End of training

README.md

@@ -3,11 +3,24 @@ license: apache-2.0
 base_model: openai/whisper-large-v3
 tags:
 - generated_from_trainer
+datasets:
+- DewiBrynJones/commonvoice_18_0_cy
 metrics:
 - wer
 model-index:
 - name: whisper-large-v3-ft-cv-cy
-  results: []
+  results:
+  - task:
+      name: Automatic Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: DewiBrynJones/commonvoice_18_0_cy default
+      type: DewiBrynJones/commonvoice_18_0_cy
+      args: default
+    metrics:
+    - name: Wer
+      type: wer
+      value: 0.2026720744363593
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -15,7 +28,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # whisper-large-v3-ft-cv-cy
 
-This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on an unknown dataset.
+This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the DewiBrynJones/commonvoice_18_0_cy default dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.4426
 - Wer: 0.2027

all_results.json

@@ -1,15 +1,15 @@
 {
     "epoch": 20.080321285140563,
-    "eval_loss": 0.41853219270706177,
-    "eval_runtime": 2160.3731,
+    "eval_loss": 0.44258150458335876,
+    "eval_runtime": 2083.698,
     "eval_samples": 5381,
-    "eval_samples_per_second": 2.491,
-    "eval_steps_per_second": 0.156,
-    "eval_wer": 0.19108115630840192,
+    "eval_samples_per_second": 2.582,
+    "eval_steps_per_second": 0.162,
+    "eval_wer": 0.2026720744363593,
     "total_flos": 5.435997290496e+20,
-    "train_loss": 0.05159526972509921,
-    "train_runtime": 59448.8268,
+    "train_loss": 0.0536484617867507,
+    "train_runtime": 58113.0051,
     "train_samples": 7968,
-    "train_samples_per_second": 2.691,
-    "train_steps_per_second": 0.084
+    "train_samples_per_second": 2.753,
+    "train_steps_per_second": 0.086
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 20.080321285140563,
-    "eval_loss": 0.41853219270706177,
-    "eval_runtime": 2160.3731,
+    "eval_loss": 0.44258150458335876,
+    "eval_runtime": 2083.698,
     "eval_samples": 5381,
-    "eval_samples_per_second": 2.491,
-    "eval_steps_per_second": 0.156,
-    "eval_wer": 0.19108115630840192
+    "eval_samples_per_second": 2.582,
+    "eval_steps_per_second": 0.162,
+    "eval_wer": 0.2026720744363593
 }

runs/Aug20_09-25-54_5e64a28d78ff/events.out.tfevents.1724202810.5e64a28d78ff.595.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b048ae3df49efa8b3787cc657e14cab84bc592595315e9ac4c2a67030608a22
+size 406

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 20.080321285140563,
     "total_flos": 5.435997290496e+20,
-    "train_loss": 0.05159526972509921,
-    "train_runtime": 59448.8268,
+    "train_loss": 0.0536484617867507,
+    "train_runtime": 58113.0051,
     "train_samples": 7968,
-    "train_samples_per_second": 2.691,
-    "train_steps_per_second": 0.084
+    "train_samples_per_second": 2.753,
+    "train_steps_per_second": 0.086
 }

trainer_state.json

@@ -10,1457 +10,1457 @@
   "log_history": [
     {
       "epoch": 0.10040160642570281,
-      "grad_norm": 5.931593418121338,
+      "grad_norm": 5.799849987030029,
       "learning_rate": 5.000000000000001e-07,
-      "loss": 1.0153,
+      "loss": 1.0201,
       "step": 25
     },
     {
       "epoch": 0.20080321285140562,
-      "grad_norm": 4.555520534515381,
+      "grad_norm": 5.0020341873168945,
       "learning_rate": 1.0000000000000002e-06,
-      "loss": 0.7001,
+      "loss": 0.7191,
       "step": 50
     },
     {
       "epoch": 0.30120481927710846,
-      "grad_norm": 3.582505941390991,
+      "grad_norm": 3.9180057048797607,
       "learning_rate": 1.5e-06,
-      "loss": 0.4808,
+      "loss": 0.4981,
       "step": 75
     },
     {
       "epoch": 0.40160642570281124,
-      "grad_norm": 4.382739067077637,
+      "grad_norm": 4.2526445388793945,
       "learning_rate": 2.0000000000000003e-06,
-      "loss": 0.4384,
+      "loss": 0.4558,
       "step": 100
     },
     {
       "epoch": 0.5020080321285141,
-      "grad_norm": 4.169747352600098,
+      "grad_norm": 4.253071308135986,
       "learning_rate": 2.5e-06,
-      "loss": 0.4109,
+      "loss": 0.4253,
       "step": 125
     },
     {
       "epoch": 0.6024096385542169,
-      "grad_norm": 3.8426809310913086,
+      "grad_norm": 3.9886934757232666,
       "learning_rate": 3e-06,
-      "loss": 0.3934,
+      "loss": 0.4031,
       "step": 150
     },
     {
       "epoch": 0.7028112449799196,
-      "grad_norm": 3.972766399383545,
+      "grad_norm": 4.073461055755615,
       "learning_rate": 3.5e-06,
-      "loss": 0.3788,
+      "loss": 0.3945,
       "step": 175
     },
     {
       "epoch": 0.8032128514056225,
-      "grad_norm": 3.828845739364624,
+      "grad_norm": 3.7386837005615234,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 0.372,
+      "loss": 0.3886,
       "step": 200
     },
     {
       "epoch": 0.9036144578313253,
-      "grad_norm": 4.017838954925537,
+      "grad_norm": 4.179929256439209,
       "learning_rate": 4.5e-06,
-      "loss": 0.362,
