diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 16.0,
+  "epoch": 160.0,
   "eval_steps": 500,
-  "global_step": 2000,
+  "global_step": 20000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -389,19 +389,3439 @@
       "step": 2000
     },
     {
-      "epoch": 16.0,
-      "step": 2000,
-      "total_flos": 2.80447288344576e+17,
-      "train_loss": 0.1418005952835083,
-      "train_runtime": 624.6204,
-      "train_samples_per_second": 25.616,
-      "train_steps_per_second": 3.202
+      "epoch": 16.8,
+      "grad_norm": 1.944307565689087,
+      "learning_rate": 5.37105e-05,
+      "loss": 0.1072,
+      "step": 2100
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy_no_text": 0.9776347749308564,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9776347749308564,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10060456395149231,
+      "eval_mean_accuracy": 0.9776347749308564,
+      "eval_mean_iou": 0.4888173874654282,
+      "eval_overall_accuracy": 0.9776347749308564,
+      "eval_runtime": 8.4116,
+      "eval_samples_per_second": 30.196,
+      "eval_steps_per_second": 3.804,
+      "step": 2125
+    },
+    {
+      "epoch": 17.6,
+      "grad_norm": 0.7363667488098145,
+      "learning_rate": 5.3411e-05,
+      "loss": 0.1046,
+      "step": 2200
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy_no_text": 0.9800149224455696,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9800149224455696,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09380830079317093,
+      "eval_mean_accuracy": 0.9800149224455696,
+      "eval_mean_iou": 0.4900074612227848,
+      "eval_overall_accuracy": 0.9800149224455696,
+      "eval_runtime": 8.0943,
+      "eval_samples_per_second": 31.38,
+      "eval_steps_per_second": 3.953,
+      "step": 2250
+    },
+    {
+      "epoch": 18.4,
+      "grad_norm": 1.9069602489471436,
+      "learning_rate": 5.31115e-05,
+      "loss": 0.1072,
+      "step": 2300
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy_no_text": 0.9799700770156704,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9799700770156704,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09618178755044937,
+      "eval_mean_accuracy": 0.9799700770156704,
+      "eval_mean_iou": 0.4899850385078352,
+      "eval_overall_accuracy": 0.9799700770156704,
+      "eval_runtime": 8.171,
+      "eval_samples_per_second": 31.086,
+      "eval_steps_per_second": 3.916,
+      "step": 2375
+    },
+    {
+      "epoch": 19.2,
+      "grad_norm": 0.44612982869148254,
+      "learning_rate": 5.2812e-05,
+      "loss": 0.1062,
+      "step": 2400
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 1.044952392578125,
+      "learning_rate": 5.25125e-05,
+      "loss": 0.1127,
+      "step": 2500
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy_no_text": 0.98404203872741,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.98404203872741,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09183745831251144,
+      "eval_mean_accuracy": 0.98404203872741,
+      "eval_mean_iou": 0.492021019363705,
+      "eval_overall_accuracy": 0.98404203872741,
+      "eval_runtime": 8.4621,
+      "eval_samples_per_second": 30.016,
+      "eval_steps_per_second": 3.782,
+      "step": 2500
+    },
+    {
+      "epoch": 20.8,
+      "grad_norm": 3.0239295959472656,
+      "learning_rate": 5.2213e-05,
+      "loss": 0.1017,
+      "step": 2600
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy_no_text": 0.9781973615966435,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9781973615966435,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09400010854005814,
+      "eval_mean_accuracy": 0.9781973615966435,
+      "eval_mean_iou": 0.48909868079832175,
+      "eval_overall_accuracy": 0.9781973615966435,
+      "eval_runtime": 8.2162,
+      "eval_samples_per_second": 30.915,
+      "eval_steps_per_second": 3.895,
+      "step": 2625
+    },
+    {
+      "epoch": 21.6,
+      "grad_norm": 1.0735955238342285,
+      "learning_rate": 5.1913500000000004e-05,
+      "loss": 0.0961,
+      "step": 2700
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy_no_text": 0.9784010685142297,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9784010685142297,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09638506919145584,
+      "eval_mean_accuracy": 0.9784010685142297,
+      "eval_mean_iou": 0.48920053425711485,
+      "eval_overall_accuracy": 0.9784010685142297,
+      "eval_runtime": 8.6739,
+      "eval_samples_per_second": 29.283,
+      "eval_steps_per_second": 3.689,
+      "step": 2750
+    },
+    {
+      "epoch": 22.4,
+      "grad_norm": 0.7935839891433716,
+      "learning_rate": 5.1614000000000004e-05,
+      "loss": 0.0951,
+      "step": 2800
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy_no_text": 0.9820980998348329,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9820980998348329,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09395471215248108,
+      "eval_mean_accuracy": 0.9820980998348329,
+      "eval_mean_iou": 0.49104904991741644,
+      "eval_overall_accuracy": 0.9820980998348329,
+      "eval_runtime": 8.6006,
+      "eval_samples_per_second": 29.533,
+      "eval_steps_per_second": 3.721,
+      "step": 2875
+    },
+    {
+      "epoch": 23.2,
+      "grad_norm": 1.0572441816329956,
+      "learning_rate": 5.13145e-05,
+      "loss": 0.101,
+      "step": 2900
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 0.9966760873794556,
+      "learning_rate": 5.1015e-05,
+      "loss": 0.0938,
+      "step": 3000
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy_no_text": 0.983619405027922,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.983619405027922,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10053908079862595,
+      "eval_mean_accuracy": 0.983619405027922,
+      "eval_mean_iou": 0.491809702513961,
+      "eval_overall_accuracy": 0.983619405027922,
+      "eval_runtime": 8.5467,
+      "eval_samples_per_second": 29.719,
+      "eval_steps_per_second": 3.744,
+      "step": 3000
+    },
+    {
+      "epoch": 24.8,
+      "grad_norm": 1.7212049961090088,
+      "learning_rate": 5.07155e-05,
+      "loss": 0.0949,
+      "step": 3100
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy_no_text": 0.9802790768155358,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9802790768155358,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10030657052993774,
+      "eval_mean_accuracy": 0.9802790768155358,
+      "eval_mean_iou": 0.4901395384077679,
+      "eval_overall_accuracy": 0.9802790768155358,
+      "eval_runtime": 8.5434,
+      "eval_samples_per_second": 29.731,
+      "eval_steps_per_second": 3.746,
+      "step": 3125
+    },
+    {
+      "epoch": 25.6,
+      "grad_norm": 0.9479594230651855,
+      "learning_rate": 5.0416e-05,
+      "loss": 0.0949,
+      "step": 3200
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy_no_text": 0.9815445520404297,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9815445520404297,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10147000104188919,
+      "eval_mean_accuracy": 0.9815445520404297,
+      "eval_mean_iou": 0.49077227602021484,
+      "eval_overall_accuracy": 0.9815445520404297,
+      "eval_runtime": 8.3101,
+      "eval_samples_per_second": 30.565,
+      "eval_steps_per_second": 3.851,
+      "step": 3250
+    },
+    {
+      "epoch": 26.4,
+      "grad_norm": 2.8663785457611084,
+      "learning_rate": 5.01165e-05,
+      "loss": 0.0949,
+      "step": 3300
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy_no_text": 0.9779668869919259,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9779668869919259,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0970182940363884,
+      "eval_mean_accuracy": 0.9779668869919259,
+      "eval_mean_iou": 0.48898344349596295,
+      "eval_overall_accuracy": 0.9779668869919259,
+      "eval_runtime": 8.507,
+      "eval_samples_per_second": 29.858,
+      "eval_steps_per_second": 3.762,
+      "step": 3375
+    },
+    {
+      "epoch": 27.2,
+      "grad_norm": 0.29338690638542175,
+      "learning_rate": 5.15e-05,
+      "loss": 0.0977,
+      "step": 3400
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 0.5164900422096252,
+      "learning_rate": 5.125e-05,
+      "loss": 0.0883,
+      "step": 3500
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy_no_text": 0.9779090481844123,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9779090481844123,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09671156853437424,
+      "eval_mean_accuracy": 0.9779090481844123,
+      "eval_mean_iou": 0.48895452409220613,
+      "eval_overall_accuracy": 0.9779090481844123,
+      "eval_runtime": 8.4255,
+      "eval_samples_per_second": 30.147,
+      "eval_steps_per_second": 3.798,
+      "step": 3500
+    },
+    {
+      "epoch": 28.8,
+      "grad_norm": 0.9270356893539429,
+      "learning_rate": 5.1e-05,
+      "loss": 0.0846,
+      "step": 3600
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy_no_text": 0.984867849291109,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.984867849291109,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09728837013244629,
+      "eval_mean_accuracy": 0.984867849291109,
+      "eval_mean_iou": 0.4924339246455545,
+      "eval_overall_accuracy": 0.984867849291109,
+      "eval_runtime": 8.3583,
+      "eval_samples_per_second": 30.389,
+      "eval_steps_per_second": 3.829,
+      "step": 3625
+    },
+    {
+      "epoch": 29.6,
+      "grad_norm": 54.255714416503906,
+      "learning_rate": 5.075000000000001e-05,
+      "loss": 0.0842,
+      "step": 3700
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy_no_text": 0.9820055344810097,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9820055344810097,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0946170911192894,
+      "eval_mean_accuracy": 0.9820055344810097,
+      "eval_mean_iou": 0.49100276724050484,
+      "eval_overall_accuracy": 0.9820055344810097,
+      "eval_runtime": 8.1044,
+      "eval_samples_per_second": 31.341,
+      "eval_steps_per_second": 3.948,
+      "step": 3750
+    },
+    {
+      "epoch": 30.4,
+      "grad_norm": 0.4858163297176361,
+      "learning_rate": 5.050000000000001e-05,
+      "loss": 0.0814,
+      "step": 3800
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy_no_text": 0.9818622916313758,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9818622916313758,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09355071932077408,
+      "eval_mean_accuracy": 0.9818622916313758,
+      "eval_mean_iou": 0.4909311458156879,
+      "eval_overall_accuracy": 0.9818622916313758,
+      "eval_runtime": 8.3948,
+      "eval_samples_per_second": 30.257,
+      "eval_steps_per_second": 3.812,
+      "step": 3875
+    },
+    {
+      "epoch": 31.2,
+      "grad_norm": 0.7054563760757446,
+      "learning_rate": 5.0249999999999995e-05,
+      "loss": 0.082,
+      "step": 3900
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 1.9450538158416748,
+      "learning_rate": 5e-05,
+      "loss": 0.0813,
+      "step": 4000
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy_no_text": 0.9812941719799806,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9812941719799806,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09379926323890686,
+      "eval_mean_accuracy": 0.9812941719799806,
+      "eval_mean_iou": 0.4906470859899903,
+      "eval_overall_accuracy": 0.9812941719799806,
+      "eval_runtime": 8.3235,
+      "eval_samples_per_second": 30.516,
+      "eval_steps_per_second": 3.845,
+      "step": 4000
+    },
+    {
+      "epoch": 32.8,
+      "grad_norm": 0.48973309993743896,
+      "learning_rate": 4.975e-05,
+      "loss": 0.0817,
+      "step": 4100
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy_no_text": 0.9812250511988094,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9812250511988094,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09455174952745438,
+      "eval_mean_accuracy": 0.9812250511988094,
+      "eval_mean_iou": 0.4906125255994047,
+      "eval_overall_accuracy": 0.9812250511988094,
+      "eval_runtime": 8.2561,
+      "eval_samples_per_second": 30.765,
+      "eval_steps_per_second": 3.876,
+      "step": 4125
+    },
+    {
+      "epoch": 33.6,
+      "grad_norm": 1.3802294731140137,
+      "learning_rate": 4.95e-05,
+      "loss": 0.0836,
+      "step": 4200
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy_no_text": 0.9775148603448678,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9775148603448678,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0939713567495346,
+      "eval_mean_accuracy": 0.9775148603448678,
+      "eval_mean_iou": 0.4887574301724339,
+      "eval_overall_accuracy": 0.9775148603448678,
