    },
+    {
+      "epoch": 143.22,
+      "grad_norm": 72.71974182128906,
+      "learning_rate": 5.032534277737644e-08,
+      "loss": 0.2177,
+      "step": 71610
+    },
+    {
+      "epoch": 143.28,
+      "grad_norm": 66.9245834350586,
+      "learning_rate": 4.944003428535349e-08,
+      "loss": 0.252,
+      "step": 71640
+    },
+    {
+      "epoch": 143.34,
+      "grad_norm": 150.7808074951172,
+      "learning_rate": 4.856254340325051e-08,
+      "loss": 0.2234,
+      "step": 71670
+    },
+    {
+      "epoch": 143.4,
+      "grad_norm": 41.08302688598633,
+      "learning_rate": 4.769287151674407e-08,
+      "loss": 0.2017,
+      "step": 71700
+    },
+    {
+      "epoch": 143.46,
+      "grad_norm": 60.249473571777344,
+      "learning_rate": 4.683101999916562e-08,
+      "loss": 0.1932,
+      "step": 71730
+    },
+    {
+      "epoch": 143.52,
+      "grad_norm": 40.887020111083984,
+      "learning_rate": 4.597699021149649e-08,
+      "loss": 0.2198,
+      "step": 71760
+    },
+    {
+      "epoch": 143.58,
+      "grad_norm": 86.38909149169922,
+      "learning_rate": 4.5130783502365106e-08,
+      "loss": 0.2149,
+      "step": 71790
+    },
+    {
+      "epoch": 143.64,
+      "grad_norm": 70.79454040527344,
+      "learning_rate": 4.429240120804923e-08,
+      "loss": 0.1897,
+      "step": 71820
+    },
+    {
+      "epoch": 143.7,
+      "grad_norm": 58.97731018066406,
+      "learning_rate": 4.346184465246761e-08,
+      "loss": 0.2016,
+      "step": 71850
+    },
+    {
+      "epoch": 143.76,
+      "grad_norm": 30.449399948120117,
+      "learning_rate": 4.263911514718222e-08,
+      "loss": 0.1932,
+      "step": 71880
+    },
+    {
+      "epoch": 143.82,
+      "grad_norm": 59.97642135620117,
+      "learning_rate": 4.1824213991396024e-08,
+      "loss": 0.2003,
+      "step": 71910
+    },
+    {
+      "epoch": 143.88,
+      "grad_norm": 50.71462631225586,
+      "learning_rate": 4.10171424719491e-08,
+      "loss": 0.2051,
+      "step": 71940
+    },
+    {
+      "epoch": 143.94,
+      "grad_norm": 117.72753143310547,
+      "learning_rate": 4.0217901863317534e-08,
+      "loss": 0.1994,
+      "step": 71970
+    },
+    {
+      "epoch": 144.0,
+      "grad_norm": 62.03313064575195,
+      "learning_rate": 3.9426493427611177e-08,
+      "loss": 0.2031,
+      "step": 72000
+    },
+    {
+      "epoch": 144.0,
+      "eval_loss": 0.30865278840065,
+      "eval_map": 0.8065,
+      "eval_map_50": 0.9475,
+      "eval_map_75": 0.9062,
+      "eval_map_chicken": 0.8062,
+      "eval_map_duck": 0.7383,
+      "eval_map_large": 0.8107,
+      "eval_map_medium": 0.8091,
+      "eval_map_plant": 0.8752,
+      "eval_map_small": 0.3766,
+      "eval_mar_1": 0.3191,
+      "eval_mar_10": 0.843,
+      "eval_mar_100": 0.8458,
+      "eval_mar_100_chicken": 0.8476,
+      "eval_mar_100_duck": 0.7897,
+      "eval_mar_100_plant": 0.9,
+      "eval_mar_large": 0.8384,
+      "eval_mar_medium": 0.8511,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 13.6489,
+      "eval_samples_per_second": 7.327,
+      "eval_steps_per_second": 0.952,