+      "loss": 0.3816,
       "step": 225
     },
     {
       "epoch": 1.0040160642570282,
-      "grad_norm": 2.762294054031372,
+      "grad_norm": 2.9208106994628906,
       "learning_rate": 5e-06,
-      "loss": 0.3486,
+      "loss": 0.3655,
       "step": 250
     },
     {
       "epoch": 1.104417670682731,
-      "grad_norm": 3.553866386413574,
+      "grad_norm": 3.4144461154937744,
       "learning_rate": 5.500000000000001e-06,
-      "loss": 0.2512,
+      "loss": 0.2631,
       "step": 275
     },
     {
       "epoch": 1.2048192771084336,
-      "grad_norm": 3.850424289703369,
+      "grad_norm": 4.025992393493652,
       "learning_rate": 6e-06,
-      "loss": 0.2581,
+      "loss": 0.268,
       "step": 300
     },
     {
       "epoch": 1.3052208835341366,
-      "grad_norm": 3.749298572540283,
+      "grad_norm": 3.375342845916748,
       "learning_rate": 6.5000000000000004e-06,
-      "loss": 0.2579,
+      "loss": 0.2701,
       "step": 325
     },
     {
       "epoch": 1.4056224899598393,
-      "grad_norm": 3.389807939529419,
+      "grad_norm": 3.536499500274658,
       "learning_rate": 7e-06,
-      "loss": 0.2721,
+      "loss": 0.2834,
       "step": 350
     },
     {
       "epoch": 1.5060240963855422,
-      "grad_norm": 3.3439886569976807,
+      "grad_norm": 3.2381362915039062,
       "learning_rate": 7.500000000000001e-06,
-      "loss": 0.2516,
+      "loss": 0.2627,
       "step": 375
     },
     {
       "epoch": 1.606425702811245,
-      "grad_norm": 3.6428940296173096,
+      "grad_norm": 4.001467704772949,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.2577,
+      "loss": 0.2694,
       "step": 400
     },
     {
       "epoch": 1.7068273092369477,
-      "grad_norm": 3.5750339031219482,
+      "grad_norm": 3.698154926300049,
       "learning_rate": 8.5e-06,
-      "loss": 0.2484,
+      "loss": 0.261,
       "step": 425
     },
     {
       "epoch": 1.8072289156626506,
-      "grad_norm": 3.3576269149780273,
+      "grad_norm": 3.6060898303985596,
       "learning_rate": 9e-06,
-      "loss": 0.2418,
+      "loss": 0.2494,
       "step": 450
     },
     {
       "epoch": 1.9076305220883534,
-      "grad_norm": 3.252156972885132,
+      "grad_norm": 3.474760055541992,
       "learning_rate": 9.5e-06,
-      "loss": 0.247,
+      "loss": 0.2569,
       "step": 475
     },
     {
       "epoch": 2.0080321285140563,
-      "grad_norm": 2.6496663093566895,
+      "grad_norm": 2.3090035915374756,
       "learning_rate": 1e-05,
-      "loss": 0.2256,
+      "loss": 0.2416,
       "step": 500
     },
     {
       "epoch": 2.108433734939759,
-      "grad_norm": 3.2420847415924072,
+      "grad_norm": 2.8201253414154053,
       "learning_rate": 9.944444444444445e-06,
-      "loss": 0.1409,
+      "loss": 0.1463,
       "step": 525
     },
     {
       "epoch": 2.208835341365462,
-      "grad_norm": 2.1582603454589844,
+      "grad_norm": 2.083610773086548,
       "learning_rate": 9.88888888888889e-06,
-      "loss": 0.1367,
+      "loss": 0.1438,
       "step": 550
     },
     {
       "epoch": 2.3092369477911645,
-      "grad_norm": 2.708833932876587,
+      "grad_norm": 2.2323787212371826,
       "learning_rate": 9.833333333333333e-06,
-      "loss": 0.1284,
+      "loss": 0.1362,
       "step": 575
     },
     {
       "epoch": 2.4096385542168672,
-      "grad_norm": 2.957946300506592,
+      "grad_norm": 2.5994038581848145,
       "learning_rate": 9.777777777777779e-06,
-      "loss": 0.1352,
+      "loss": 0.1448,
       "step": 600
     },
     {
       "epoch": 2.5100401606425704,
-      "grad_norm": 2.999298334121704,
+      "grad_norm": 3.3458213806152344,
       "learning_rate": 9.722222222222223e-06,
-      "loss": 0.1346,
+      "loss": 0.1432,
       "step": 625
     },
     {
       "epoch": 2.610441767068273,
-      "grad_norm": 2.4836103916168213,
+      "grad_norm": 2.83992075920105,
       "learning_rate": 9.666666666666667e-06,
-      "loss": 0.1257,
+      "loss": 0.1357,
       "step": 650
     },
     {
       "epoch": 2.710843373493976,
-      "grad_norm": 2.5539424419403076,
+      "grad_norm": 2.6329407691955566,
       "learning_rate": 9.611111111111112e-06,
-      "loss": 0.1359,
+      "loss": 0.1432,
       "step": 675
     },
     {
       "epoch": 2.8112449799196786,
-      "grad_norm": 2.5347061157226562,
+      "grad_norm": 2.340303659439087,
       "learning_rate": 9.555555555555556e-06,
-      "loss": 0.1363,
+      "loss": 0.1458,
       "step": 700
     },
     {
       "epoch": 2.9116465863453813,
-      "grad_norm": 2.7126588821411133,
+      "grad_norm": 2.580652952194214,
       "learning_rate": 9.5e-06,
-      "loss": 0.137,
+      "loss": 0.1422,
       "step": 725
     },
     {
       "epoch": 3.0120481927710845,
-      "grad_norm": 1.696603775024414,
+      "grad_norm": 1.4325968027114868,
       "learning_rate": 9.444444444444445e-06,
-      "loss": 0.1269,
+      "loss": 0.1279,
       "step": 750
     },
     {
       "epoch": 3.112449799196787,