+      "eval_runtime": 8.2502,
+      "eval_samples_per_second": 30.787,
+      "eval_steps_per_second": 3.879,
+      "step": 4250
+    },
+    {
+      "epoch": 34.4,
+      "grad_norm": 3.26483416557312,
+      "learning_rate": 4.925e-05,
+      "loss": 0.0836,
+      "step": 4300
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy_no_text": 0.9811469413929715,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9811469413929715,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09151175618171692,
+      "eval_mean_accuracy": 0.9811469413929715,
+      "eval_mean_iou": 0.49057347069648577,
+      "eval_overall_accuracy": 0.9811469413929715,
+      "eval_runtime": 8.2647,
+      "eval_samples_per_second": 30.733,
+      "eval_steps_per_second": 3.872,
+      "step": 4375
+    },
+    {
+      "epoch": 35.2,
+      "grad_norm": 1.5455337762832642,
+      "learning_rate": 4.9e-05,
+      "loss": 0.0784,
+      "step": 4400
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 3.040316581726074,
+      "learning_rate": 4.875e-05,
+      "loss": 0.0785,
+      "step": 4500
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy_no_text": 0.9815795942826148,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9815795942826148,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09513239562511444,
+      "eval_mean_accuracy": 0.9815795942826148,
+      "eval_mean_iou": 0.4907897971413074,
+      "eval_overall_accuracy": 0.9815795942826148,
+      "eval_runtime": 8.2374,
+      "eval_samples_per_second": 30.835,
+      "eval_steps_per_second": 3.885,
+      "step": 4500
+    },
+    {
+      "epoch": 36.8,
+      "grad_norm": 2.6754534244537354,
+      "learning_rate": 4.85e-05,
+      "loss": 0.0746,
+      "step": 4600
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy_no_text": 0.9757280549204483,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9757280549204483,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0951041430234909,
+      "eval_mean_accuracy": 0.9757280549204483,
+      "eval_mean_iou": 0.48786402746022417,
+      "eval_overall_accuracy": 0.9757280549204483,
+      "eval_runtime": 8.2295,
+      "eval_samples_per_second": 30.864,
+      "eval_steps_per_second": 3.888,
+      "step": 4625
+    },
+    {
+      "epoch": 37.6,
+      "grad_norm": 0.629497766494751,
+      "learning_rate": 4.825000000000001e-05,
+      "loss": 0.0819,
+      "step": 4700
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy_no_text": 0.9800142910538185,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9800142910538185,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09518762677907944,
+      "eval_mean_accuracy": 0.9800142910538185,
+      "eval_mean_iou": 0.49000714552690927,
+      "eval_overall_accuracy": 0.9800142910538185,
+      "eval_runtime": 8.1766,
+      "eval_samples_per_second": 31.064,
+      "eval_steps_per_second": 3.914,
+      "step": 4750
+    },
+    {
+      "epoch": 38.4,
+      "grad_norm": 0.9247422218322754,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 0.0731,
+      "step": 4800
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy_no_text": 0.9797100599232311,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9797100599232311,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09217899292707443,
+      "eval_mean_accuracy": 0.9797100599232311,
+      "eval_mean_iou": 0.48985502996161556,
+      "eval_overall_accuracy": 0.9797100599232311,
+      "eval_runtime": 8.1339,
+      "eval_samples_per_second": 31.227,
+      "eval_steps_per_second": 3.934,
+      "step": 4875
+    },
+    {
+      "epoch": 39.2,
+      "grad_norm": 0.5285217761993408,
+      "learning_rate": 4.775e-05,
+      "loss": 0.0826,
+      "step": 4900
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 0.7555281519889832,
+      "learning_rate": 4.75e-05,
+      "loss": 0.0745,
+      "step": 5000
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy_no_text": 0.9798135915548363,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9798135915548363,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09387236088514328,
+      "eval_mean_accuracy": 0.9798135915548363,
+      "eval_mean_iou": 0.48990679577741814,
+      "eval_overall_accuracy": 0.9798135915548363,
+      "eval_runtime": 8.2225,
+      "eval_samples_per_second": 30.891,
+      "eval_steps_per_second": 3.892,
+      "step": 5000
+    },
+    {
+      "epoch": 40.8,
+      "grad_norm": 0.6379430294036865,
+      "learning_rate": 4.7249999999999997e-05,
+      "loss": 0.0755,
+      "step": 5100
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy_no_text": 0.980224079270902,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.980224079270902,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.094584159553051,
+      "eval_mean_accuracy": 0.980224079270902,
+      "eval_mean_iou": 0.490112039635451,
+      "eval_overall_accuracy": 0.980224079270902,
+      "eval_runtime": 8.2271,
+      "eval_samples_per_second": 30.874,
+      "eval_steps_per_second": 3.89,
+      "step": 5125
+    },
+    {
+      "epoch": 41.6,
+      "grad_norm": 0.8154006004333496,
+      "learning_rate": 4.7000000000000004e-05,
+      "loss": 0.0692,
+      "step": 5200
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy_no_text": 0.975735365772303,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.975735365772303,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09759368002414703,
+      "eval_mean_accuracy": 0.975735365772303,
+      "eval_mean_iou": 0.4878676828861515,
+      "eval_overall_accuracy": 0.975735365772303,
+      "eval_runtime": 8.4223,
+      "eval_samples_per_second": 30.158,
+      "eval_steps_per_second": 3.799,
+      "step": 5250
+    },
+    {
+      "epoch": 42.4,
+      "grad_norm": 0.7524943351745605,
+      "learning_rate": 4.6750000000000005e-05,
+      "loss": 0.0798,
+      "step": 5300
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy_no_text": 0.9804404638702271,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9804404638702271,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09877783805131912,
+      "eval_mean_accuracy": 0.9804404638702271,
+      "eval_mean_iou": 0.49022023193511355,
+      "eval_overall_accuracy": 0.9804404638702271,
+      "eval_runtime": 8.2237,
+      "eval_samples_per_second": 30.886,
+      "eval_steps_per_second": 3.891,
+      "step": 5375
+    },
+    {
+      "epoch": 43.2,
+      "grad_norm": 0.3496994972229004,
+      "learning_rate": 4.6500000000000005e-05,
+      "loss": 0.072,
+      "step": 5400
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 1.7848222255706787,
+      "learning_rate": 4.625e-05,
+      "loss": 0.076,
+      "step": 5500
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy_no_text": 0.9797547724285511,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9797547724285511,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09654530137777328,
+      "eval_mean_accuracy": 0.9797547724285511,
+      "eval_mean_iou": 0.48987738621427557,
+      "eval_overall_accuracy": 0.9797547724285511,
+      "eval_runtime": 8.1971,
+      "eval_samples_per_second": 30.987,
+      "eval_steps_per_second": 3.904,
+      "step": 5500
+    },
+    {
+      "epoch": 44.8,
+      "grad_norm": 0.5874524712562561,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.0757,
+      "step": 5600
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy_no_text": 0.9823000122706003,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9823000122706003,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09140286594629288,
+      "eval_mean_accuracy": 0.9823000122706003,
+      "eval_mean_iou": 0.49115000613530013,
+      "eval_overall_accuracy": 0.9823000122706003,
+      "eval_runtime": 8.2597,
+      "eval_samples_per_second": 30.752,
+      "eval_steps_per_second": 3.874,
+      "step": 5625
+    },
+    {
+      "epoch": 45.6,
+      "grad_norm": 0.46248531341552734,
+      "learning_rate": 4.575000000000001e-05,
+      "loss": 0.0702,
+      "step": 5700
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy_no_text": 0.9780766660787523,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9780766660787523,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09352780878543854,
+      "eval_mean_accuracy": 0.9780766660787523,
+      "eval_mean_iou": 0.48903833303937616,
+      "eval_overall_accuracy": 0.9780766660787523,
+      "eval_runtime": 8.3183,
+      "eval_samples_per_second": 30.535,
+      "eval_steps_per_second": 3.847,
+      "step": 5750
+    },
+    {
+      "epoch": 46.4,
+      "grad_norm": 1.1901108026504517,
+      "learning_rate": 4.55e-05,
+      "loss": 0.0765,
+      "step": 5800
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy_no_text": 0.9809388645798449,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9809388645798449,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09659279137849808,
+      "eval_mean_accuracy": 0.9809388645798449,
+      "eval_mean_iou": 0.49046943228992246,
+      "eval_overall_accuracy": 0.9809388645798449,
+      "eval_runtime": 8.247,
+      "eval_samples_per_second": 30.799,
+      "eval_steps_per_second": 3.88,
+      "step": 5875
+    },
+    {
+      "epoch": 47.2,
+      "grad_norm": 0.2969698905944824,
+      "learning_rate": 4.525000000000001e-05,
+      "loss": 0.0722,
+      "step": 5900
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 5.192142486572266,
+      "learning_rate": 4.5e-05,
+      "loss": 0.0724,
+      "step": 6000
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy_no_text": 0.9832674872045554,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9832674872045554,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09365521371364594,
+      "eval_mean_accuracy": 0.9832674872045554,
+      "eval_mean_iou": 0.4916337436022777,
+      "eval_overall_accuracy": 0.9832674872045554,
+      "eval_runtime": 8.1901,
+      "eval_samples_per_second": 31.013,
+      "eval_steps_per_second": 3.907,
+      "step": 6000
+    },
+    {
+      "epoch": 48.8,
+      "grad_norm": 0.8341678380966187,
+      "learning_rate": 4.4750000000000004e-05,
+      "loss": 0.0713,
+      "step": 6100
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy_no_text": 0.9761614388952771,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9761614388952771,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10168309509754181,
+      "eval_mean_accuracy": 0.9761614388952771,
+      "eval_mean_iou": 0.48808071944763853,
+      "eval_overall_accuracy": 0.9761614388952771,
+      "eval_runtime": 8.2425,
+      "eval_samples_per_second": 30.816,
+      "eval_steps_per_second": 3.882,
+      "step": 6125
+    },
+    {
+      "epoch": 49.6,
+      "grad_norm": 1.4235390424728394,
+      "learning_rate": 4.4500000000000004e-05,
+      "loss": 0.0677,
+      "step": 6200
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy_no_text": 0.9803571201590842,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9803571201590842,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09322305768728256,
+      "eval_mean_accuracy": 0.9803571201590842,
+      "eval_mean_iou": 0.4901785600795421,
+      "eval_overall_accuracy": 0.9803571201590842,
+      "eval_runtime": 8.1322,
+      "eval_samples_per_second": 31.234,
+      "eval_steps_per_second": 3.935,
+      "step": 6250
+    },
+    {
+      "epoch": 50.4,
+      "grad_norm": 4.009937763214111,
+      "learning_rate": 4.4250000000000005e-05,
+      "loss": 0.0715,
+      "step": 6300
+    },
+    {
+      "epoch": 51.0,
+      "eval_accuracy_no_text": 0.9781054442501438,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9781054442501438,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09748787432909012,
+      "eval_mean_accuracy": 0.9781054442501438,
+      "eval_mean_iou": 0.4890527221250719,
+      "eval_overall_accuracy": 0.9781054442501438,
+      "eval_runtime": 8.3284,
+      "eval_samples_per_second": 30.498,
+      "eval_steps_per_second": 3.842,
+      "step": 6375
+    },
+    {
+      "epoch": 51.2,
+      "grad_norm": 0.7396467328071594,
+      "learning_rate": 4.4e-05,
+      "loss": 0.07,
+      "step": 6400
+    },
+    {
+      "epoch": 52.0,
+      "grad_norm": 0.4652855694293976,
+      "learning_rate": 4.3750000000000006e-05,
+      "loss": 0.0713,
+      "step": 6500
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy_no_text": 0.9833075639651767,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9833075639651767,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09451069682836533,
+      "eval_mean_accuracy": 0.9833075639651767,
+      "eval_mean_iou": 0.49165378198258836,
+      "eval_overall_accuracy": 0.9833075639651767,
+      "eval_runtime": 8.2755,
+      "eval_samples_per_second": 30.693,
+      "eval_steps_per_second": 3.867,
+      "step": 6500
+    },
+    {