+      "step": 72000
+    },
+    {
+      "epoch": 144.06,
+      "grad_norm": 35.4929313659668,
+      "learning_rate": 3.864291841457146e-08,
+      "loss": 0.1833,
+      "step": 72030
+    },
+    {
+      "epoch": 144.12,
+      "grad_norm": 87.02384185791016,
+      "learning_rate": 3.786717806157136e-08,
+      "loss": 0.2104,
+      "step": 72060
+    },
+    {
+      "epoch": 144.18,
+      "grad_norm": 178.31637573242188,
+      "learning_rate": 3.7099273593609316e-08,
+      "loss": 0.2086,
+      "step": 72090
+    },
+    {
+      "epoch": 144.24,
+      "grad_norm": 53.136329650878906,
+      "learning_rate": 3.633920622331311e-08,
+      "loss": 0.2387,
+      "step": 72120
+    },
+    {
+      "epoch": 144.3,
+      "grad_norm": 78.11143493652344,
+      "learning_rate": 3.558697715093207e-08,
+      "loss": 0.2213,
+      "step": 72150
+    },
+    {
+      "epoch": 144.36,
+      "grad_norm": 46.717445373535156,
+      "learning_rate": 3.4842587564337674e-08,
+      "loss": 0.2149,
+      "step": 72180
+    },
+    {
+      "epoch": 144.42,
+      "grad_norm": 30.410228729248047,
+      "learning_rate": 3.410603863902406e-08,
+      "loss": 0.2304,
+      "step": 72210
+    },
+    {
+      "epoch": 144.48,
+      "grad_norm": 99.21456909179688,
+      "learning_rate": 3.337733153810141e-08,
+      "loss": 0.2044,
+      "step": 72240
+    },
+    {
+      "epoch": 144.54,
+      "grad_norm": 67.04376983642578,
+      "learning_rate": 3.2656467412298665e-08,
+      "loss": 0.2052,
+      "step": 72270
+    },
+    {
+      "epoch": 144.6,
+      "grad_norm": 64.57819366455078,
+      "learning_rate": 3.194344739995803e-08,
+      "loss": 0.2304,
+      "step": 72300
+    },
+    {
+      "epoch": 144.66,
+      "grad_norm": 93.20011901855469,
+      "learning_rate": 3.1238272627035494e-08,
+      "loss": 0.1735,
+      "step": 72330
+    },
+    {
+      "epoch": 144.72,
+      "grad_norm": 44.79425048828125,
+      "learning_rate": 3.054094420709863e-08,
+      "loss": 0.2073,
+      "step": 72360
+    },
+    {
+      "epoch": 144.78,
+      "grad_norm": 83.33869934082031,
+      "learning_rate": 2.985146324132438e-08,
+      "loss": 0.1901,
+      "step": 72390
+    },
+    {
+      "epoch": 144.84,
+      "grad_norm": 53.1854248046875,
+      "learning_rate": 2.9169830818496226e-08,
+      "loss": 0.2056,
+      "step": 72420
+    },
+    {
+      "epoch": 144.9,
+      "grad_norm": 43.01469802856445,
+      "learning_rate": 2.8496048015005385e-08,
+      "loss": 0.2146,
+      "step": 72450
+    },
+    {
+      "epoch": 144.96,
+      "grad_norm": 209.90834045410156,
+      "learning_rate": 2.783011589484741e-08,
+      "loss": 0.2042,
+      "step": 72480
+    },
+    {
+      "epoch": 145.0,
+      "eval_loss": 0.30925503373146057,
+      "eval_map": 0.8056,
+      "eval_map_50": 0.9476,
+      "eval_map_75": 0.9048,
+      "eval_map_chicken": 0.8045,
+      "eval_map_duck": 0.737,
+      "eval_map_large": 0.8105,
+      "eval_map_medium": 0.8082,
+      "eval_map_plant": 0.8752,
+      "eval_map_small": 0.3781,
+      "eval_mar_1": 0.3189,
+      "eval_mar_10": 0.8425,