-      "grad_norm": 2.1536343097686768,
+      "grad_norm": 1.9051955938339233,
       "learning_rate": 9.38888888888889e-06,
-      "loss": 0.0626,
+      "loss": 0.0634,
       "step": 775
     },
     {
       "epoch": 3.21285140562249,
-      "grad_norm": 2.163684844970703,
+      "grad_norm": 2.539659023284912,
       "learning_rate": 9.333333333333334e-06,
-      "loss": 0.0625,
+      "loss": 0.066,
       "step": 800
     },
     {
       "epoch": 3.3132530120481927,
-      "grad_norm": 1.543696641921997,
+      "grad_norm": 1.864598274230957,
       "learning_rate": 9.277777777777778e-06,
-      "loss": 0.0633,
+      "loss": 0.0651,
       "step": 825
     },
     {
       "epoch": 3.4136546184738954,
-      "grad_norm": 2.221062660217285,
+      "grad_norm": 2.48823618888855,
       "learning_rate": 9.222222222222224e-06,
-      "loss": 0.0619,
+      "loss": 0.0641,
       "step": 850
     },
     {
       "epoch": 3.5140562248995986,
-      "grad_norm": 2.847231149673462,
+      "grad_norm": 1.932178020477295,
       "learning_rate": 9.166666666666666e-06,
-      "loss": 0.065,
+      "loss": 0.0729,
       "step": 875
     },
     {
       "epoch": 3.6144578313253013,
-      "grad_norm": 2.0118203163146973,
+      "grad_norm": 1.8803843259811401,
       "learning_rate": 9.111111111111112e-06,
-      "loss": 0.0671,
+      "loss": 0.072,
       "step": 900
     },
     {
       "epoch": 3.714859437751004,
-      "grad_norm": 1.9212677478790283,
+      "grad_norm": 1.7528445720672607,
       "learning_rate": 9.055555555555556e-06,
-      "loss": 0.0656,
+      "loss": 0.0719,
       "step": 925
     },
     {
       "epoch": 3.8152610441767068,
-      "grad_norm": 2.1838502883911133,
+      "grad_norm": 1.9137065410614014,
       "learning_rate": 9e-06,
-      "loss": 0.0693,
+      "loss": 0.0748,
       "step": 950
     },
     {
       "epoch": 3.9156626506024095,
-      "grad_norm": 2.106473445892334,
+      "grad_norm": 2.316129446029663,
       "learning_rate": 8.944444444444446e-06,
-      "loss": 0.0658,
+      "loss": 0.0711,
       "step": 975
     },
     {
       "epoch": 4.016064257028113,
-      "grad_norm": 1.0204880237579346,
+      "grad_norm": 1.2647976875305176,
       "learning_rate": 8.888888888888888e-06,
-      "loss": 0.0619,
+      "loss": 0.0611,
       "step": 1000
     },
     {
       "epoch": 4.016064257028113,
-      "eval_loss": 0.3070617914199829,
-      "eval_runtime": 2205.2781,
-      "eval_samples_per_second": 2.44,
-      "eval_steps_per_second": 0.153,
-      "eval_wer": 0.21380571792118971,
+      "eval_loss": 0.32311680912971497,
+      "eval_runtime": 2118.544,
+      "eval_samples_per_second": 2.54,
+      "eval_steps_per_second": 0.159,
+      "eval_wer": 0.22724561612787786,
       "step": 1000
     },
     {
       "epoch": 4.116465863453815,
-      "grad_norm": 1.0052565336227417,
+      "grad_norm": 1.5836769342422485,
       "learning_rate": 8.833333333333334e-06,
-      "loss": 0.0339,
+      "loss": 0.0356,
       "step": 1025
     },
     {
       "epoch": 4.216867469879518,
-      "grad_norm": 1.4552688598632812,
+      "grad_norm": 1.5816694498062134,
       "learning_rate": 8.777777777777778e-06,
-      "loss": 0.033,
+      "loss": 0.0346,
       "step": 1050
     },
     {
       "epoch": 4.317269076305221,
-      "grad_norm": 1.3099792003631592,
+      "grad_norm": 1.5410749912261963,
       "learning_rate": 8.722222222222224e-06,
-      "loss": 0.0346,
+      "loss": 0.0349,
       "step": 1075
     },
     {
       "epoch": 4.417670682730924,
-      "grad_norm": 1.9144798517227173,
+      "grad_norm": 0.6259779334068298,
       "learning_rate": 8.666666666666668e-06,
-      "loss": 0.0343,
+      "loss": 0.0333,
       "step": 1100
     },
     {
       "epoch": 4.518072289156627,
-      "grad_norm": 1.7304869890213013,
+      "grad_norm": 1.7501243352890015,
       "learning_rate": 8.611111111111112e-06,
-      "loss": 0.0353,
+      "loss": 0.0321,
       "step": 1125
     },
     {
       "epoch": 4.618473895582329,
-      "grad_norm": 1.9012507200241089,
+      "grad_norm": 1.2653108835220337,
       "learning_rate": 8.555555555555556e-06,
-      "loss": 0.0338,
+      "loss": 0.0339,
       "step": 1150
     },
     {
       "epoch": 4.718875502008032,
-      "grad_norm": 1.9311727285385132,
+      "grad_norm": 1.5895155668258667,
       "learning_rate": 8.5e-06,
-      "loss": 0.0377,
+      "loss": 0.0387,
       "step": 1175
     },
     {
       "epoch": 4.8192771084337345,
-      "grad_norm": 2.47721266746521,
+      "grad_norm": 2.180783271789551,
       "learning_rate": 8.444444444444446e-06,
-      "loss": 0.0376,
+      "loss": 0.0394,
       "step": 1200
     },
     {
       "epoch": 4.919678714859438,
-      "grad_norm": 2.12300181388855,
+      "grad_norm": 1.5165903568267822,
       "learning_rate": 8.38888888888889e-06,
-      "loss": 0.0407,
+      "loss": 0.0425,
       "step": 1225
     },
     {
       "epoch": 5.020080321285141,
-      "grad_norm": 1.4431580305099487,