+      "epoch": 52.8,
+      "grad_norm": 0.32952389121055603,
+      "learning_rate": 4.35e-05,
+      "loss": 0.0695,
+      "step": 6600
+    },
+    {
+      "epoch": 53.0,
+      "eval_accuracy_no_text": 0.9819285545340948,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9819285545340948,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09514930099248886,
+      "eval_mean_accuracy": 0.9819285545340948,
+      "eval_mean_iou": 0.4909642772670474,
+      "eval_overall_accuracy": 0.9819285545340948,
+      "eval_runtime": 8.5513,
+      "eval_samples_per_second": 29.703,
+      "eval_steps_per_second": 3.742,
+      "step": 6625
+    },
+    {
+      "epoch": 53.6,
+      "grad_norm": 1.479300856590271,
+      "learning_rate": 4.325e-05,
+      "loss": 0.0648,
+      "step": 6700
+    },
+    {
+      "epoch": 54.0,
+      "eval_accuracy_no_text": 0.9824639248922958,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9824639248922958,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09646683931350708,
+      "eval_mean_accuracy": 0.9824639248922958,
+      "eval_mean_iou": 0.4912319624461479,
+      "eval_overall_accuracy": 0.9824639248922958,
+      "eval_runtime": 8.3734,
+      "eval_samples_per_second": 30.334,
+      "eval_steps_per_second": 3.822,
+      "step": 6750
+    },
+    {
+      "epoch": 54.4,
+      "grad_norm": 0.37164923548698425,
+      "learning_rate": 4.3e-05,
+      "loss": 0.0694,
+      "step": 6800
+    },
+    {
+      "epoch": 55.0,
+      "eval_accuracy_no_text": 0.9809235284065225,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9809235284065225,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09463436901569366,
+      "eval_mean_accuracy": 0.9809235284065225,
+      "eval_mean_iou": 0.49046176420326126,
+      "eval_overall_accuracy": 0.9809235284065225,
+      "eval_runtime": 8.3659,
+      "eval_samples_per_second": 30.361,
+      "eval_steps_per_second": 3.825,
+      "step": 6875
+    },
+    {
+      "epoch": 55.2,
+      "grad_norm": 0.3387889862060547,
+      "learning_rate": 4.275e-05,
+      "loss": 0.0658,
+      "step": 6900
+    },
+    {
+      "epoch": 56.0,
+      "grad_norm": 0.8153048157691956,
+      "learning_rate": 4.25e-05,
+      "loss": 0.0665,
+      "step": 7000
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy_no_text": 0.9824437868185507,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9824437868185507,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1007036343216896,
+      "eval_mean_accuracy": 0.9824437868185507,
+      "eval_mean_iou": 0.49122189340927536,
+      "eval_overall_accuracy": 0.9824437868185507,
+      "eval_runtime": 8.3927,
+      "eval_samples_per_second": 30.265,
+      "eval_steps_per_second": 3.813,
+      "step": 7000
+    },
+    {
+      "epoch": 56.8,
+      "grad_norm": 0.8299281001091003,
+      "learning_rate": 4.225e-05,
+      "loss": 0.0635,
+      "step": 7100
+    },
+    {
+      "epoch": 57.0,
+      "eval_accuracy_no_text": 0.9831196252257952,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9831196252257952,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09709285199642181,
+      "eval_mean_accuracy": 0.9831196252257952,
+      "eval_mean_iou": 0.4915598126128976,
+      "eval_overall_accuracy": 0.9831196252257952,
+      "eval_runtime": 8.2632,
+      "eval_samples_per_second": 30.739,
+      "eval_steps_per_second": 3.873,
+      "step": 7125
+    },
+    {
+      "epoch": 57.6,
+      "grad_norm": 0.6725646257400513,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 0.0628,
+      "step": 7200
+    },
+    {
+      "epoch": 58.0,
+      "eval_accuracy_no_text": 0.9785253031490414,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9785253031490414,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10017971694469452,
+      "eval_mean_accuracy": 0.9785253031490414,
+      "eval_mean_iou": 0.4892626515745207,
+      "eval_overall_accuracy": 0.9785253031490414,
+      "eval_runtime": 8.2012,
+      "eval_samples_per_second": 30.971,
+      "eval_steps_per_second": 3.902,
+      "step": 7250
+    },
+    {
+      "epoch": 58.4,
+      "grad_norm": 0.5249119400978088,
+      "learning_rate": 4.175e-05,
+      "loss": 0.0668,
+      "step": 7300
+    },
+    {
+      "epoch": 59.0,
+      "eval_accuracy_no_text": 0.9812511708786174,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9812511708786174,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09598462283611298,
+      "eval_mean_accuracy": 0.9812511708786174,
+      "eval_mean_iou": 0.4906255854393087,
+      "eval_overall_accuracy": 0.9812511708786174,
+      "eval_runtime": 8.1994,
+      "eval_samples_per_second": 30.978,
+      "eval_steps_per_second": 3.903,
+      "step": 7375
+    },
+    {
+      "epoch": 59.2,
+      "grad_norm": 0.7958447933197021,
+      "learning_rate": 4.15e-05,
+      "loss": 0.0688,
+      "step": 7400
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 0.3779708743095398,
+      "learning_rate": 4.125e-05,
+      "loss": 0.0648,
+      "step": 7500
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy_no_text": 0.9796383969594832,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9796383969594832,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0938754603266716,
+      "eval_mean_accuracy": 0.9796383969594832,
+      "eval_mean_iou": 0.4898191984797416,
+      "eval_overall_accuracy": 0.9796383969594832,
+      "eval_runtime": 8.2596,
+      "eval_samples_per_second": 30.752,
+      "eval_steps_per_second": 3.874,
+      "step": 7500
+    },
+    {
+      "epoch": 60.8,
+      "grad_norm": 0.30460646748542786,
+      "learning_rate": 4.1e-05,
+      "loss": 0.064,
+      "step": 7600
+    },
+    {
+      "epoch": 61.0,
+      "eval_accuracy_no_text": 0.9785537157778401,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9785537157778401,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09466074407100677,
+      "eval_mean_accuracy": 0.9785537157778401,
+      "eval_mean_iou": 0.48927685788892006,
+      "eval_overall_accuracy": 0.9785537157778401,
+      "eval_runtime": 8.2318,
+      "eval_samples_per_second": 30.856,
+      "eval_steps_per_second": 3.887,
+      "step": 7625
+    },
+    {
+      "epoch": 61.6,
+      "grad_norm": 1.3763538599014282,
+      "learning_rate": 4.0749999999999994e-05,
+      "loss": 0.0636,
+      "step": 7700
+    },
+    {
+      "epoch": 62.0,
+      "eval_accuracy_no_text": 0.9788354161922574,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9788354161922574,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09854825586080551,
+      "eval_mean_accuracy": 0.9788354161922574,
+      "eval_mean_iou": 0.4894177080961287,
+      "eval_overall_accuracy": 0.9788354161922574,
+      "eval_runtime": 8.1667,
+      "eval_samples_per_second": 31.102,
+      "eval_steps_per_second": 3.918,
+      "step": 7750
+    },
+    {
+      "epoch": 62.4,
+      "grad_norm": 0.5202354788780212,
+      "learning_rate": 4.05e-05,
+      "loss": 0.0653,
+      "step": 7800
+    },
+    {
+      "epoch": 63.0,
+      "eval_accuracy_no_text": 0.9812044645046096,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9812044645046096,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0914328396320343,
+      "eval_mean_accuracy": 0.9812044645046096,
+      "eval_mean_iou": 0.4906022322523048,
+      "eval_overall_accuracy": 0.9812044645046096,
+      "eval_runtime": 8.3443,
+      "eval_samples_per_second": 30.44,
+      "eval_steps_per_second": 3.835,
+      "step": 7875
+    },
+    {
+      "epoch": 63.2,
+      "grad_norm": 0.6701187491416931,
+      "learning_rate": 4.0249999999999996e-05,
+      "loss": 0.061,
+      "step": 7900
+    },
+    {
+      "epoch": 64.0,
+      "grad_norm": 0.389220654964447,
+      "learning_rate": 3.9999999999999996e-05,
+      "loss": 0.0594,
+      "step": 8000
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy_no_text": 0.9782057192295591,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9782057192295591,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09664972871541977,
+      "eval_mean_accuracy": 0.9782057192295591,
+      "eval_mean_iou": 0.48910285961477956,
+      "eval_overall_accuracy": 0.9782057192295591,
+      "eval_runtime": 8.3923,
+      "eval_samples_per_second": 30.266,
+      "eval_steps_per_second": 3.813,
+      "step": 8000
+    },
+    {
+      "epoch": 64.8,
+      "grad_norm": 0.408395379781723,
+      "learning_rate": 3.975e-05,
+      "loss": 0.0608,
+      "step": 8100
+    },
+    {
+      "epoch": 65.0,
+      "eval_accuracy_no_text": 0.9794022730601506,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9794022730601506,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09610570222139359,
+      "eval_mean_accuracy": 0.9794022730601506,
+      "eval_mean_iou": 0.4897011365300753,
+      "eval_overall_accuracy": 0.9794022730601506,
+      "eval_runtime": 8.4338,
+      "eval_samples_per_second": 30.117,
+      "eval_steps_per_second": 3.794,
+      "step": 8125
+    },
+    {
+      "epoch": 65.6,
+      "grad_norm": 0.7247848510742188,
+      "learning_rate": 3.9500000000000005e-05,
+      "loss": 0.0625,
+      "step": 8200
+    },
+    {
+      "epoch": 66.0,
+      "eval_accuracy_no_text": 0.9813712682359023,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9813712682359023,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09536296874284744,
+      "eval_mean_accuracy": 0.9813712682359023,
+      "eval_mean_iou": 0.49068563411795113,
+      "eval_overall_accuracy": 0.9813712682359023,
+      "eval_runtime": 8.3567,
+      "eval_samples_per_second": 30.395,
+      "eval_steps_per_second": 3.829,
+      "step": 8250
+    },
+    {
+      "epoch": 66.4,
+      "grad_norm": 0.7799036502838135,
+      "learning_rate": 3.924999999999999e-05,
+      "loss": 0.0646,
+      "step": 8300
+    },
+    {
+      "epoch": 67.0,
+      "eval_accuracy_no_text": 0.9800713489294296,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9800713489294296,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09814047813415527,
+      "eval_mean_accuracy": 0.9800713489294296,
+      "eval_mean_iou": 0.4900356744647148,
+      "eval_overall_accuracy": 0.9800713489294296,
+      "eval_runtime": 8.317,
+      "eval_samples_per_second": 30.54,
+      "eval_steps_per_second": 3.848,
+      "step": 8375
+    },
+    {
+      "epoch": 67.2,
+      "grad_norm": 0.9421979188919067,
+      "learning_rate": 3.9e-05,
+      "loss": 0.0616,
+      "step": 8400
+    },
+    {
+      "epoch": 68.0,
+      "grad_norm": 0.7759032249450684,
+      "learning_rate": 3.874999999999999e-05,
+      "loss": 0.0634,
+      "step": 8500
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy_no_text": 0.9822642223276573,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9822642223276573,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09961166232824326,
+      "eval_mean_accuracy": 0.9822642223276573,
+      "eval_mean_iou": 0.49113211116382866,
+      "eval_overall_accuracy": 0.9822642223276573,
+      "eval_runtime": 8.4898,
+      "eval_samples_per_second": 29.918,
+      "eval_steps_per_second": 3.769,
+      "step": 8500
+    },
+    {
+      "epoch": 68.8,
+      "grad_norm": 0.9563459753990173,
+      "learning_rate": 3.850000000000001e-05,
+      "loss": 0.0611,
+      "step": 8600
+    },
+    {
+      "epoch": 69.0,
+      "eval_accuracy_no_text": 0.981046533488977,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.981046533488977,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10069240629673004,
+      "eval_mean_accuracy": 0.981046533488977,
+      "eval_mean_iou": 0.4905232667444885,
+      "eval_overall_accuracy": 0.981046533488977,
+      "eval_runtime": 8.3357,
+      "eval_samples_per_second": 30.472,
+      "eval_steps_per_second": 3.839,
+      "step": 8625
+    },
+    {
+      "epoch": 69.6,
+      "grad_norm": 0.3676619231700897,
+      "learning_rate": 3.825e-05,
+      "loss": 0.0599,
+      "step": 8700
+    },
+    {
+      "epoch": 70.0,
+      "eval_accuracy_no_text": 0.9792718242012625,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9792718242012625,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09285294264554977,
+      "eval_mean_accuracy": 0.9792718242012625,
+      "eval_mean_iou": 0.48963591210063123,
+      "eval_overall_accuracy": 0.9792718242012625,
+      "eval_runtime": 8.2408,
+      "eval_samples_per_second": 30.822,
+      "eval_steps_per_second": 3.883,
+      "step": 8750
+    },
+    {
+      "epoch": 70.4,
+      "grad_norm": 0.6337301135063171,
+      "learning_rate": 3.800000000000001e-05,
+      "loss": 0.0583,
+      "step": 8800
+    },
+    {
+      "epoch": 71.0,
+      "eval_accuracy_no_text": 0.9825129740620114,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9825129740620114,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09883657097816467,