+      "eval_mar_100": 0.8452,
+      "eval_mar_100_chicken": 0.846,
+      "eval_mar_100_duck": 0.7897,
+      "eval_mar_100_plant": 0.9,
+      "eval_mar_large": 0.8374,
+      "eval_mar_medium": 0.8506,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 12.735,
+      "eval_samples_per_second": 7.852,
+      "eval_steps_per_second": 1.021,
+      "step": 72500
+    },
+    {
+      "epoch": 145.02,
+      "grad_norm": 41.61909484863281,
+      "learning_rate": 2.7172035509619442e-08,
+      "loss": 0.2153,
+      "step": 72510
+    },
+    {
+      "epoch": 145.08,
+      "grad_norm": 92.02596282958984,
+      "learning_rate": 2.6521807898520214e-08,
+      "loss": 0.2094,
+      "step": 72540
+    },
+    {
+      "epoch": 145.14,
+      "grad_norm": 56.887611389160156,
+      "learning_rate": 2.5879434088348364e-08,
+      "loss": 0.2085,
+      "step": 72570
+    },
+    {
+      "epoch": 145.2,
+      "grad_norm": 26.000112533569336,
+      "learning_rate": 2.5244915093499134e-08,
+      "loss": 0.2036,
+      "step": 72600
+    },
+    {
+      "epoch": 145.26,
+      "grad_norm": 32.90476989746094,
+      "learning_rate": 2.46182519159649e-08,
+      "loss": 0.1947,
+      "step": 72630
+    },
+    {
+      "epoch": 145.32,
+      "grad_norm": 55.35662078857422,
+      "learning_rate": 2.3999445545332955e-08,
+      "loss": 0.2019,
+      "step": 72660
+    },
+    {
+      "epoch": 145.38,
+      "grad_norm": 91.33636474609375,
+      "learning_rate": 2.3388496958782203e-08,
+      "loss": 0.252,
+      "step": 72690
+    },
+    {
+      "epoch": 145.44,
+      "grad_norm": 83.00819396972656,
+      "learning_rate": 2.2785407121084236e-08,
+      "loss": 0.2069,
+      "step": 72720
+    },
+    {
+      "epoch": 145.5,
+      "grad_norm": 42.86649703979492,
+      "learning_rate": 2.219017698460002e-08,
+      "loss": 0.2156,
+      "step": 72750
+    },
+    {
+      "epoch": 145.56,
+      "grad_norm": 86.97473907470703,
+      "learning_rate": 2.1602807489279344e-08,
+      "loss": 0.2129,
+      "step": 72780
+    },
+    {
+      "epoch": 145.62,
+      "grad_norm": 56.99577713012695,
+      "learning_rate": 2.1023299562658584e-08,
+      "loss": 0.2005,
+      "step": 72810
+    },
+    {
+      "epoch": 145.68,
+      "grad_norm": 61.787715911865234,
+      "learning_rate": 2.0451654119860164e-08,
+      "loss": 0.2379,
+      "step": 72840
+    },
+    {
+      "epoch": 145.74,
+      "grad_norm": 989.9368286132812,
+      "learning_rate": 1.988787206359033e-08,
+      "loss": 0.2225,
+      "step": 72870
+    },
+    {
+      "epoch": 145.8,
+      "grad_norm": 75.80166625976562,
+      "learning_rate": 1.9331954284137476e-08,
+      "loss": 0.209,
+      "step": 72900
+    },
+    {
+      "epoch": 145.86,
+      "grad_norm": 48.081207275390625,
+      "learning_rate": 1.8783901659372162e-08,
+      "loss": 0.2186,
+      "step": 72930
+    },
+    {
+      "epoch": 145.92,
+      "grad_norm": 43.130218505859375,
+      "learning_rate": 1.8243715054744315e-08,
+      "loss": 0.185,
+      "step": 72960
+    },
+    {
+      "epoch": 145.98,