+      "grad_norm": 1.0201085805892944,
       "learning_rate": 8.333333333333334e-06,
-      "loss": 0.0345,
+      "loss": 0.0342,
       "step": 1250
     },
     {
       "epoch": 5.120481927710843,
-      "grad_norm": 1.7874999046325684,
+      "grad_norm": 1.3962773084640503,
       "learning_rate": 8.277777777777778e-06,
-      "loss": 0.0204,
+      "loss": 0.02,
       "step": 1275
     },
     {
       "epoch": 5.220883534136546,
-      "grad_norm": 0.6316199898719788,
+      "grad_norm": 1.0199594497680664,
       "learning_rate": 8.222222222222222e-06,
-      "loss": 0.02,
+      "loss": 0.024,
       "step": 1300
     },
     {
       "epoch": 5.321285140562249,
-      "grad_norm": 1.5528149604797363,
+      "grad_norm": 1.3515305519104004,
       "learning_rate": 8.166666666666668e-06,
-      "loss": 0.019,
+      "loss": 0.0192,
       "step": 1325
     },
     {
       "epoch": 5.421686746987952,
-      "grad_norm": 0.8070461750030518,
+      "grad_norm": 0.9697476625442505,
       "learning_rate": 8.111111111111112e-06,
-      "loss": 0.019,
+      "loss": 0.0191,
       "step": 1350
     },
     {
       "epoch": 5.522088353413655,
-      "grad_norm": 1.8495326042175293,
+      "grad_norm": 1.726500153541565,
       "learning_rate": 8.055555555555557e-06,
-      "loss": 0.0248,
+      "loss": 0.0254,
       "step": 1375
     },
     {
       "epoch": 5.622489959839357,
-      "grad_norm": 0.8098943829536438,
+      "grad_norm": 1.5264352560043335,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.022,
+      "loss": 0.0246,
       "step": 1400
     },
     {
       "epoch": 5.72289156626506,
-      "grad_norm": 1.6946097612380981,
+      "grad_norm": 0.7463297247886658,
       "learning_rate": 7.944444444444445e-06,
-      "loss": 0.0202,
+      "loss": 0.0214,
       "step": 1425
     },
     {
       "epoch": 5.823293172690763,
-      "grad_norm": 3.2773447036743164,
+      "grad_norm": 0.7417141795158386,
       "learning_rate": 7.88888888888889e-06,
-      "loss": 0.0199,
+      "loss": 0.0198,
       "step": 1450
     },
     {
       "epoch": 5.923694779116466,
-      "grad_norm": 1.4831568002700806,
+      "grad_norm": 1.0998048782348633,
       "learning_rate": 7.833333333333333e-06,
-      "loss": 0.0216,
+      "loss": 0.024,
       "step": 1475
     },
     {
       "epoch": 6.024096385542169,
-      "grad_norm": 0.6036717891693115,
+      "grad_norm": 0.6888712048530579,
       "learning_rate": 7.77777777777778e-06,
-      "loss": 0.0226,
+      "loss": 0.0212,
       "step": 1500
     },
     {
       "epoch": 6.124497991967871,
-      "grad_norm": 0.7774052619934082,
+      "grad_norm": 0.7003078460693359,
       "learning_rate": 7.722222222222223e-06,
-      "loss": 0.0138,
+      "loss": 0.0119,
       "step": 1525
     },
     {
       "epoch": 6.224899598393574,
-      "grad_norm": 0.9542578458786011,
+      "grad_norm": 0.5591835379600525,
       "learning_rate": 7.666666666666667e-06,
-      "loss": 0.0116,
+      "loss": 0.0125,
       "step": 1550
     },
     {
       "epoch": 6.325301204819277,
-      "grad_norm": 0.5289311408996582,
+      "grad_norm": 0.6486973166465759,
       "learning_rate": 7.611111111111111e-06,
-      "loss": 0.0117,
+      "loss": 0.0135,
       "step": 1575
     },
     {
       "epoch": 6.42570281124498,
-      "grad_norm": 0.8694401383399963,
+      "grad_norm": 0.5251615643501282,
       "learning_rate": 7.555555555555556e-06,
-      "loss": 0.0126,
+      "loss": 0.0122,
       "step": 1600
     },
     {
       "epoch": 6.526104417670683,
-      "grad_norm": 1.1825799942016602,
+      "grad_norm": 1.032357931137085,
       "learning_rate": 7.500000000000001e-06,
-      "loss": 0.0116,
+      "loss": 0.0132,
       "step": 1625
     },
     {
       "epoch": 6.626506024096385,
-      "grad_norm": 1.5035139322280884,
+      "grad_norm": 1.0197020769119263,
       "learning_rate": 7.444444444444445e-06,
-      "loss": 0.0141,
+      "loss": 0.0139,
       "step": 1650
     },
     {
       "epoch": 6.7269076305220885,
-      "grad_norm": 0.9342186450958252,
+      "grad_norm": 1.1973915100097656,
       "learning_rate": 7.38888888888889e-06,
-      "loss": 0.0128,
+      "loss": 0.0151,
       "step": 1675
     },
     {
       "epoch": 6.827309236947791,
-      "grad_norm": 1.0788260698318481,
+      "grad_norm": 2.7104268074035645,
       "learning_rate": 7.333333333333333e-06,
-      "loss": 0.0126,
+      "loss": 0.0135,
       "step": 1700
     },
     {
       "epoch": 6.927710843373494,
-      "grad_norm": 1.3436901569366455,
+      "grad_norm": 1.0425926446914673,
       "learning_rate": 7.277777777777778e-06,
-      "loss": 0.0137,
+      "loss": 0.0143,
       "step": 1725
     },
     {
       "epoch": 7.028112449799197,
-      "grad_norm": 0.681151807308197,
+      "grad_norm": 1.0216535329818726,
       "learning_rate": 7.222222222222223e-06,
-      "loss": 0.0124,
+      "loss": 0.013,
       "step": 1750
     },
     {
       "epoch": 7.128514056224899,