+      "eval_mean_accuracy": 0.9825129740620114,
+      "eval_mean_iou": 0.4912564870310057,
+      "eval_overall_accuracy": 0.9825129740620114,
+      "eval_runtime": 8.1683,
+      "eval_samples_per_second": 31.096,
+      "eval_steps_per_second": 3.918,
+      "step": 8875
+    },
+    {
+      "epoch": 71.2,
+      "grad_norm": 1.0939313173294067,
+      "learning_rate": 3.7749999999999996e-05,
+      "loss": 0.0618,
+      "step": 8900
+    },
+    {
+      "epoch": 72.0,
+      "grad_norm": 0.729832112789154,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 0.0596,
+      "step": 9000
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy_no_text": 0.9790404523556354,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9790404523556354,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09549351781606674,
+      "eval_mean_accuracy": 0.9790404523556354,
+      "eval_mean_iou": 0.4895202261778177,
+      "eval_overall_accuracy": 0.9790404523556354,
+      "eval_runtime": 8.3006,
+      "eval_samples_per_second": 30.6,
+      "eval_steps_per_second": 3.855,
+      "step": 9000
+    },
+    {
+      "epoch": 72.8,
+      "grad_norm": 0.31644830107688904,
+      "learning_rate": 3.725e-05,
+      "loss": 0.0598,
+      "step": 9100
+    },
+    {
+      "epoch": 73.0,
+      "eval_accuracy_no_text": 0.9799757761570025,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9799757761570025,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10248930752277374,
+      "eval_mean_accuracy": 0.9799757761570025,
+      "eval_mean_iou": 0.48998788807850124,
+      "eval_overall_accuracy": 0.9799757761570025,
+      "eval_runtime": 8.2995,
+      "eval_samples_per_second": 30.604,
+      "eval_steps_per_second": 3.856,
+      "step": 9125
+    },
+    {
+      "epoch": 73.6,
+      "grad_norm": 0.7597707509994507,
+      "learning_rate": 3.7000000000000005e-05,
+      "loss": 0.0623,
+      "step": 9200
+    },
+    {
+      "epoch": 74.0,
+      "eval_accuracy_no_text": 0.9836081064386921,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9836081064386921,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09970316290855408,
+      "eval_mean_accuracy": 0.9836081064386921,
+      "eval_mean_iou": 0.49180405321934606,
+      "eval_overall_accuracy": 0.9836081064386921,
+      "eval_runtime": 8.2563,
+      "eval_samples_per_second": 30.764,
+      "eval_steps_per_second": 3.876,
+      "step": 9250
+    },
+    {
+      "epoch": 74.4,
+      "grad_norm": 1.544348955154419,
+      "learning_rate": 3.675e-05,
+      "loss": 0.0637,
+      "step": 9300
+    },
+    {
+      "epoch": 75.0,
+      "eval_accuracy_no_text": 0.9782330019994349,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9782330019994349,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09712851047515869,
+      "eval_mean_accuracy": 0.9782330019994349,
+      "eval_mean_iou": 0.48911650099971743,
+      "eval_overall_accuracy": 0.9782330019994349,
+      "eval_runtime": 8.2372,
+      "eval_samples_per_second": 30.836,
+      "eval_steps_per_second": 3.885,
+      "step": 9375
+    },
+    {
+      "epoch": 75.2,
+      "grad_norm": 1.103569507598877,
+      "learning_rate": 3.6500000000000006e-05,
+      "loss": 0.0577,
+      "step": 9400
+    },
+    {
+      "epoch": 76.0,
+      "grad_norm": 3.0431878566741943,
+      "learning_rate": 3.625e-05,
+      "loss": 0.0627,
+      "step": 9500
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy_no_text": 0.9806131328985758,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9806131328985758,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09336179494857788,
+      "eval_mean_accuracy": 0.9806131328985758,
+      "eval_mean_iou": 0.4903065664492879,
+      "eval_overall_accuracy": 0.9806131328985758,
+      "eval_runtime": 8.3291,
+      "eval_samples_per_second": 30.496,
+      "eval_steps_per_second": 3.842,
+      "step": 9500
+    },
+    {
+      "epoch": 76.8,
+      "grad_norm": 0.6158745288848877,
+      "learning_rate": 3.6e-05,
+      "loss": 0.0566,
+      "step": 9600
+    },
+    {
+      "epoch": 77.0,
+      "eval_accuracy_no_text": 0.9830253983147323,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9830253983147323,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1015550047159195,
+      "eval_mean_accuracy": 0.9830253983147323,
+      "eval_mean_iou": 0.4915126991573662,
+      "eval_overall_accuracy": 0.9830253983147323,
+      "eval_runtime": 8.1139,
+      "eval_samples_per_second": 31.304,
+      "eval_steps_per_second": 3.944,
+      "step": 9625
+    },
+    {
+      "epoch": 77.6,
+      "grad_norm": 0.6829688549041748,
+      "learning_rate": 3.575e-05,
+      "loss": 0.0585,
+      "step": 9700
+    },
+    {
+      "epoch": 78.0,
+      "eval_accuracy_no_text": 0.9816801849579057,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9816801849579057,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09153465181589127,
+      "eval_mean_accuracy": 0.9816801849579057,
+      "eval_mean_iou": 0.4908400924789528,
+      "eval_overall_accuracy": 0.9816801849579057,
+      "eval_runtime": 8.1774,
+      "eval_samples_per_second": 31.061,
+      "eval_steps_per_second": 3.913,
+      "step": 9750
+    },
+    {
+      "epoch": 78.4,
+      "grad_norm": 0.5731038451194763,
+      "learning_rate": 3.55e-05,
+      "loss": 0.0574,
+      "step": 9800
+    },
+    {
+      "epoch": 79.0,
+      "eval_accuracy_no_text": 0.981351246471164,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.981351246471164,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09391193091869354,
+      "eval_mean_accuracy": 0.981351246471164,
+      "eval_mean_iou": 0.490675623235582,
+      "eval_overall_accuracy": 0.981351246471164,
+      "eval_runtime": 8.219,
+      "eval_samples_per_second": 30.904,
+      "eval_steps_per_second": 3.893,
+      "step": 9875
+    },
+    {
+      "epoch": 79.2,
+      "grad_norm": 0.2290593832731247,
+      "learning_rate": 3.5249999999999996e-05,
+      "loss": 0.0561,
+      "step": 9900
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 0.826732337474823,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 0.0579,
+      "step": 10000
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy_no_text": 0.9797084980594257,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9797084980594257,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0996052473783493,
+      "eval_mean_accuracy": 0.9797084980594257,
+      "eval_mean_iou": 0.48985424902971286,
+      "eval_overall_accuracy": 0.9797084980594257,
+      "eval_runtime": 8.2761,
+      "eval_samples_per_second": 30.691,
+      "eval_steps_per_second": 3.867,
+      "step": 10000
+    },
+    {
+      "epoch": 80.8,
+      "grad_norm": 0.6924729943275452,
+      "learning_rate": 3.4750000000000004e-05,
+      "loss": 0.0564,
+      "step": 10100
+    },
+    {
+      "epoch": 81.0,
+      "eval_accuracy_no_text": 0.9801340560996555,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9801340560996555,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09884705394506454,
+      "eval_mean_accuracy": 0.9801340560996555,
+      "eval_mean_iou": 0.49006702804982777,
+      "eval_overall_accuracy": 0.9801340560996555,
+      "eval_runtime": 8.235,
+      "eval_samples_per_second": 30.844,
+      "eval_steps_per_second": 3.886,
+      "step": 10125
+    },
+    {
+      "epoch": 81.6,
+      "grad_norm": 0.6362507939338684,
+      "learning_rate": 3.45e-05,
+      "loss": 0.0614,
+      "step": 10200
+    },
+    {
+      "epoch": 82.0,
+      "eval_accuracy_no_text": 0.9835959936864148,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9835959936864148,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1010938361287117,
+      "eval_mean_accuracy": 0.9835959936864148,
+      "eval_mean_iou": 0.4917979968432074,
+      "eval_overall_accuracy": 0.9835959936864148,
+      "eval_runtime": 8.1531,
+      "eval_samples_per_second": 31.154,
+      "eval_steps_per_second": 3.925,
+      "step": 10250
+    },
+    {
+      "epoch": 82.4,
+      "grad_norm": 0.2358619123697281,
+      "learning_rate": 3.4250000000000006e-05,
+      "loss": 0.0556,
+      "step": 10300
+    },
+    {
+      "epoch": 83.0,
+      "eval_accuracy_no_text": 0.9816712457799561,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9816712457799561,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09844139963388443,
+      "eval_mean_accuracy": 0.9816712457799561,
+      "eval_mean_iou": 0.49083562288997806,
+      "eval_overall_accuracy": 0.9816712457799561,
+      "eval_runtime": 8.4411,
+      "eval_samples_per_second": 30.091,
+      "eval_steps_per_second": 3.791,
+      "step": 10375
+    },
+    {
+      "epoch": 83.2,
+      "grad_norm": 0.4608982801437378,
+      "learning_rate": 3.4e-05,
+      "loss": 0.0585,
+      "step": 10400
+    },
+    {
+      "epoch": 84.0,
+      "grad_norm": 0.537953794002533,
+      "learning_rate": 3.3749999999999994e-05,
+      "loss": 0.0582,
+      "step": 10500
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy_no_text": 0.9811251916087039,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9811251916087039,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.0963573008775711,
+      "eval_mean_accuracy": 0.9811251916087039,
+      "eval_mean_iou": 0.49056259580435196,
+      "eval_overall_accuracy": 0.9811251916087039,
+      "eval_runtime": 8.3426,
+      "eval_samples_per_second": 30.446,
+      "eval_steps_per_second": 3.836,
+      "step": 10500
+    },
+    {
+      "epoch": 84.8,
+      "grad_norm": 0.3407799303531647,
+      "learning_rate": 3.35e-05,
+      "loss": 0.057,
+      "step": 10600
+    },
+    {
+      "epoch": 85.0,
+      "eval_accuracy_no_text": 0.9821431280360285,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9821431280360285,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09560161828994751,
+      "eval_mean_accuracy": 0.9821431280360285,
+      "eval_mean_iou": 0.4910715640180143,
+      "eval_overall_accuracy": 0.9821431280360285,
+      "eval_runtime": 8.3368,
+      "eval_samples_per_second": 30.467,
+      "eval_steps_per_second": 3.838,
+      "step": 10625
+    },
+    {
+      "epoch": 85.6,
+      "grad_norm": 0.7103127241134644,
+      "learning_rate": 3.325e-05,
+      "loss": 0.0552,
+      "step": 10700
+    },
+    {
+      "epoch": 86.0,
+      "eval_accuracy_no_text": 0.9803550099813897,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9803550099813897,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10003025084733963,
+      "eval_mean_accuracy": 0.9803550099813897,
+      "eval_mean_iou": 0.49017750499069485,
+      "eval_overall_accuracy": 0.9803550099813897,
+      "eval_runtime": 8.3658,
+      "eval_samples_per_second": 30.362,
+      "eval_steps_per_second": 3.825,
+      "step": 10750
+    },
+    {
+      "epoch": 86.4,
+      "grad_norm": 0.22427453100681305,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 0.059,
+      "step": 10800
+    },
+    {
+      "epoch": 87.0,
+      "eval_accuracy_no_text": 0.9828363297164279,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9828363297164279,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09896524995565414,
+      "eval_mean_accuracy": 0.9828363297164279,
+      "eval_mean_iou": 0.49141816485821394,
+      "eval_overall_accuracy": 0.9828363297164279,
+      "eval_runtime": 8.1925,
+      "eval_samples_per_second": 31.004,
+      "eval_steps_per_second": 3.906,
+      "step": 10875
+    },
+    {
+      "epoch": 87.2,
+      "grad_norm": 0.7796413898468018,
+      "learning_rate": 3.2749999999999996e-05,
+      "loss": 0.0596,
+      "step": 10900
+    },
+    {
+      "epoch": 88.0,
+      "grad_norm": 0.44246774911880493,
+      "learning_rate": 3.25e-05,
+      "loss": 0.0547,
+      "step": 11000
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy_no_text": 0.981057084377449,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.981057084377449,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09589246660470963,
+      "eval_mean_accuracy": 0.981057084377449,
+      "eval_mean_iou": 0.4905285421887245,
+      "eval_overall_accuracy": 0.981057084377449,
+      "eval_runtime": 8.2104,
+      "eval_samples_per_second": 30.936,
+      "eval_steps_per_second": 3.898,
+      "step": 11000
+    },
+    {
+      "epoch": 88.8,
+      "grad_norm": 0.19626279175281525,
+      "learning_rate": 3.225e-05,
+      "loss": 0.0532,
+      "step": 11100
+    },
+    {
+      "epoch": 89.0,
+      "eval_accuracy_no_text": 0.9818968520219614,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9818968520219614,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09797363728284836,