+      "grad_norm": 77.66969299316406,
+      "learning_rate": 1.7711395323281588e-08,
+      "loss": 0.1978,
+      "step": 72990
+    },
+    {
+      "epoch": 146.0,
+      "eval_loss": 0.30891692638397217,
+      "eval_map": 0.8065,
+      "eval_map_50": 0.9475,
+      "eval_map_75": 0.9047,
+      "eval_map_chicken": 0.8064,
+      "eval_map_duck": 0.7372,
+      "eval_map_large": 0.815,
+      "eval_map_medium": 0.8091,
+      "eval_map_plant": 0.876,
+      "eval_map_small": 0.3781,
+      "eval_mar_1": 0.3197,
+      "eval_mar_10": 0.8436,
+      "eval_mar_100": 0.8463,
+      "eval_mar_100_chicken": 0.848,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.9003,
+      "eval_mar_large": 0.8418,
+      "eval_mar_medium": 0.8515,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 12.7636,
+      "eval_samples_per_second": 7.835,
+      "eval_steps_per_second": 1.019,
+      "step": 73000
+    },
+    {
+      "epoch": 146.04,
+      "grad_norm": 51.415863037109375,
+      "learning_rate": 1.71869433055899e-08,
+      "loss": 0.2065,
+      "step": 73020
+    },
+    {
+      "epoch": 146.1,
+      "grad_norm": 55.093109130859375,
+      "learning_rate": 1.6670359829850657e-08,
+      "loss": 0.199,
+      "step": 73050
+    },
+    {
+      "epoch": 146.16,
+      "grad_norm": 67.57818603515625,
+      "learning_rate": 1.6161645711819664e-08,
+      "loss": 0.2043,
+      "step": 73080
+    },
+    {
+      "epoch": 146.22,
+      "grad_norm": 78.64994812011719,
+      "learning_rate": 1.5660801754825983e-08,
+      "loss": 0.1761,
+      "step": 73110
+    },
+    {
+      "epoch": 146.28,
+      "grad_norm": 44.29743957519531,
+      "learning_rate": 1.5167828749770853e-08,
+      "loss": 0.2297,
+      "step": 73140
+    },
+    {
+      "epoch": 146.34,
+      "grad_norm": 56.25002670288086,
+      "learning_rate": 1.4682727475124891e-08,
+      "loss": 0.1894,
+      "step": 73170
+    },
+    {
+      "epoch": 146.4,
+      "grad_norm": 81.73661041259766,
+      "learning_rate": 1.4205498696930332e-08,
+      "loss": 0.1852,
+      "step": 73200
+    },
+    {
+      "epoch": 146.46,
+      "grad_norm": 29.35874366760254,
+      "learning_rate": 1.3736143168796012e-08,
+      "loss": 0.2025,
+      "step": 73230
+    },
+    {
+      "epoch": 146.52,
+      "grad_norm": 29.727802276611328,
+      "learning_rate": 1.3274661631899055e-08,
+      "loss": 0.2379,
+      "step": 73260
+    },
+    {
+      "epoch": 146.58,
+      "grad_norm": 66.46299743652344,
+      "learning_rate": 1.2821054814980971e-08,
+      "loss": 0.2061,
+      "step": 73290
+    },
+    {
+      "epoch": 146.64,
+      "grad_norm": 51.29338455200195,
+      "learning_rate": 1.2375323434348773e-08,
+      "loss": 0.2488,
+      "step": 73320
+    },
+    {
+      "epoch": 146.7,
+      "grad_norm": 35.33070373535156,
+      "learning_rate": 1.1937468193873869e-08,
+      "loss": 0.2273,
+      "step": 73350
+    },
+    {
+      "epoch": 146.76,
+      "grad_norm": 45.7594108581543,
+      "learning_rate": 1.1507489784989278e-08,
+      "loss": 0.2235,
+      "step": 73380
+    },
+    {
+      "epoch": 146.82,