-      "grad_norm": 0.9661208987236023,
+      "grad_norm": 1.7165677547454834,
       "learning_rate": 7.166666666666667e-06,
-      "loss": 0.008,
+      "loss": 0.0089,
       "step": 1775
     },
     {
       "epoch": 7.228915662650603,
-      "grad_norm": 1.0090680122375488,
+      "grad_norm": 0.5909074544906616,
       "learning_rate": 7.111111111111112e-06,
-      "loss": 0.0074,
+      "loss": 0.0063,
       "step": 1800
     },
     {
       "epoch": 7.329317269076305,
-      "grad_norm": 1.4430378675460815,
+      "grad_norm": 0.5838289856910706,
       "learning_rate": 7.055555555555557e-06,
-      "loss": 0.0079,
+      "loss": 0.0072,
       "step": 1825
     },
     {
       "epoch": 7.429718875502008,
-      "grad_norm": 1.111413836479187,
+      "grad_norm": 0.6081740260124207,
       "learning_rate": 7e-06,
-      "loss": 0.0085,
+      "loss": 0.0108,
       "step": 1850
     },
     {
       "epoch": 7.530120481927711,
-      "grad_norm": 0.9819089770317078,
+      "grad_norm": 0.46646830439567566,
       "learning_rate": 6.944444444444445e-06,
-      "loss": 0.0083,
+      "loss": 0.0084,
       "step": 1875
     },
     {
       "epoch": 7.6305220883534135,
-      "grad_norm": 0.962616503238678,
+      "grad_norm": 1.0998950004577637,
       "learning_rate": 6.88888888888889e-06,
-      "loss": 0.008,
+      "loss": 0.01,
       "step": 1900
     },
     {
       "epoch": 7.730923694779117,
-      "grad_norm": 0.8837612271308899,
+      "grad_norm": 0.7490992546081543,
       "learning_rate": 6.833333333333334e-06,
-      "loss": 0.01,
+      "loss": 0.0089,
       "step": 1925
     },
     {
       "epoch": 7.831325301204819,
-      "grad_norm": 0.9247878789901733,
+      "grad_norm": 0.5501118898391724,
       "learning_rate": 6.777777777777779e-06,
-      "loss": 0.0095,
+      "loss": 0.0082,
       "step": 1950
     },
     {
       "epoch": 7.931726907630522,
-      "grad_norm": 0.6101934909820557,
+      "grad_norm": 0.8928245902061462,
       "learning_rate": 6.7222222222222235e-06,
-      "loss": 0.0092,
+      "loss": 0.0097,
       "step": 1975
     },
     {
       "epoch": 8.032128514056225,
-      "grad_norm": 0.9556750655174255,
+      "grad_norm": 0.43832483887672424,
       "learning_rate": 6.666666666666667e-06,
-      "loss": 0.0069,
+      "loss": 0.0088,
       "step": 2000
     },
     {
       "epoch": 8.032128514056225,
-      "eval_loss": 0.36700183153152466,
-      "eval_runtime": 2231.9147,
-      "eval_samples_per_second": 2.411,
-      "eval_steps_per_second": 0.151,
-      "eval_wer": 0.20452105451509006,
+      "eval_loss": 0.3692984879016876,
+      "eval_runtime": 2080.2644,
+      "eval_samples_per_second": 2.587,
+      "eval_steps_per_second": 0.162,
+      "eval_wer": 0.2145015706389916,
       "step": 2000
     },
     {
       "epoch": 8.132530120481928,
-      "grad_norm": 0.511352002620697,
+      "grad_norm": 0.46993330121040344,
       "learning_rate": 6.6111111111111115e-06,
-      "loss": 0.0051,
+      "loss": 0.0063,
       "step": 2025
     },
     {
       "epoch": 8.23293172690763,
-      "grad_norm": 0.762175977230072,
+      "grad_norm": 0.1431354582309723,
       "learning_rate": 6.555555555555556e-06,
       "loss": 0.0058,
       "step": 2050
     },
     {
       "epoch": 8.333333333333334,
-      "grad_norm": 0.09607477486133575,
+      "grad_norm": 0.4667530953884125,
       "learning_rate": 6.5000000000000004e-06,
-      "loss": 0.0038,
+      "loss": 0.0065,
       "step": 2075
     },
     {
       "epoch": 8.433734939759036,
-      "grad_norm": 0.6300207376480103,
+      "grad_norm": 0.6301947236061096,
       "learning_rate": 6.444444444444445e-06,
-      "loss": 0.0048,
+      "loss": 0.0049,
       "step": 2100
     },
     {
       "epoch": 8.534136546184738,
-      "grad_norm": 0.21270623803138733,
+      "grad_norm": 0.14466367661952972,
       "learning_rate": 6.3888888888888885e-06,
-      "loss": 0.0091,
+      "loss": 0.0068,
       "step": 2125
     },
     {
       "epoch": 8.634538152610443,
-      "grad_norm": 0.6666727066040039,
+      "grad_norm": 1.5687779188156128,
       "learning_rate": 6.333333333333333e-06,
-      "loss": 0.0058,
+      "loss": 0.0057,
       "step": 2150
     },
     {
       "epoch": 8.734939759036145,
-      "grad_norm": 1.9148590564727783,
+      "grad_norm": 0.33468949794769287,
       "learning_rate": 6.277777777777778e-06,
-      "loss": 0.0063,
+      "loss": 0.006,
       "step": 2175
     },
     {
       "epoch": 8.835341365461847,
-      "grad_norm": 0.272132009267807,
+      "grad_norm": 0.4165504276752472,
       "learning_rate": 6.222222222222223e-06,
-      "loss": 0.0056,
+      "loss": 0.0057,
       "step": 2200
     },
     {
       "epoch": 8.93574297188755,
-      "grad_norm": 1.3808581829071045,
+      "grad_norm": 0.9954782128334045,
       "learning_rate": 6.166666666666667e-06,
-      "loss": 0.0062,
+      "loss": 0.0061,
       "step": 2225
     },
     {
       "epoch": 9.036144578313253,