+      "eval_mean_accuracy": 0.9818968520219614,
+      "eval_mean_iou": 0.4909484260109807,
+      "eval_overall_accuracy": 0.9818968520219614,
+      "eval_runtime": 8.1435,
+      "eval_samples_per_second": 31.19,
+      "eval_steps_per_second": 3.929,
+      "step": 11125
+    },
+    {
+      "epoch": 89.6,
+      "grad_norm": 1.401407241821289,
+      "learning_rate": 3.2e-05,
+      "loss": 0.0578,
+      "step": 11200
+    },
+    {
+      "epoch": 90.0,
+      "eval_accuracy_no_text": 0.9829336138928118,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9829336138928118,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09540460258722305,
+      "eval_mean_accuracy": 0.9829336138928118,
+      "eval_mean_iou": 0.4914668069464059,
+      "eval_overall_accuracy": 0.9829336138928118,
+      "eval_runtime": 7.9565,
+      "eval_samples_per_second": 31.924,
+      "eval_steps_per_second": 4.022,
+      "step": 11250
+    },
+    {
+      "epoch": 90.4,
+      "grad_norm": 0.3004220426082611,
+      "learning_rate": 3.1750000000000006e-05,
+      "loss": 0.0552,
+      "step": 11300
+    },
+    {
+      "epoch": 91.0,
+      "eval_accuracy_no_text": 0.9817473783326788,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9817473783326788,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10131113231182098,
+      "eval_mean_accuracy": 0.9817473783326788,
+      "eval_mean_iou": 0.4908736891663394,
+      "eval_overall_accuracy": 0.9817473783326788,
+      "eval_runtime": 7.9524,
+      "eval_samples_per_second": 31.94,
+      "eval_steps_per_second": 4.024,
+      "step": 11375
+    },
+    {
+      "epoch": 91.2,
+      "grad_norm": 0.3821699321269989,
+      "learning_rate": 3.15e-05,
+      "loss": 0.0555,
+      "step": 11400
+    },
+    {
+      "epoch": 92.0,
+      "grad_norm": 0.6809713840484619,
+      "learning_rate": 3.1249999999999994e-05,
+      "loss": 0.0584,
+      "step": 11500
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy_no_text": 0.9802022796399172,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9802022796399172,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09861049801111221,
+      "eval_mean_accuracy": 0.9802022796399172,
+      "eval_mean_iou": 0.4901011398199586,
+      "eval_overall_accuracy": 0.9802022796399172,
+      "eval_runtime": 8.4024,
+      "eval_samples_per_second": 30.23,
+      "eval_steps_per_second": 3.808,
+      "step": 11500
+    },
+    {
+      "epoch": 92.8,
+      "grad_norm": 2.616292953491211,
+      "learning_rate": 3.1e-05,
+      "loss": 0.0528,
+      "step": 11600
+    },
+    {
+      "epoch": 93.0,
+      "eval_accuracy_no_text": 0.9805875116859397,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9805875116859397,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10085444152355194,
+      "eval_mean_accuracy": 0.9805875116859397,
+      "eval_mean_iou": 0.49029375584296986,
+      "eval_overall_accuracy": 0.9805875116859397,
+      "eval_runtime": 8.1295,
+      "eval_samples_per_second": 31.244,
+      "eval_steps_per_second": 3.936,
+      "step": 11625
+    },
+    {
+      "epoch": 93.6,
+      "grad_norm": 0.6591981053352356,
+      "learning_rate": 3.075e-05,
+      "loss": 0.0566,
+      "step": 11700
+    },
+    {
+      "epoch": 94.0,
+      "eval_accuracy_no_text": 0.9802413096194776,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9802413096194776,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.098316490650177,
+      "eval_mean_accuracy": 0.9802413096194776,
+      "eval_mean_iou": 0.4901206548097388,
+      "eval_overall_accuracy": 0.9802413096194776,
+      "eval_runtime": 8.5005,
+      "eval_samples_per_second": 29.881,
+      "eval_steps_per_second": 3.764,
+      "step": 11750
+    },
+    {
+      "epoch": 94.4,
+      "grad_norm": 1.771093487739563,
+      "learning_rate": 3.05e-05,
+      "loss": 0.0541,
+      "step": 11800
+    },
+    {
+      "epoch": 95.0,
+      "eval_accuracy_no_text": 0.9805775257269292,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9805775257269292,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10317344218492508,
+      "eval_mean_accuracy": 0.9805775257269292,
+      "eval_mean_iou": 0.4902887628634646,
+      "eval_overall_accuracy": 0.9805775257269292,
+      "eval_runtime": 8.3723,
+      "eval_samples_per_second": 30.338,
+      "eval_steps_per_second": 3.822,
+      "step": 11875
+    },
+    {
+      "epoch": 95.2,
+      "grad_norm": 1.3780484199523926,
+      "learning_rate": 3.025e-05,
+      "loss": 0.0534,
+      "step": 11900
+    },
+    {
+      "epoch": 96.0,
+      "grad_norm": 0.5944967865943909,
+      "learning_rate": 3e-05,
+      "loss": 0.0577,
+      "step": 12000
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy_no_text": 0.9799943357513698,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9799943357513698,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10296148806810379,
+      "eval_mean_accuracy": 0.9799943357513698,
+      "eval_mean_iou": 0.4899971678756849,
+      "eval_overall_accuracy": 0.9799943357513698,
+      "eval_runtime": 8.2647,
+      "eval_samples_per_second": 30.733,
+      "eval_steps_per_second": 3.872,
+      "step": 12000
+    },
+    {
+      "epoch": 96.8,
+      "grad_norm": 0.368669718503952,
+      "learning_rate": 2.975e-05,
+      "loss": 0.0567,
+      "step": 12100
+    },
+    {
+      "epoch": 97.0,
+      "eval_accuracy_no_text": 0.9795809070789898,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9795809070789898,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10392536222934723,
+      "eval_mean_accuracy": 0.9795809070789898,
+      "eval_mean_iou": 0.4897904535394949,
+      "eval_overall_accuracy": 0.9795809070789898,
+      "eval_runtime": 8.318,
+      "eval_samples_per_second": 30.536,
+      "eval_steps_per_second": 3.847,
+      "step": 12125
+    },
+    {
+      "epoch": 97.6,
+      "grad_norm": 0.6572363376617432,
+      "learning_rate": 2.95e-05,
+      "loss": 0.056,
+      "step": 12200
+    },
+    {
+      "epoch": 98.0,
+      "eval_accuracy_no_text": 0.9788773871281321,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9788773871281321,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10201127082109451,
+      "eval_mean_accuracy": 0.9788773871281321,
+      "eval_mean_iou": 0.48943869356406605,
+      "eval_overall_accuracy": 0.9788773871281321,
+      "eval_runtime": 8.2133,
+      "eval_samples_per_second": 30.925,
+      "eval_steps_per_second": 3.896,
+      "step": 12250
+    },
+    {
+      "epoch": 98.4,
+      "grad_norm": 0.1991199553012848,
+      "learning_rate": 2.9249999999999996e-05,
+      "loss": 0.0517,
+      "step": 12300
+    },
+    {
+      "epoch": 99.0,
+      "eval_accuracy_no_text": 0.981906671825248,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.981906671825248,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10037293285131454,
+      "eval_mean_accuracy": 0.981906671825248,
+      "eval_mean_iou": 0.490953335912624,
+      "eval_overall_accuracy": 0.981906671825248,
+      "eval_runtime": 8.1469,
+      "eval_samples_per_second": 31.178,
+      "eval_steps_per_second": 3.928,
+      "step": 12375
+    },
+    {
+      "epoch": 99.2,
+      "grad_norm": 0.3791349530220032,
+      "learning_rate": 2.9e-05,
+      "loss": 0.0562,
+      "step": 12400
+    },
+    {
+      "epoch": 100.0,
+      "grad_norm": 0.8666055798530579,
+      "learning_rate": 2.875e-05,
+      "loss": 0.051,
+      "step": 12500
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy_no_text": 0.9826272227378127,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9826272227378127,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09898550808429718,
+      "eval_mean_accuracy": 0.9826272227378127,
+      "eval_mean_iou": 0.49131361136890633,
+      "eval_overall_accuracy": 0.9826272227378127,
+      "eval_runtime": 8.2239,
+      "eval_samples_per_second": 30.886,
+      "eval_steps_per_second": 3.891,
+      "step": 12500
+    },
+    {
+      "epoch": 100.8,
+      "grad_norm": 1.307895541191101,
+      "learning_rate": 2.8500000000000005e-05,
+      "loss": 0.0523,
+      "step": 12600
+    },
+    {
+      "epoch": 101.0,
+      "eval_accuracy_no_text": 0.9826056889559862,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9826056889559862,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09835489839315414,
+      "eval_mean_accuracy": 0.9826056889559862,
+      "eval_mean_iou": 0.4913028444779931,
+      "eval_overall_accuracy": 0.9826056889559862,
+      "eval_runtime": 8.309,
+      "eval_samples_per_second": 30.569,
+      "eval_steps_per_second": 3.851,
+      "step": 12625
+    },
+    {
+      "epoch": 101.6,
+      "grad_norm": 0.38712045550346375,
+      "learning_rate": 2.825e-05,
+      "loss": 0.0521,
+      "step": 12700
+    },
+    {
+      "epoch": 102.0,
+      "eval_accuracy_no_text": 0.9799104769574827,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9799104769574827,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.098680280148983,
+      "eval_mean_accuracy": 0.9799104769574827,
+      "eval_mean_iou": 0.48995523847874134,
+      "eval_overall_accuracy": 0.9799104769574827,
+      "eval_runtime": 8.3754,
+      "eval_samples_per_second": 30.327,
+      "eval_steps_per_second": 3.821,
+      "step": 12750
+    },
+    {
+      "epoch": 102.4,
+      "grad_norm": 0.2037675976753235,
+      "learning_rate": 2.7999999999999996e-05,
+      "loss": 0.0518,
+      "step": 12800
+    },
+    {
+      "epoch": 103.0,
+      "eval_accuracy_no_text": 0.9818977160317262,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9818977160317262,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10650473833084106,
+      "eval_mean_accuracy": 0.9818977160317262,
+      "eval_mean_iou": 0.4909488580158631,
+      "eval_overall_accuracy": 0.9818977160317262,
+      "eval_runtime": 8.2263,
+      "eval_samples_per_second": 30.876,
+      "eval_steps_per_second": 3.89,
+      "step": 12875
+    },
+    {
+      "epoch": 103.2,
+      "grad_norm": 0.6227338314056396,
+      "learning_rate": 2.7750000000000004e-05,
+      "loss": 0.0551,
+      "step": 12900
+    },
+    {
+      "epoch": 104.0,
+      "grad_norm": 0.6191376447677612,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 0.0521,
+      "step": 13000
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy_no_text": 0.9808840830376523,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9808840830376523,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10519281029701233,
+      "eval_mean_accuracy": 0.9808840830376523,
+      "eval_mean_iou": 0.49044204151882614,
+      "eval_overall_accuracy": 0.9808840830376523,
+      "eval_runtime": 8.3514,
+      "eval_samples_per_second": 30.414,
+      "eval_steps_per_second": 3.832,
+      "step": 13000
+    },
+    {
+      "epoch": 104.8,
+      "grad_norm": 0.30404117703437805,
+      "learning_rate": 2.725e-05,
+      "loss": 0.0556,
+      "step": 13100
+    },
+    {
+      "epoch": 105.0,
+      "eval_accuracy_no_text": 0.9817842316722551,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9817842316722551,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10063216090202332,
+      "eval_mean_accuracy": 0.9817842316722551,
+      "eval_mean_iou": 0.4908921158361276,
+      "eval_overall_accuracy": 0.9817842316722551,
+      "eval_runtime": 8.2778,
+      "eval_samples_per_second": 30.685,
+      "eval_steps_per_second": 3.866,
+      "step": 13125
+    },
+    {
+      "epoch": 105.6,
+      "grad_norm": 0.7706215977668762,
+      "learning_rate": 2.7e-05,
+      "loss": 0.0544,
+      "step": 13200
+    },
+    {
+      "epoch": 106.0,
+      "eval_accuracy_no_text": 0.9808936702229253,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9808936702229253,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10449624061584473,
+      "eval_mean_accuracy": 0.9808936702229253,
+      "eval_mean_iou": 0.49044683511146264,
+      "eval_overall_accuracy": 0.9808936702229253,
+      "eval_runtime": 8.4473,
+      "eval_samples_per_second": 30.069,
+      "eval_steps_per_second": 3.788,
+      "step": 13250
+    },
+    {
+      "epoch": 106.4,
+      "grad_norm": 0.3159944713115692,
+      "learning_rate": 2.675e-05,
+      "loss": 0.0549,
+      "step": 13300
+    },
+    {
+      "epoch": 107.0,
+      "eval_accuracy_no_text": 0.9823117428647125,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9823117428647125,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.101432666182518,
+      "eval_mean_accuracy": 0.9823117428647125,
+      "eval_mean_iou": 0.49115587143235623,
+      "eval_overall_accuracy": 0.9823117428647125,
+      "eval_runtime": 8.2997,
+      "eval_samples_per_second": 30.604,
+      "eval_steps_per_second": 3.856,