+      "grad_norm": 50.467529296875,
+      "learning_rate": 1.1085388886689085e-08,
+      "loss": 0.2128,
+      "step": 73410
+    },
+    {
+      "epoch": 146.88,
+      "grad_norm": 61.66533660888672,
+      "learning_rate": 1.067116616552899e-08,
+      "loss": 0.2061,
+      "step": 73440
+    },
+    {
+      "epoch": 146.94,
+      "grad_norm": 58.77129364013672,
+      "learning_rate": 1.026482227562242e-08,
+      "loss": 0.228,
+      "step": 73470
+    },
+    {
+      "epoch": 147.0,
+      "grad_norm": 32.76871109008789,
+      "learning_rate": 9.866357858642206e-09,
+      "loss": 0.1969,
+      "step": 73500
+    },
+    {
+      "epoch": 147.0,
+      "eval_loss": 0.30892306566238403,
+      "eval_map": 0.8063,
+      "eval_map_50": 0.9477,
+      "eval_map_75": 0.9049,
+      "eval_map_chicken": 0.8065,
+      "eval_map_duck": 0.7377,
+      "eval_map_large": 0.8145,
+      "eval_map_medium": 0.8093,
+      "eval_map_plant": 0.8748,
+      "eval_map_small": 0.3766,
+      "eval_mar_1": 0.3195,
+      "eval_mar_10": 0.8434,
+      "eval_mar_100": 0.8462,
+      "eval_mar_100_chicken": 0.8476,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.9003,
+      "eval_mar_large": 0.8408,
+      "eval_mar_medium": 0.8515,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 12.1018,
+      "eval_samples_per_second": 8.263,
+      "eval_steps_per_second": 1.074,
+      "step": 73500
+    },
+    {
+      "epoch": 147.06,
+      "grad_norm": 63.24929428100586,
+      "learning_rate": 9.475773543818345e-09,
+      "loss": 0.2078,
+      "step": 73530
+    },
+    {
+      "epoch": 147.12,
+      "grad_norm": 45.290950775146484,
+      "learning_rate": 9.09306994793635e-09,
+      "loss": 0.2573,
+      "step": 73560
+    },
+    {
+      "epoch": 147.18,
+      "grad_norm": 52.7389030456543,
+      "learning_rate": 8.718247675337243e-09,
+      "loss": 0.193,
+      "step": 73590
+    },
+    {
+      "epoch": 147.24,
+      "grad_norm": 38.60297393798828,
+      "learning_rate": 8.351307317917002e-09,
+      "loss": 0.221,
+      "step": 73620
+    },
+    {
+      "epoch": 147.3,
+      "grad_norm": 35.21743392944336,
+      "learning_rate": 7.992249455124889e-09,
+      "loss": 0.2271,
+      "step": 73650
+    },
+    {
+      "epoch": 147.36,
+      "grad_norm": 44.66568374633789,
+      "learning_rate": 7.641074653961244e-09,
+      "loss": 0.2287,
+      "step": 73680
+    },
+    {
+      "epoch": 147.42,
+      "grad_norm": 30.76506805419922,
+      "learning_rate": 7.297783468980246e-09,
+      "loss": 0.2162,
+      "step": 73710
+    },
+    {
+      "epoch": 147.48,
+      "grad_norm": 44.26003646850586,
+      "learning_rate": 6.962376442284368e-09,
+      "loss": 0.22,
+      "step": 73740
+    },
+    {
+      "epoch": 147.54,
+      "grad_norm": 78.09246826171875,
+      "learning_rate": 6.63485410352771e-09,
+      "loss": 0.1989,
+      "step": 73770
+    },
+    {
+      "epoch": 147.6,
+      "grad_norm": 64.34059143066406,
+      "learning_rate": 6.315216969912663e-09,
+      "loss": 0.1954,
+      "step": 73800
+    },
+    {
+      "epoch": 147.66,