-      "grad_norm": 0.6878290772438049,
+      "grad_norm": 0.2389974594116211,
       "learning_rate": 6.111111111111112e-06,
-      "loss": 0.0034,
+      "loss": 0.0053,
       "step": 2250
     },
     {
       "epoch": 9.136546184738956,
-      "grad_norm": 0.1612684726715088,
+      "grad_norm": 0.1630544662475586,
       "learning_rate": 6.055555555555555e-06,
-      "loss": 0.0034,
+      "loss": 0.0028,
       "step": 2275
     },
     {
       "epoch": 9.236947791164658,
-      "grad_norm": 0.4624500572681427,
+      "grad_norm": 0.4344353973865509,
       "learning_rate": 6e-06,
-      "loss": 0.003,
+      "loss": 0.0038,
       "step": 2300
     },
     {
       "epoch": 9.337349397590362,
-      "grad_norm": 0.5339928865432739,
+      "grad_norm": 1.4055259227752686,
       "learning_rate": 5.944444444444445e-06,
-      "loss": 0.0024,
+      "loss": 0.0028,
       "step": 2325
     },
     {
       "epoch": 9.437751004016064,
-      "grad_norm": 0.4839150905609131,
+      "grad_norm": 0.8303292393684387,
       "learning_rate": 5.88888888888889e-06,
-      "loss": 0.0033,
+      "loss": 0.0041,
       "step": 2350
     },
     {
       "epoch": 9.538152610441767,
-      "grad_norm": 0.11668159067630768,
+      "grad_norm": 0.12666375935077667,
       "learning_rate": 5.833333333333334e-06,
-      "loss": 0.0033,
+      "loss": 0.004,
       "step": 2375
     },
     {
       "epoch": 9.638554216867469,
-      "grad_norm": 0.2360651195049286,
+      "grad_norm": 0.19957666099071503,
       "learning_rate": 5.777777777777778e-06,
-      "loss": 0.0031,
+      "loss": 0.0037,
       "step": 2400
     },
     {
       "epoch": 9.738955823293173,
-      "grad_norm": 0.4457962214946747,
+      "grad_norm": 1.1443361043930054,
       "learning_rate": 5.722222222222222e-06,
-      "loss": 0.0025,
+      "loss": 0.0024,
       "step": 2425
     },
     {
       "epoch": 9.839357429718875,
-      "grad_norm": 0.41899001598358154,
+      "grad_norm": 0.7928004264831543,
       "learning_rate": 5.666666666666667e-06,
-      "loss": 0.0031,
+      "loss": 0.0042,
       "step": 2450
     },
     {
       "epoch": 9.939759036144578,
-      "grad_norm": 0.4257371425628662,
+      "grad_norm": 1.8562343120574951,
       "learning_rate": 5.611111111111112e-06,
-      "loss": 0.0052,
+      "loss": 0.0041,
       "step": 2475
     },
     {
       "epoch": 10.040160642570282,
-      "grad_norm": 0.4485076367855072,
+      "grad_norm": 0.47182267904281616,
       "learning_rate": 5.555555555555557e-06,
-      "loss": 0.0025,
+      "loss": 0.0031,
       "step": 2500
     },
     {
       "epoch": 10.140562248995984,
-      "grad_norm": 0.30146104097366333,
+      "grad_norm": 0.424947053194046,
       "learning_rate": 5.500000000000001e-06,
-      "loss": 0.0021,
+      "loss": 0.0023,
       "step": 2525
     },
     {
       "epoch": 10.240963855421686,
-      "grad_norm": 0.13998575508594513,
+      "grad_norm": 0.08587998151779175,
       "learning_rate": 5.444444444444445e-06,
-      "loss": 0.0025,
+      "loss": 0.002,
       "step": 2550
     },
     {
       "epoch": 10.34136546184739,
-      "grad_norm": 1.0045596361160278,
+      "grad_norm": 0.4150608777999878,
       "learning_rate": 5.388888888888889e-06,
-      "loss": 0.0023,
+      "loss": 0.0024,
       "step": 2575
     },
     {
       "epoch": 10.441767068273093,
-      "grad_norm": 0.07078930735588074,
+      "grad_norm": 1.4853534698486328,
       "learning_rate": 5.333333333333334e-06,
-      "loss": 0.0017,
+      "loss": 0.0021,
       "step": 2600
     },
     {
       "epoch": 10.542168674698795,
-      "grad_norm": 0.042963068932294846,
+      "grad_norm": 0.8874973654747009,
       "learning_rate": 5.2777777777777785e-06,
-      "loss": 0.0019,
+      "loss": 0.0024,
       "step": 2625
     },
     {
       "epoch": 10.642570281124499,
-      "grad_norm": 0.13671617209911346,
+      "grad_norm": 1.0584465265274048,
       "learning_rate": 5.2222222222222226e-06,
-      "loss": 0.0018,
+      "loss": 0.0032,
       "step": 2650
     },
     {
       "epoch": 10.742971887550201,
-      "grad_norm": 0.6172053217887878,
+      "grad_norm": 0.08437704294919968,
       "learning_rate": 5.1666666666666675e-06,
-      "loss": 0.0048,
+      "loss": 0.0025,
       "step": 2675
     },
     {
       "epoch": 10.843373493975903,
-      "grad_norm": 2.069375991821289,
+      "grad_norm": 0.10795915871858597,
       "learning_rate": 5.1111111111111115e-06,
       "loss": 0.0022,
       "step": 2700
     },
     {
       "epoch": 10.943775100401606,
-      "grad_norm": 0.2148715853691101,
+      "grad_norm": 0.528668224811554,
       "learning_rate": 5.0555555555555555e-06,
-      "loss": 0.0021,
+      "loss": 0.003,
       "step": 2725
     },
     {
       "epoch": 11.04417670682731,
-      "grad_norm": 0.045111846178770065,
+      "grad_norm": 0.11841586232185364,
       "learning_rate": 5e-06,
-      "loss": 0.0012,
+      "loss": 0.0017,
       "step": 2750
     },
     {
       "epoch": 11.144578313253012,