+      "step": 13375
+    },
+    {
+      "epoch": 107.2,
+      "grad_norm": 0.7384445667266846,
+      "learning_rate": 2.6500000000000004e-05,
+      "loss": 0.0506,
+      "step": 13400
+    },
+    {
+      "epoch": 108.0,
+      "grad_norm": 1.998416543006897,
+      "learning_rate": 2.625e-05,
+      "loss": 0.054,
+      "step": 13500
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy_no_text": 0.9808719370542301,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9808719370542301,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10262825340032578,
+      "eval_mean_accuracy": 0.9808719370542301,
+      "eval_mean_iou": 0.4904359685271151,
+      "eval_overall_accuracy": 0.9808719370542301,
+      "eval_runtime": 8.2241,
+      "eval_samples_per_second": 30.885,
+      "eval_steps_per_second": 3.891,
+      "step": 13500
+    },
+    {
+      "epoch": 108.8,
+      "grad_norm": 0.29045435786247253,
+      "learning_rate": 2.5999999999999995e-05,
+      "loss": 0.0526,
+      "step": 13600
+    },
+    {
+      "epoch": 109.0,
+      "eval_accuracy_no_text": 0.9836949560356109,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9836949560356109,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10518956929445267,
+      "eval_mean_accuracy": 0.9836949560356109,
+      "eval_mean_iou": 0.49184747801780543,
+      "eval_overall_accuracy": 0.9836949560356109,
+      "eval_runtime": 8.301,
+      "eval_samples_per_second": 30.599,
+      "eval_steps_per_second": 3.855,
+      "step": 13625
+    },
+    {
+      "epoch": 109.6,
+      "grad_norm": 0.6301660537719727,
+      "learning_rate": 2.575e-05,
+      "loss": 0.0524,
+      "step": 13700
+    },
+    {
+      "epoch": 110.0,
+      "eval_accuracy_no_text": 0.9830132357157378,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9830132357157378,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.09869378805160522,
+      "eval_mean_accuracy": 0.9830132357157378,
+      "eval_mean_iou": 0.4915066178578689,
+      "eval_overall_accuracy": 0.9830132357157378,
+      "eval_runtime": 8.2194,
+      "eval_samples_per_second": 30.903,
+      "eval_steps_per_second": 3.893,
+      "step": 13750
+    },
+    {
+      "epoch": 110.4,
+      "grad_norm": 5.493782043457031,
+      "learning_rate": 2.55e-05,
+      "loss": 0.0487,
+      "step": 13800
+    },
+    {
+      "epoch": 111.0,
+      "eval_accuracy_no_text": 0.9800823152072116,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9800823152072116,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10276078432798386,
+      "eval_mean_accuracy": 0.9800823152072116,
+      "eval_mean_iou": 0.4900411576036058,
+      "eval_overall_accuracy": 0.9800823152072116,
+      "eval_runtime": 11.9277,
+      "eval_samples_per_second": 21.295,
+      "eval_steps_per_second": 2.683,
+      "step": 13875
+    },
+    {
+      "epoch": 111.2,
+      "grad_norm": 0.2871710956096649,
+      "learning_rate": 2.5250000000000004e-05,
+      "loss": 0.0536,
+      "step": 13900
+    },
+    {
+      "epoch": 112.0,
+      "grad_norm": 0.9292570948600769,
+      "learning_rate": 2.5000000000000005e-05,
+      "loss": 0.054,
+      "step": 14000
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy_no_text": 0.9829168155491185,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9829168155491185,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1070096418261528,
+      "eval_mean_accuracy": 0.9829168155491185,
+      "eval_mean_iou": 0.49145840777455924,
+      "eval_overall_accuracy": 0.9829168155491185,
+      "eval_runtime": 8.2923,
+      "eval_samples_per_second": 30.631,
+      "eval_steps_per_second": 3.859,
+      "step": 14000
+    },
+    {
+      "epoch": 112.8,
+      "grad_norm": 0.38912442326545715,
+      "learning_rate": 2.4750000000000002e-05,
+      "loss": 0.0531,
+      "step": 14100
+    },
+    {
+      "epoch": 113.0,
+      "eval_accuracy_no_text": 0.9806236837870479,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9806236837870479,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10462525486946106,
+      "eval_mean_accuracy": 0.9806236837870479,
+      "eval_mean_iou": 0.49031184189352395,
+      "eval_overall_accuracy": 0.9806236837870479,
+      "eval_runtime": 8.2827,
+      "eval_samples_per_second": 30.666,
+      "eval_steps_per_second": 3.863,
+      "step": 14125
+    },
+    {
+      "epoch": 113.6,
+      "grad_norm": 0.5761541128158569,
+      "learning_rate": 2.45e-05,
+      "loss": 0.0478,
+      "step": 14200
+    },
+    {
+      "epoch": 114.0,
+      "eval_accuracy_no_text": 0.9830612214888201,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9830612214888201,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10361193120479584,
+      "eval_mean_accuracy": 0.9830612214888201,
+      "eval_mean_iou": 0.49153061074441007,
+      "eval_overall_accuracy": 0.9830612214888201,
+      "eval_runtime": 8.3565,
+      "eval_samples_per_second": 30.396,
+      "eval_steps_per_second": 3.829,
+      "step": 14250
+    },
+    {
+      "epoch": 114.4,
+      "grad_norm": 0.43338820338249207,
+      "learning_rate": 2.425e-05,
+      "loss": 0.0511,
+      "step": 14300
+    },
+    {
+      "epoch": 115.0,
+      "eval_accuracy_no_text": 0.9807316019297658,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9807316019297658,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10400665551424026,
+      "eval_mean_accuracy": 0.9807316019297658,
+      "eval_mean_iou": 0.4903658009648829,
+      "eval_overall_accuracy": 0.9807316019297658,
+      "eval_runtime": 8.1898,
+      "eval_samples_per_second": 31.014,
+      "eval_steps_per_second": 3.907,
+      "step": 14375
+    },
+    {
+      "epoch": 115.2,
+      "grad_norm": 0.4137003421783447,
+      "learning_rate": 2.4000000000000004e-05,
+      "loss": 0.0516,
+      "step": 14400
+    },
+    {
+      "epoch": 116.0,
+      "grad_norm": 0.34559884667396545,
+      "learning_rate": 2.375e-05,
+      "loss": 0.05,
+      "step": 14500
+    },
+    {
+      "epoch": 116.0,
+      "eval_accuracy_no_text": 0.9825900370867884,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9825900370867884,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10383593291044235,
+      "eval_mean_accuracy": 0.9825900370867884,
+      "eval_mean_iou": 0.4912950185433942,
+      "eval_overall_accuracy": 0.9825900370867884,
+      "eval_runtime": 8.3311,
+      "eval_samples_per_second": 30.488,
+      "eval_steps_per_second": 3.841,
+      "step": 14500
+    },
+    {
+      "epoch": 116.8,
+      "grad_norm": 0.5693605542182922,
+      "learning_rate": 2.3500000000000002e-05,
+      "loss": 0.0522,
+      "step": 14600
+    },
+    {
+      "epoch": 117.0,
+      "eval_accuracy_no_text": 0.9813530907997,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9813530907997,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10510838031768799,
+      "eval_mean_accuracy": 0.9813530907997,
+      "eval_mean_iou": 0.49067654539985,
+      "eval_overall_accuracy": 0.9813530907997,
+      "eval_runtime": 8.3739,
+      "eval_samples_per_second": 30.332,
+      "eval_steps_per_second": 3.821,
+      "step": 14625
+    },
+    {
+      "epoch": 117.6,
+      "grad_norm": 3.1022820472717285,
+      "learning_rate": 2.3250000000000003e-05,
+      "loss": 0.0492,
+      "step": 14700
+    },
+    {
+      "epoch": 118.0,
+      "eval_accuracy_no_text": 0.9816835745346747,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9816835745346747,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10122588276863098,
+      "eval_mean_accuracy": 0.9816835745346747,
+      "eval_mean_iou": 0.49084178726733735,
+      "eval_overall_accuracy": 0.9816835745346747,
+      "eval_runtime": 8.251,
+      "eval_samples_per_second": 30.784,
+      "eval_steps_per_second": 3.878,
+      "step": 14750
+    },
+    {
+      "epoch": 118.4,
+      "grad_norm": 6.673212051391602,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 0.0526,
+      "step": 14800
+    },
+    {
+      "epoch": 119.0,
+      "eval_accuracy_no_text": 0.9810688315871336,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9810688315871336,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10411898046731949,
+      "eval_mean_accuracy": 0.9810688315871336,
+      "eval_mean_iou": 0.4905344157935668,
+      "eval_overall_accuracy": 0.9810688315871336,
+      "eval_runtime": 8.5831,
+      "eval_samples_per_second": 29.593,
+      "eval_steps_per_second": 3.728,
+      "step": 14875
+    },
+    {
+      "epoch": 119.2,
+      "grad_norm": 1.0424220561981201,
+      "learning_rate": 2.275e-05,
+      "loss": 0.0483,
+      "step": 14900
+    },
+    {
+      "epoch": 120.0,
+      "grad_norm": 0.8458845019340515,
+      "learning_rate": 2.25e-05,
+      "loss": 0.0483,
+      "step": 15000
+    },
+    {
+      "epoch": 120.0,
+      "eval_accuracy_no_text": 0.9836277958919825,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9836277958919825,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10483555495738983,
+      "eval_mean_accuracy": 0.9836277958919825,
+      "eval_mean_iou": 0.49181389794599123,
+      "eval_overall_accuracy": 0.9836277958919825,
+      "eval_runtime": 8.2865,
+      "eval_samples_per_second": 30.652,
+      "eval_steps_per_second": 3.862,
+      "step": 15000
+    },
+    {
+      "epoch": 120.8,
+      "grad_norm": 0.9494897127151489,
+      "learning_rate": 2.2250000000000002e-05,
+      "loss": 0.0496,
+      "step": 15100
+    },
+    {
+      "epoch": 121.0,
+      "eval_accuracy_no_text": 0.9807214830461761,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9807214830461761,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10666216909885406,
+      "eval_mean_accuracy": 0.9807214830461761,
+      "eval_mean_iou": 0.49036074152308806,
+      "eval_overall_accuracy": 0.9807214830461761,
+      "eval_runtime": 8.3274,
+      "eval_samples_per_second": 30.502,
+      "eval_steps_per_second": 3.843,
+      "step": 15125
+    },
+    {
+      "epoch": 121.6,
+      "grad_norm": 0.33741095662117004,
+      "learning_rate": 2.2e-05,
+      "loss": 0.0486,
+      "step": 15200
+    },
+    {
+      "epoch": 122.0,
+      "eval_accuracy_no_text": 0.9799435585621249,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9799435585621249,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10895564407110214,
+      "eval_mean_accuracy": 0.9799435585621249,
+      "eval_mean_iou": 0.48997177928106245,
+      "eval_overall_accuracy": 0.9799435585621249,
+      "eval_runtime": 8.4498,
+      "eval_samples_per_second": 30.06,
+      "eval_steps_per_second": 3.787,
+      "step": 15250
+    },
+    {
+      "epoch": 122.4,
+      "grad_norm": 0.38935738801956177,
+      "learning_rate": 2.175e-05,
+      "loss": 0.0539,
+      "step": 15300
+    },
+    {
+      "epoch": 123.0,
+      "eval_accuracy_no_text": 0.9796546303737149,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9796546303737149,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10292521864175797,
+      "eval_mean_accuracy": 0.9796546303737149,
+      "eval_mean_iou": 0.48982731518685746,
+      "eval_overall_accuracy": 0.9796546303737149,
+      "eval_runtime": 8.2096,
+      "eval_samples_per_second": 30.939,
+      "eval_steps_per_second": 3.898,
+      "step": 15375
+    },
+    {
+      "epoch": 123.2,
+      "grad_norm": 0.4297466278076172,
+      "learning_rate": 2.15e-05,
+      "loss": 0.0498,
+      "step": 15400
+    },
+    {
+      "epoch": 124.0,
+      "grad_norm": 0.7218087911605835,
+      "learning_rate": 2.1249999999999998e-05,
+      "loss": 0.0507,
+      "step": 15500
+    },
+    {
+      "epoch": 124.0,
+      "eval_accuracy_no_text": 0.9803565552296226,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9803565552296226,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10427888482809067,
+      "eval_mean_accuracy": 0.9803565552296226,
+      "eval_mean_iou": 0.4901782776148113,
+      "eval_overall_accuracy": 0.9803565552296226,
+      "eval_runtime": 8.4272,
+      "eval_samples_per_second": 30.14,
+      "eval_steps_per_second": 3.797,
+      "step": 15500
+    },
+    {
+      "epoch": 124.8,
+      "grad_norm": 0.3873758316040039,
+      "learning_rate": 2.1e-05,
+      "loss": 0.0482,
+      "step": 15600
+    },
+    {
+      "epoch": 125.0,
+      "eval_accuracy_no_text": 0.9791164519837788,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9791164519837788,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1063632071018219,
+      "eval_mean_accuracy": 0.9791164519837788,
+      "eval_mean_iou": 0.4895582259918894,
+      "eval_overall_accuracy": 0.9791164519837788,
+      "eval_runtime": 8.3041,
+      "eval_samples_per_second": 30.587,
+      "eval_steps_per_second": 3.854,