+      "grad_norm": 80.68038940429688,
+      "learning_rate": 6.003465546189358e-09,
+      "loss": 0.1898,
+      "step": 73830
+    },
+    {
+      "epoch": 147.72,
+      "grad_norm": 37.143863677978516,
+      "learning_rate": 5.699600324657328e-09,
+      "loss": 0.1896,
+      "step": 73860
+    },
+    {
+      "epoch": 147.78,
+      "grad_norm": 86.01634216308594,
+      "learning_rate": 5.403621785159407e-09,
+      "loss": 0.2144,
+      "step": 73890
+    },
+    {
+      "epoch": 147.84,
+      "grad_norm": 43.97216796875,
+      "learning_rate": 5.115530395087276e-09,
+      "loss": 0.213,
+      "step": 73920
+    },
+    {
+      "epoch": 147.9,
+      "grad_norm": 38.47873306274414,
+      "learning_rate": 4.835326609376468e-09,
+      "loss": 0.2246,
+      "step": 73950
+    },
+    {
+      "epoch": 147.96,
+      "grad_norm": 35.541812896728516,
+      "learning_rate": 4.5630108705063684e-09,
+      "loss": 0.2083,
+      "step": 73980
+    },
+    {
+      "epoch": 148.0,
+      "eval_loss": 0.30893582105636597,
+      "eval_map": 0.8062,
+      "eval_map_50": 0.9477,
+      "eval_map_75": 0.9049,
+      "eval_map_chicken": 0.8066,
+      "eval_map_duck": 0.7377,
+      "eval_map_large": 0.8149,
+      "eval_map_medium": 0.8098,
+      "eval_map_plant": 0.8744,
+      "eval_map_small": 0.3781,
+      "eval_mar_1": 0.3198,
+      "eval_mar_10": 0.8435,
+      "eval_mar_100": 0.8463,
+      "eval_mar_100_chicken": 0.848,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.9003,
+      "eval_mar_large": 0.8417,
+      "eval_mar_medium": 0.8519,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 13.6671,
+      "eval_samples_per_second": 7.317,
+      "eval_steps_per_second": 0.951,
+      "step": 74000
+    },
+    {
+      "epoch": 148.02,
+      "grad_norm": 56.26382064819336,
+      "learning_rate": 4.298583608501328e-09,
+      "loss": 0.2157,
+      "step": 74010
+    },
+    {
+      "epoch": 148.08,
+      "grad_norm": 32.970672607421875,
+      "learning_rate": 4.042045240927883e-09,
+      "loss": 0.2133,
+      "step": 74040
+    },
+    {
+      "epoch": 148.14,
+      "grad_norm": 63.352481842041016,
+      "learning_rate": 3.793396172895314e-09,
+      "loss": 0.2116,
+      "step": 74070
+    },
+    {
+      "epoch": 148.2,
+      "grad_norm": 88.44683837890625,
+      "learning_rate": 3.5526367970539765e-09,
+      "loss": 0.1971,
+      "step": 74100
+    },
+    {
+      "epoch": 148.26,
+      "grad_norm": 36.679317474365234,
+      "learning_rate": 3.31976749359586e-09,
+      "loss": 0.1952,
+      "step": 74130
+    },
+    {
+      "epoch": 148.32,
+      "grad_norm": 34.160804748535156,
+      "learning_rate": 3.094788630254031e-09,
+      "loss": 0.2392,
+      "step": 74160
+    },
+    {
+      "epoch": 148.38,
+      "grad_norm": 50.7787971496582,
+      "learning_rate": 2.8777005622998567e-09,
+      "loss": 0.2159,
+      "step": 74190
+    },
+    {
+      "epoch": 148.44,
+      "grad_norm": 105.82192993164062,
+      "learning_rate": 2.6685036325457826e-09,
+      "loss": 0.2018,
+      "step": 74220
+    },
+    {