-      "grad_norm": 0.03575390577316284,
+      "grad_norm": 0.060645826160907745,
       "learning_rate": 4.944444444444445e-06,
-      "loss": 0.0009,
+      "loss": 0.0017,
       "step": 2775
     },
     {
       "epoch": 11.244979919678714,
-      "grad_norm": 0.036444906145334244,
+      "grad_norm": 0.04768579453229904,
       "learning_rate": 4.888888888888889e-06,
       "loss": 0.0012,
       "step": 2800
     },
     {
       "epoch": 11.345381526104418,
-      "grad_norm": 0.18367384374141693,
+      "grad_norm": 0.05169716849923134,
       "learning_rate": 4.833333333333333e-06,
-      "loss": 0.0015,
+      "loss": 0.0013,
       "step": 2825
     },
     {
       "epoch": 11.44578313253012,
-      "grad_norm": 0.03629644960165024,
+      "grad_norm": 0.04550230875611305,
       "learning_rate": 4.777777777777778e-06,
-      "loss": 0.0011,
+      "loss": 0.0009,
       "step": 2850
     },
     {
       "epoch": 11.546184738955823,
-      "grad_norm": 0.08484747260808945,
+      "grad_norm": 0.05582621693611145,
       "learning_rate": 4.722222222222222e-06,
-      "loss": 0.0011,
+      "loss": 0.001,
       "step": 2875
     },
     {
       "epoch": 11.646586345381525,
-      "grad_norm": 0.2961013913154602,
+      "grad_norm": 0.02758258767426014,
       "learning_rate": 4.666666666666667e-06,
-      "loss": 0.0011,
+      "loss": 0.0008,
       "step": 2900
     },
     {
       "epoch": 11.74698795180723,
-      "grad_norm": 0.04454226791858673,
+      "grad_norm": 0.2900394797325134,
       "learning_rate": 4.611111111111112e-06,
-      "loss": 0.0019,
+      "loss": 0.0017,
       "step": 2925
     },
     {
       "epoch": 11.847389558232932,
-      "grad_norm": 0.06762082129716873,
+      "grad_norm": 0.381853848695755,
       "learning_rate": 4.555555555555556e-06,
-      "loss": 0.001,
+      "loss": 0.0015,
       "step": 2950
     },
     {
       "epoch": 11.947791164658634,
-      "grad_norm": 0.051877710968256,
+      "grad_norm": 0.06450287997722626,
       "learning_rate": 4.5e-06,
-      "loss": 0.0019,
+      "loss": 0.0011,
       "step": 2975
     },
     {
       "epoch": 12.048192771084338,
-      "grad_norm": 0.04771376773715019,
+      "grad_norm": 0.047577910125255585,
       "learning_rate": 4.444444444444444e-06,
-      "loss": 0.0013,
+      "loss": 0.0008,
       "step": 3000
     },
     {
       "epoch": 12.048192771084338,
-      "eval_loss": 0.3645249009132385,
-      "eval_runtime": 2193.5952,
-      "eval_samples_per_second": 2.453,
-      "eval_steps_per_second": 0.154,
-      "eval_wer": 0.19776134239930018,
+      "eval_loss": 0.40374720096588135,
+      "eval_runtime": 2284.8901,
+      "eval_samples_per_second": 2.355,
+      "eval_steps_per_second": 0.147,
+      "eval_wer": 0.20497832915821704,
       "step": 3000
     },
     {
       "epoch": 12.14859437751004,
-      "grad_norm": 0.020401790738105774,
+      "grad_norm": 0.021902164444327354,
       "learning_rate": 4.388888888888889e-06,
-      "loss": 0.0007,
+      "loss": 0.0006,
       "step": 3025
     },
     {
       "epoch": 12.248995983935743,
-      "grad_norm": 0.037684116512537,
+      "grad_norm": 0.02727910317480564,
       "learning_rate": 4.333333333333334e-06,
-      "loss": 0.0006,
+      "loss": 0.0007,
       "step": 3050
     },
     {
       "epoch": 12.349397590361447,
-      "grad_norm": 0.025216449052095413,
+      "grad_norm": 0.035888105630874634,
       "learning_rate": 4.277777777777778e-06,
-      "loss": 0.0005,
+      "loss": 0.0015,
       "step": 3075
     },
     {
       "epoch": 12.449799196787149,
-      "grad_norm": 0.021326890215277672,
+      "grad_norm": 0.021920297294855118,
       "learning_rate": 4.222222222222223e-06,
-      "loss": 0.0005,
+      "loss": 0.0006,
       "step": 3100
     },
     {
       "epoch": 12.550200803212851,
-      "grad_norm": 0.02904532290995121,
+      "grad_norm": 0.023950112983584404,
       "learning_rate": 4.166666666666667e-06,
-      "loss": 0.0009,
+      "loss": 0.0006,
       "step": 3125
     },
     {
       "epoch": 12.650602409638553,
-      "grad_norm": 0.15470072627067566,
+      "grad_norm": 0.0257779061794281,
       "learning_rate": 4.111111111111111e-06,
       "loss": 0.0006,
       "step": 3150
     },
     {
       "epoch": 12.751004016064257,
-      "grad_norm": 0.5353085398674011,
+      "grad_norm": 0.01860366389155388,
       "learning_rate": 4.055555555555556e-06,
-      "loss": 0.0008,
+      "loss": 0.0006,
       "step": 3175
     },
     {
       "epoch": 12.85140562248996,
-      "grad_norm": 0.1263090819120407,
+      "grad_norm": 0.01894628070294857,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 0.0005,
+      "loss": 0.0015,
       "step": 3200
     },
     {
       "epoch": 12.951807228915662,
-      "grad_norm": 0.018346522003412247,
+      "grad_norm": 0.11034880578517914,
       "learning_rate": 3.944444444444445e-06,
-      "loss": 0.0005,
+      "loss": 0.0008,
       "step": 3225
     },
     {