+      "step": 15625
+    },
+    {
+      "epoch": 125.6,
+      "grad_norm": 0.4336509704589844,
+      "learning_rate": 2.075e-05,
+      "loss": 0.0487,
+      "step": 15700
+    },
+    {
+      "epoch": 126.0,
+      "eval_accuracy_no_text": 0.981322318759621,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.981322318759621,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10696995258331299,
+      "eval_mean_accuracy": 0.981322318759621,
+      "eval_mean_iou": 0.4906611593798105,
+      "eval_overall_accuracy": 0.981322318759621,
+      "eval_runtime": 8.3258,
+      "eval_samples_per_second": 30.508,
+      "eval_steps_per_second": 3.843,
+      "step": 15750
+    },
+    {
+      "epoch": 126.4,
+      "grad_norm": 0.6076303124427795,
+      "learning_rate": 2.0499999999999997e-05,
+      "loss": 0.0492,
+      "step": 15800
+    },
+    {
+      "epoch": 127.0,
+      "eval_accuracy_no_text": 0.9836348408946788,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9836348408946788,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.11010294407606125,
+      "eval_mean_accuracy": 0.9836348408946788,
+      "eval_mean_iou": 0.4918174204473394,
+      "eval_overall_accuracy": 0.9836348408946788,
+      "eval_runtime": 8.4835,
+      "eval_samples_per_second": 29.941,
+      "eval_steps_per_second": 3.772,
+      "step": 15875
+    },
+    {
+      "epoch": 127.2,
+      "grad_norm": 0.2879132032394409,
+      "learning_rate": 2.0249999999999998e-05,
+      "loss": 0.0507,
+      "step": 15900
+    },
+    {
+      "epoch": 128.0,
+      "grad_norm": 0.3296537697315216,
+      "learning_rate": 1.9999999999999998e-05,
+      "loss": 0.0479,
+      "step": 16000
+    },
+    {
+      "epoch": 128.0,
+      "eval_accuracy_no_text": 0.9799868587437913,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9799868587437913,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10447646677494049,
+      "eval_mean_accuracy": 0.9799868587437913,
+      "eval_mean_iou": 0.48999342937189566,
+      "eval_overall_accuracy": 0.9799868587437913,
+      "eval_runtime": 8.3118,
+      "eval_samples_per_second": 30.559,
+      "eval_steps_per_second": 3.85,
+      "step": 16000
+    },
+    {
+      "epoch": 128.8,
+      "grad_norm": 0.43510717153549194,
+      "learning_rate": 1.9749999999999996e-05,
+      "loss": 0.0514,
+      "step": 16100
+    },
+    {
+      "epoch": 129.0,
+      "eval_accuracy_no_text": 0.9819989048676233,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9819989048676233,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10433124750852585,
+      "eval_mean_accuracy": 0.9819989048676233,
+      "eval_mean_iou": 0.49099945243381166,
+      "eval_overall_accuracy": 0.9819989048676233,
+      "eval_runtime": 8.1845,
+      "eval_samples_per_second": 31.034,
+      "eval_steps_per_second": 3.91,
+      "step": 16125
+    },
+    {
+      "epoch": 129.6,
+      "grad_norm": 0.9808353781700134,
+      "learning_rate": 1.9499999999999996e-05,
+      "loss": 0.0505,
+      "step": 16200
+    },
+    {
+      "epoch": 130.0,
+      "eval_accuracy_no_text": 0.9821372295078277,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9821372295078277,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1069905087351799,
+      "eval_mean_accuracy": 0.9821372295078277,
+      "eval_mean_iou": 0.49106861475391383,
+      "eval_overall_accuracy": 0.9821372295078277,
+      "eval_runtime": 8.336,
+      "eval_samples_per_second": 30.47,
+      "eval_steps_per_second": 3.839,
+      "step": 16250
+    },
+    {
+      "epoch": 130.4,
+      "grad_norm": 0.44676822423934937,
+      "learning_rate": 1.9250000000000004e-05,
+      "loss": 0.0491,
+      "step": 16300
+    },
+    {
+      "epoch": 131.0,
+      "eval_accuracy_no_text": 0.9810748464243413,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9810748464243413,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10192025452852249,
+      "eval_mean_accuracy": 0.9810748464243413,
+      "eval_mean_iou": 0.49053742321217064,
+      "eval_overall_accuracy": 0.9810748464243413,
+      "eval_runtime": 8.1762,
+      "eval_samples_per_second": 31.066,
+      "eval_steps_per_second": 3.914,
+      "step": 16375
+    },
+    {
+      "epoch": 131.2,
+      "grad_norm": 0.2674311101436615,
+      "learning_rate": 1.9000000000000008e-05,
+      "loss": 0.0479,
+      "step": 16400
+    },
+    {
+      "epoch": 132.0,
+      "grad_norm": 0.18870976567268372,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 0.0477,
+      "step": 16500
+    },
+    {
+      "epoch": 132.0,
+      "eval_accuracy_no_text": 0.9808076347890541,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9808076347890541,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10085348039865494,
+      "eval_mean_accuracy": 0.9808076347890541,
+      "eval_mean_iou": 0.49040381739452704,
+      "eval_overall_accuracy": 0.9808076347890541,
+      "eval_runtime": 8.2705,
+      "eval_samples_per_second": 30.712,
+      "eval_steps_per_second": 3.869,
+      "step": 16500
+    },
+    {
+      "epoch": 132.8,
+      "grad_norm": 1.6924952268600464,
+      "learning_rate": 1.8500000000000006e-05,
+      "loss": 0.0476,
+      "step": 16600
+    },
+    {
+      "epoch": 133.0,
+      "eval_accuracy_no_text": 0.9818020435658646,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9818020435658646,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10147301107645035,
+      "eval_mean_accuracy": 0.9818020435658646,
+      "eval_mean_iou": 0.4909010217829323,
+      "eval_overall_accuracy": 0.9818020435658646,
+      "eval_runtime": 8.2767,
+      "eval_samples_per_second": 30.688,
+      "eval_steps_per_second": 3.866,
+      "step": 16625
+    },
+    {
+      "epoch": 133.6,
+      "grad_norm": 0.26980453729629517,
+      "learning_rate": 1.8250000000000003e-05,
+      "loss": 0.0462,
+      "step": 16700
+    },
+    {
+      "epoch": 134.0,
+      "eval_accuracy_no_text": 0.9804314416144155,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9804314416144155,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10597039759159088,
+      "eval_mean_accuracy": 0.9804314416144155,
+      "eval_mean_iou": 0.49021572080720777,
+      "eval_overall_accuracy": 0.9804314416144155,
+      "eval_runtime": 8.2302,
+      "eval_samples_per_second": 30.862,
+      "eval_steps_per_second": 3.888,
+      "step": 16750
+    },
+    {
+      "epoch": 134.4,
+      "grad_norm": 0.5653632283210754,
+      "learning_rate": 1.8000000000000004e-05,
+      "loss": 0.0485,
+      "step": 16800
+    },
+    {
+      "epoch": 135.0,
+      "eval_accuracy_no_text": 0.9795444691287234,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9795444691287234,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10182661563158035,
+      "eval_mean_accuracy": 0.9795444691287234,
+      "eval_mean_iou": 0.4897722345643617,
+      "eval_overall_accuracy": 0.9795444691287234,
+      "eval_runtime": 8.1917,
+      "eval_samples_per_second": 31.007,
+      "eval_steps_per_second": 3.906,
+      "step": 16875
+    },
+    {
+      "epoch": 135.2,
+      "grad_norm": 0.603354275226593,
+      "learning_rate": 1.775e-05,
+      "loss": 0.0518,
+      "step": 16900
+    },
+    {
+      "epoch": 136.0,
+      "grad_norm": 0.7962479591369629,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 0.0483,
+      "step": 17000
+    },
+    {
+      "epoch": 136.0,
+      "eval_accuracy_no_text": 0.9795505836593654,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9795505836593654,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10557083040475845,
+      "eval_mean_accuracy": 0.9795505836593654,
+      "eval_mean_iou": 0.4897752918296827,
+      "eval_overall_accuracy": 0.9795505836593654,
+      "eval_runtime": 8.2516,
+      "eval_samples_per_second": 30.782,
+      "eval_steps_per_second": 3.878,
+      "step": 17000
+    },
+    {
+      "epoch": 136.8,
+      "grad_norm": 0.2562846839427948,
+      "learning_rate": 1.725e-05,
+      "loss": 0.0503,
+      "step": 17100
+    },
+    {
+      "epoch": 137.0,
+      "eval_accuracy_no_text": 0.9820201229535742,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9820201229535742,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10440485179424286,
+      "eval_mean_accuracy": 0.9820201229535742,
+      "eval_mean_iou": 0.4910100614767871,
+      "eval_overall_accuracy": 0.9820201229535742,
+      "eval_runtime": 8.1927,
+      "eval_samples_per_second": 31.003,
+      "eval_steps_per_second": 3.906,
+      "step": 17125
+    },
+    {
+      "epoch": 137.6,
+      "grad_norm": 0.45571181178092957,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 0.0514,
+      "step": 17200
+    },
+    {
+      "epoch": 138.0,
+      "eval_accuracy_no_text": 0.9812987744935345,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9812987744935345,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10530504584312439,
+      "eval_mean_accuracy": 0.9812987744935345,
+      "eval_mean_iou": 0.49064938724676727,
+      "eval_overall_accuracy": 0.9812987744935345,
+      "eval_runtime": 8.2797,
+      "eval_samples_per_second": 30.678,
+      "eval_steps_per_second": 3.865,
+      "step": 17250
+    },
+    {
+      "epoch": 138.4,
+      "grad_norm": 0.21188737452030182,
+      "learning_rate": 1.675e-05,
+      "loss": 0.0446,
+      "step": 17300
+    },
+    {
+      "epoch": 139.0,
+      "eval_accuracy_no_text": 0.9807978980636295,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9807978980636295,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10513133555650711,
+      "eval_mean_accuracy": 0.9807978980636295,
+      "eval_mean_iou": 0.49039894903181475,
+      "eval_overall_accuracy": 0.9807978980636295,
+      "eval_runtime": 8.251,
+      "eval_samples_per_second": 30.784,
+      "eval_steps_per_second": 3.878,
+      "step": 17375
+    },
+    {
+      "epoch": 139.2,
+      "grad_norm": 0.8840853571891785,
+      "learning_rate": 1.65e-05,
+      "loss": 0.0504,
+      "step": 17400
+    },
+    {
+      "epoch": 140.0,
+      "grad_norm": 0.943657398223877,
+      "learning_rate": 1.625e-05,
+      "loss": 0.047,
+      "step": 17500
+    },
+    {
+      "epoch": 140.0,
+      "eval_accuracy_no_text": 0.9806808579716656,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9806808579716656,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1071048378944397,
+      "eval_mean_accuracy": 0.9806808579716656,
+      "eval_mean_iou": 0.4903404289858328,
+      "eval_overall_accuracy": 0.9806808579716656,
+      "eval_runtime": 8.5372,
+      "eval_samples_per_second": 29.752,
+      "eval_steps_per_second": 3.748,
+      "step": 17500
+    },
+    {
+      "epoch": 140.8,
+      "grad_norm": 0.7239159941673279,
+      "learning_rate": 1.6e-05,
+      "loss": 0.0467,
+      "step": 17600
+    },
+    {
+      "epoch": 141.0,
+      "eval_accuracy_no_text": 0.9828105755792126,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9828105755792126,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10849156975746155,
+      "eval_mean_accuracy": 0.9828105755792126,
+      "eval_mean_iou": 0.4914052877896063,
+      "eval_overall_accuracy": 0.9828105755792126,
+      "eval_runtime": 8.3714,
+      "eval_samples_per_second": 30.341,
+      "eval_steps_per_second": 3.823,
+      "step": 17625
+    },
+    {
+      "epoch": 141.6,
+      "grad_norm": 0.34657156467437744,
+      "learning_rate": 1.5749999999999997e-05,
+      "loss": 0.0476,
+      "step": 17700
+    },
+    {
+      "epoch": 142.0,
+      "eval_accuracy_no_text": 0.9832416998361954,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9832416998361954,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1076694130897522,
+      "eval_mean_accuracy": 0.9832416998361954,
+      "eval_mean_iou": 0.4916208499180977,
+      "eval_overall_accuracy": 0.9832416998361954,
+      "eval_runtime": 8.3324,
+      "eval_samples_per_second": 30.483,
+      "eval_steps_per_second": 3.84,
+      "step": 17750
+    },
+    {
+      "epoch": 142.4,
+      "grad_norm": 1.23021399974823,
+      "learning_rate": 1.55e-05,
+      "loss": 0.0472,
+      "step": 17800
+    },
+    {
+      "epoch": 143.0,
+      "eval_accuracy_no_text": 0.9818141397025696,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9818141397025696,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.11224336177110672,
+      "eval_mean_accuracy": 0.9818141397025696,
+      "eval_mean_iou": 0.4909070698512848,
+      "eval_overall_accuracy": 0.9818141397025696,
+      "eval_runtime": 8.346,
+      "eval_samples_per_second": 30.434,
+      "eval_steps_per_second": 3.834,
+      "step": 17875
+    },
+    {
+      "epoch": 143.2,
+      "grad_norm": 0.43515679240226746,