+      "epoch": 148.5,
+      "grad_norm": 52.19423294067383,
+      "learning_rate": 2.4671981713420003e-09,
+      "loss": 0.2117,
+      "step": 74250
+    },
+    {
+      "epoch": 148.56,
+      "grad_norm": 54.85720443725586,
+      "learning_rate": 2.2737844965775578e-09,
+      "loss": 0.2018,
+      "step": 74280
+    },
+    {
+      "epoch": 148.62,
+      "grad_norm": 67.35590362548828,
+      "learning_rate": 2.088262913679251e-09,
+      "loss": 0.1987,
+      "step": 74310
+    },
+    {
+      "epoch": 148.68,
+      "grad_norm": 65.11038208007812,
+      "learning_rate": 1.9106337156099553e-09,
+      "loss": 0.2329,
+      "step": 74340
+    },
+    {
+      "epoch": 148.74,
+      "grad_norm": 74.10694122314453,
+      "learning_rate": 1.740897182871404e-09,
+      "loss": 0.2376,
+      "step": 74370
+    },
+    {
+      "epoch": 148.8,
+      "grad_norm": 73.60809326171875,
+      "learning_rate": 1.5790535835003006e-09,
+      "loss": 0.2149,
+      "step": 74400
+    },
+    {
+      "epoch": 148.86,
+      "grad_norm": 42.631622314453125,
+      "learning_rate": 1.425103173069986e-09,
+      "loss": 0.1995,
+      "step": 74430
+    },
+    {
+      "epoch": 148.92,
+      "grad_norm": 63.96980667114258,
+      "learning_rate": 1.2790461946887712e-09,
+      "loss": 0.1897,
+      "step": 74460
+    },
+    {
+      "epoch": 148.98,
+      "grad_norm": 71.41777801513672,
+      "learning_rate": 1.1408828790010484e-09,
+      "loss": 0.1991,
+      "step": 74490
+    },
+    {
+      "epoch": 149.0,
+      "eval_loss": 0.3090154230594635,
+      "eval_map": 0.8062,
+      "eval_map_50": 0.9477,
+      "eval_map_75": 0.9049,
+      "eval_map_chicken": 0.8066,
+      "eval_map_duck": 0.7377,
+      "eval_map_large": 0.8149,
+      "eval_map_medium": 0.8098,
+      "eval_map_plant": 0.8745,
+      "eval_map_small": 0.3781,
+      "eval_mar_1": 0.3198,
+      "eval_mar_10": 0.8436,
+      "eval_mar_100": 0.8464,
+      "eval_mar_100_chicken": 0.848,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.9006,
+      "eval_mar_large": 0.8418,
+      "eval_mar_medium": 0.8519,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 13.9979,
+      "eval_samples_per_second": 7.144,
+      "eval_steps_per_second": 0.929,
+      "step": 74500
+    },
+    {
+      "epoch": 149.04,
+      "grad_norm": 102.14067840576172,
+      "learning_rate": 1.0106134441850712e-09,
+      "loss": 0.2216,
+      "step": 74520
+    },
+    {
+      "epoch": 149.1,
+      "grad_norm": 42.29328155517578,
+      "learning_rate": 8.88238095955174e-10,
+      "loss": 0.1898,
+      "step": 74550
+    },
+    {
+      "epoch": 149.16,
+      "grad_norm": 47.901248931884766,
+      "learning_rate": 7.737570275573314e-10,
+      "loss": 0.2068,
+      "step": 74580
+    },
+    {
+      "epoch": 149.22,
+      "grad_norm": 41.09926986694336,
+      "learning_rate": 6.671704197735995e-10,
+      "loss": 0.2043,
+      "step": 74610
+    },
+    {
+      "epoch": 149.28,
+      "grad_norm": 95.37877655029297,
+      "learning_rate": 5.684784409182298e-10,
+      "loss": 0.2002,
+      "step": 74640
+    },
+    {
+      "epoch": 149.34,
+      "grad_norm": 86.34088897705078,
+      "learning_rate": 4.776812468398895e-10,
+      "loss": 0.2129,
+      "step": 74670
+    },
+    {
+      "epoch": 149.4,
+      "grad_norm": 50.384944915771484,
+      "learning_rate": 3.9477898091944135e-10,
+      "loss": 0.2064,
+      "step": 74700
+    },
+    {
+      "epoch": 149.46,
+      "grad_norm": 44.60343933105469,
+      "learning_rate": 3.1977177407105376e-10,
+      "loss": 0.2034,
+      "step": 74730
+    },
+    {
+      "epoch": 149.52,
+      "grad_norm": 58.12319564819336,
+      "learning_rate": 2.5265974474109054e-10,
+      "loss": 0.221,
+      "step": 74760
+    },
+    {
+      "epoch": 149.58,
+      "grad_norm": 49.90918731689453,
+      "learning_rate": 1.9344299890866614e-10,
+      "loss": 0.2165,
+      "step": 74790
+    },
+    {
+      "epoch": 149.64,
+      "grad_norm": 56.82841873168945,
+      "learning_rate": 1.4212163008509028e-10,
+      "loss": 0.1935,
+      "step": 74820
+    },
+    {
+      "epoch": 149.7,
+      "grad_norm": 54.94172668457031,
+      "learning_rate": 9.869571931442334e-11,
+      "loss": 0.264,
+      "step": 74850
+    },
+    {
+      "epoch": 149.76,
+      "grad_norm": 49.350772857666016,
+      "learning_rate": 6.316533517125578e-11,
+      "loss": 0.217,
+      "step": 74880
+    },
+    {
+      "epoch": 149.82,
+      "grad_norm": 45.621150970458984,
+      "learning_rate": 3.55305337634837e-11,
+      "loss": 0.2009,
+      "step": 74910
+    },
+    {
+      "epoch": 149.88,
+      "grad_norm": 50.17906951904297,
+      "learning_rate": 1.57913587295333e-11,
+      "loss": 0.2077,
+      "step": 74940
+    },
+    {
+      "epoch": 149.94,
+      "grad_norm": 19.391836166381836,
+      "learning_rate": 3.947841241136452e-12,
+      "loss": 0.224,
+      "step": 74970
+    },
+    {
+      "epoch": 150.0,
+      "grad_norm": 35.242279052734375,
+      "learning_rate": 0.0,
+      "loss": 0.2337,
+      "step": 75000
+    },
+    {
+      "epoch": 150.0,
+      "eval_loss": 0.30901747941970825,
+      "eval_map": 0.8062,
+      "eval_map_50": 0.9477,
+      "eval_map_75": 0.9049,
+      "eval_map_chicken": 0.8066,
+      "eval_map_duck": 0.7377,
+      "eval_map_large": 0.8149,
+      "eval_map_medium": 0.8098,
+      "eval_map_plant": 0.8745,
+      "eval_map_small": 0.3781,
+      "eval_mar_1": 0.3198,
+      "eval_mar_10": 0.8436,
+      "eval_mar_100": 0.8464,
+      "eval_mar_100_chicken": 0.848,
+      "eval_mar_100_duck": 0.7907,
+      "eval_mar_100_plant": 0.9006,
+      "eval_mar_large": 0.8418,
+      "eval_mar_medium": 0.8519,
+      "eval_mar_small": 0.42,
+      "eval_runtime": 17.4562,
+      "eval_samples_per_second": 5.729,
+      "eval_steps_per_second": 0.745,
+      "step": 75000
     }
   ],
   "logging_steps": 30,
@@ -15995,12 +21421,12 @@
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": false
+        "should_training_stop": true
       },
       "attributes": {}
     }
   },
-  "total_flos": 1.92652943228928e+19,
+  "total_flos": 2.580173346816e+19,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null