+      "learning_rate": 1.525e-05,
+      "loss": 0.0474,
+      "step": 17900
+    },
+    {
+      "epoch": 144.0,
+      "grad_norm": 0.26305532455444336,
+      "learning_rate": 1.5e-05,
+      "loss": 0.0477,
+      "step": 18000
+    },
+    {
+      "epoch": 144.0,
+      "eval_accuracy_no_text": 0.9808217912567363,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9808217912567363,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10432812571525574,
+      "eval_mean_accuracy": 0.9808217912567363,
+      "eval_mean_iou": 0.49041089562836815,
+      "eval_overall_accuracy": 0.9808217912567363,
+      "eval_runtime": 8.2922,
+      "eval_samples_per_second": 30.631,
+      "eval_steps_per_second": 3.859,
+      "step": 18000
+    },
+    {
+      "epoch": 144.8,
+      "grad_norm": 0.24621394276618958,
+      "learning_rate": 1.4749999999999998e-05,
+      "loss": 0.0467,
+      "step": 18100
+    },
+    {
+      "epoch": 145.0,
+      "eval_accuracy_no_text": 0.9796973988570646,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9796973988570646,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10507868975400925,
+      "eval_mean_accuracy": 0.9796973988570646,
+      "eval_mean_iou": 0.4898486994285323,
+      "eval_overall_accuracy": 0.9796973988570646,
+      "eval_runtime": 8.3989,
+      "eval_samples_per_second": 30.242,
+      "eval_steps_per_second": 3.81,
+      "step": 18125
+    },
+    {
+      "epoch": 145.6,
+      "grad_norm": 0.8053872585296631,
+      "learning_rate": 1.45e-05,
+      "loss": 0.0493,
+      "step": 18200
+    },
+    {
+      "epoch": 146.0,
+      "eval_accuracy_no_text": 0.9794659273180032,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9794659273180032,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1048658937215805,
+      "eval_mean_accuracy": 0.9794659273180032,
+      "eval_mean_iou": 0.4897329636590016,
+      "eval_overall_accuracy": 0.9794659273180032,
+      "eval_runtime": 8.2794,
+      "eval_samples_per_second": 30.678,
+      "eval_steps_per_second": 3.865,
+      "step": 18250
+    },
+    {
+      "epoch": 146.4,
+      "grad_norm": 0.3059195876121521,
+      "learning_rate": 1.4249999999999999e-05,
+      "loss": 0.0485,
+      "step": 18300
+    },
+    {
+      "epoch": 147.0,
+      "eval_accuracy_no_text": 0.9809577398700878,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9809577398700878,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10593844205141068,
+      "eval_mean_accuracy": 0.9809577398700878,
+      "eval_mean_iou": 0.4904788699350439,
+      "eval_overall_accuracy": 0.9809577398700878,
+      "eval_runtime": 8.3317,
+      "eval_samples_per_second": 30.486,
+      "eval_steps_per_second": 3.841,
+      "step": 18375
+    },
+    {
+      "epoch": 147.2,
+      "grad_norm": 0.41579416394233704,
+      "learning_rate": 1.3999999999999998e-05,
+      "loss": 0.0469,
+      "step": 18400
+    },
+    {
+      "epoch": 148.0,
+      "grad_norm": 0.320431113243103,
+      "learning_rate": 1.375e-05,
+      "loss": 0.0462,
+      "step": 18500
+    },
+    {
+      "epoch": 148.0,
+      "eval_accuracy_no_text": 0.9786900963960741,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9786900963960741,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10566265136003494,
+      "eval_mean_accuracy": 0.9786900963960741,
+      "eval_mean_iou": 0.48934504819803704,
+      "eval_overall_accuracy": 0.9786900963960741,
+      "eval_runtime": 8.2787,
+      "eval_samples_per_second": 30.681,
+      "eval_steps_per_second": 3.865,
+      "step": 18500
+    },
+    {
+      "epoch": 148.8,
+      "grad_norm": 0.44983476400375366,
+      "learning_rate": 1.35e-05,
+      "loss": 0.0474,
+      "step": 18600
+    },
+    {
+      "epoch": 149.0,
+      "eval_accuracy_no_text": 0.9800260382635032,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9800260382635032,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10372522473335266,
+      "eval_mean_accuracy": 0.9800260382635032,
+      "eval_mean_iou": 0.4900130191317516,
+      "eval_overall_accuracy": 0.9800260382635032,
+      "eval_runtime": 8.1632,
+      "eval_samples_per_second": 31.115,
+      "eval_steps_per_second": 3.92,
+      "step": 18625
+    },
+    {
+      "epoch": 149.6,
+      "grad_norm": 0.1674959510564804,
+      "learning_rate": 1.3249999999999999e-05,
+      "loss": 0.0506,
+      "step": 18700
+    },
+    {
+      "epoch": 150.0,
+      "eval_accuracy_no_text": 0.9814382456082342,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9814382456082342,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10518930852413177,
+      "eval_mean_accuracy": 0.9814382456082342,
+      "eval_mean_iou": 0.4907191228041171,
+      "eval_overall_accuracy": 0.9814382456082342,
+      "eval_runtime": 8.3279,
+      "eval_samples_per_second": 30.5,
+      "eval_steps_per_second": 3.843,
+      "step": 18750
+    },
+    {
+      "epoch": 150.4,
+      "grad_norm": 0.31716248393058777,
+      "learning_rate": 1.3000000000000004e-05,
+      "loss": 0.0479,
+      "step": 18800
+    },
+    {
+      "epoch": 151.0,
+      "eval_accuracy_no_text": 0.9805025396487019,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9805025396487019,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10686381161212921,
+      "eval_mean_accuracy": 0.9805025396487019,
+      "eval_mean_iou": 0.49025126982435097,
+      "eval_overall_accuracy": 0.9805025396487019,
+      "eval_runtime": 8.3922,
+      "eval_samples_per_second": 30.266,
+      "eval_steps_per_second": 3.813,
+      "step": 18875
+    },
+    {
+      "epoch": 151.2,
+      "grad_norm": 0.9992444515228271,
+      "learning_rate": 1.2750000000000003e-05,
+      "loss": 0.0464,
+      "step": 18900
+    },
+    {
+      "epoch": 152.0,
+      "grad_norm": 0.19730743765830994,
+      "learning_rate": 1.2500000000000002e-05,
+      "loss": 0.0439,
+      "step": 19000
+    },
+    {
+      "epoch": 152.0,
+      "eval_accuracy_no_text": 0.98159488060922,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.98159488060922,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1079552099108696,
+      "eval_mean_accuracy": 0.98159488060922,
+      "eval_mean_iou": 0.49079744030461,
+      "eval_overall_accuracy": 0.98159488060922,
+      "eval_runtime": 8.3262,
+      "eval_samples_per_second": 30.506,
+      "eval_steps_per_second": 3.843,
+      "step": 19000
+    },
+    {
+      "epoch": 152.8,
+      "grad_norm": 0.3917344808578491,
+      "learning_rate": 1.2250000000000003e-05,
+      "loss": 0.0492,
+      "step": 19100
+    },
+    {
+      "epoch": 153.0,
+      "eval_accuracy_no_text": 0.9807935613992339,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9807935613992339,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10192196071147919,
+      "eval_mean_accuracy": 0.9807935613992339,
+      "eval_mean_iou": 0.49039678069961695,
+      "eval_overall_accuracy": 0.9807935613992339,
+      "eval_runtime": 8.236,
+      "eval_samples_per_second": 30.84,
+      "eval_steps_per_second": 3.885,
+      "step": 19125
+    },
+    {
+      "epoch": 153.6,
+      "grad_norm": 0.3044929504394531,
+      "learning_rate": 1.2000000000000002e-05,
+      "loss": 0.0442,
+      "step": 19200
+    },
+    {
+      "epoch": 154.0,
+      "eval_accuracy_no_text": 0.9820950259539395,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9820950259539395,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.1053338274359703,
+      "eval_mean_accuracy": 0.9820950259539395,
+      "eval_mean_iou": 0.49104751297696975,
+      "eval_overall_accuracy": 0.9820950259539395,
+      "eval_runtime": 8.244,
+      "eval_samples_per_second": 30.81,
+      "eval_steps_per_second": 3.882,
+      "step": 19250
+    },
+    {
+      "epoch": 154.4,
+      "grad_norm": 0.5873416662216187,
+      "learning_rate": 1.1750000000000003e-05,
+      "loss": 0.0484,
+      "step": 19300
+    },
+    {
+      "epoch": 155.0,
+      "eval_accuracy_no_text": 0.9818892088586589,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9818892088586589,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10322853177785873,
+      "eval_mean_accuracy": 0.9818892088586589,
+      "eval_mean_iou": 0.49094460442932947,
+      "eval_overall_accuracy": 0.9818892088586589,
+      "eval_runtime": 8.2069,
+      "eval_samples_per_second": 30.95,
+      "eval_steps_per_second": 3.899,
+      "step": 19375
+    },
+    {
+      "epoch": 155.2,
+      "grad_norm": 1.538596272468567,
+      "learning_rate": 1.1500000000000002e-05,
+      "loss": 0.0456,
+      "step": 19400
+    },
+    {
+      "epoch": 156.0,
+      "grad_norm": 1.1352208852767944,
+      "learning_rate": 1.1250000000000002e-05,
+      "loss": 0.0466,
+      "step": 19500
+    },
+    {
+      "epoch": 156.0,
+      "eval_accuracy_no_text": 0.9812151483176609,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9812151483176609,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10393723100423813,
+      "eval_mean_accuracy": 0.9812151483176609,
+      "eval_mean_iou": 0.49060757415883044,
+      "eval_overall_accuracy": 0.9812151483176609,
+      "eval_runtime": 8.309,
+      "eval_samples_per_second": 30.569,
+      "eval_steps_per_second": 3.851,
+      "step": 19500
+    },
+    {
+      "epoch": 156.8,
+      "grad_norm": 0.30518653988838196,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.0444,
+      "step": 19600
+    },
+    {
+      "epoch": 157.0,
+      "eval_accuracy_no_text": 0.9808679825479997,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9808679825479997,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.103802390396595,
+      "eval_mean_accuracy": 0.9808679825479997,
+      "eval_mean_iou": 0.49043399127399984,
+      "eval_overall_accuracy": 0.9808679825479997,
+      "eval_runtime": 8.2717,
+      "eval_samples_per_second": 30.707,
+      "eval_steps_per_second": 3.869,
+      "step": 19625
+    },
+    {
+      "epoch": 157.6,
+      "grad_norm": 0.371191143989563,
+      "learning_rate": 1.075e-05,
+      "loss": 0.0463,
+      "step": 19700
+    },
+    {
+      "epoch": 158.0,
+      "eval_accuracy_no_text": 0.9814478494090796,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9814478494090796,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10381000488996506,
+      "eval_mean_accuracy": 0.9814478494090796,
+      "eval_mean_iou": 0.4907239247045398,
+      "eval_overall_accuracy": 0.9814478494090796,
+      "eval_runtime": 8.5122,
+      "eval_samples_per_second": 29.839,
+      "eval_steps_per_second": 3.759,
+      "step": 19750
+    },
+    {
+      "epoch": 158.4,
+      "grad_norm": 0.6957194209098816,
+      "learning_rate": 1.0500000000000001e-05,
+      "loss": 0.0465,
+      "step": 19800
+    },
+    {
+      "epoch": 159.0,
+      "eval_accuracy_no_text": 0.9814518703775997,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9814518703775997,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10540702193975449,
+      "eval_mean_accuracy": 0.9814518703775997,
+      "eval_mean_iou": 0.49072593518879987,
+      "eval_overall_accuracy": 0.9814518703775997,
+      "eval_runtime": 8.1775,
+      "eval_samples_per_second": 31.061,
+      "eval_steps_per_second": 3.913,
+      "step": 19875
+    },
+    {
+      "epoch": 159.2,
+      "grad_norm": 0.21183647215366364,
+      "learning_rate": 1.025e-05,
+      "loss": 0.0464,
+      "step": 19900
+    },
+    {
+      "epoch": 160.0,
+      "grad_norm": 0.3967491090297699,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 0.046,
+      "step": 20000
+    },
+    {
+      "epoch": 160.0,
+      "eval_accuracy_no_text": 0.9804176340737537,
+      "eval_accuracy_text": NaN,
+      "eval_iou_no_text": 0.9804176340737537,
+      "eval_iou_text": 0.0,
+      "eval_loss": 0.10416771471500397,
+      "eval_mean_accuracy": 0.9804176340737537,
+      "eval_mean_iou": 0.49020881703687685,
+      "eval_overall_accuracy": 0.9804176340737537,
+      "eval_runtime": 8.2603,
+      "eval_samples_per_second": 30.75,
+      "eval_steps_per_second": 3.874,
+      "step": 20000
+    },
+    {
+      "epoch": 160.0,
+      "step": 20000,
+      "total_flos": 2.80447288344576e+18,
+      "train_loss": 0.01519945946931839,
+      "train_runtime": 1952.9271,
+      "train_samples_per_second": 81.928,
+      "train_steps_per_second": 10.241
     }
   ],
   "logging_steps": 100,
-  "max_steps": 2000,
+  "max_steps": 20000,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 16,
+  "num_train_epochs": 160,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -415,7 +3835,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 2.80447288344576e+17,
+  "total_flos": 2.80447288344576e+18,
   "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null