       "epoch": 13.052208835341366,
-      "grad_norm": 0.012502867728471756,
+      "grad_norm": 0.02306952327489853,
       "learning_rate": 3.88888888888889e-06,
       "loss": 0.0005,
       "step": 3250
     },
     {
       "epoch": 13.152610441767068,
-      "grad_norm": 0.035849809646606445,
+      "grad_norm": 0.019928133115172386,
       "learning_rate": 3.833333333333334e-06,
       "loss": 0.0005,
       "step": 3275
     },
     {
       "epoch": 13.25301204819277,
-      "grad_norm": 0.013340278528630733,
+      "grad_norm": 0.01607375033199787,
       "learning_rate": 3.777777777777778e-06,
-      "loss": 0.0004,
+      "loss": 0.0005,
       "step": 3300
     },
     {
       "epoch": 13.353413654618475,
-      "grad_norm": 0.01812613196671009,
+      "grad_norm": 0.018628498539328575,
       "learning_rate": 3.7222222222222225e-06,
       "loss": 0.0004,
       "step": 3325
     },
     {
       "epoch": 13.453815261044177,
-      "grad_norm": 0.016993574798107147,
+      "grad_norm": 0.021479720249772072,
       "learning_rate": 3.6666666666666666e-06,
       "loss": 0.0004,
       "step": 3350
     },
     {
       "epoch": 13.55421686746988,
-      "grad_norm": 0.012604492716491222,
+      "grad_norm": 0.016515526920557022,
       "learning_rate": 3.6111111111111115e-06,
       "loss": 0.0004,
       "step": 3375
     },
     {
       "epoch": 13.654618473895582,
-      "grad_norm": 0.013149112462997437,
+      "grad_norm": 0.02035825327038765,
       "learning_rate": 3.555555555555556e-06,
-      "loss": 0.0007,
+      "loss": 0.0004,
       "step": 3400
     },
     {
       "epoch": 13.755020080321286,
-      "grad_norm": 0.01593812368810177,
+      "grad_norm": 0.01709391362965107,
       "learning_rate": 3.5e-06,
       "loss": 0.0004,
       "step": 3425
     },
     {
       "epoch": 13.855421686746988,
-      "grad_norm": 0.01589050143957138,
+      "grad_norm": 0.01652124896645546,
       "learning_rate": 3.444444444444445e-06,
       "loss": 0.0004,
       "step": 3450
     },
     {
       "epoch": 13.95582329317269,
-      "grad_norm": 0.014971195720136166,
+      "grad_norm": 0.01660318858921528,
       "learning_rate": 3.3888888888888893e-06,
-      "loss": 0.0004,
+      "loss": 0.0005,
       "step": 3475
     },
     {
       "epoch": 14.056224899598394,
-      "grad_norm": 0.01241573691368103,
+      "grad_norm": 0.011111569590866566,
       "learning_rate": 3.3333333333333333e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3500
     },
     {
       "epoch": 14.156626506024097,
-      "grad_norm": 0.011940378695726395,
+      "grad_norm": 0.015211747027933598,
       "learning_rate": 3.277777777777778e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3525
     },
     {
       "epoch": 14.257028112449799,
-      "grad_norm": 0.01354218740016222,
+      "grad_norm": 0.012760845012962818,
       "learning_rate": 3.2222222222222227e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3550
     },
     {
       "epoch": 14.357429718875501,
-      "grad_norm": 0.011458562687039375,
+      "grad_norm": 0.015234610997140408,
       "learning_rate": 3.1666666666666667e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3575
     },
     {
       "epoch": 14.457831325301205,
-      "grad_norm": 0.010744108818471432,
+      "grad_norm": 0.011275901459157467,
       "learning_rate": 3.1111111111111116e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3600
     },
     {
       "epoch": 14.558232931726907,
-      "grad_norm": 0.01174489688128233,
+      "grad_norm": 0.012572256848216057,
       "learning_rate": 3.055555555555556e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3625
     },
     {
       "epoch": 14.65863453815261,
-      "grad_norm": 0.01333660539239645,
+      "grad_norm": 0.012502871453762054,
       "learning_rate": 3e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3650
     },
     {
       "epoch": 14.759036144578314,
-      "grad_norm": 0.012421938590705395,
+      "grad_norm": 0.012110814452171326,
       "learning_rate": 2.944444444444445e-06,
       "loss": 0.0004,
       "step": 3675
     },
     {
       "epoch": 14.859437751004016,
-      "grad_norm": 0.01224998664110899,
+      "grad_norm": 0.012966180220246315,
       "learning_rate": 2.888888888888889e-06,
-      "loss": 0.0003,
+      "loss": 0.0018,
       "step": 3700
     },
     {
       "epoch": 14.959839357429718,
-      "grad_norm": 0.015473966486752033,
+      "grad_norm": 0.026951424777507782,
       "learning_rate": 2.8333333333333335e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3725
     },
     {
       "epoch": 15.060240963855422,
-      "grad_norm": 0.012373683042824268,
+      "grad_norm": 0.015452017076313496,
       "learning_rate": 2.7777777777777783e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3750
     },
     {
       "epoch": 15.160642570281125,
-      "grad_norm": 0.010334338992834091,
+      "grad_norm": 0.014466413296759129,
       "learning_rate": 2.7222222222222224e-06,
       "loss": 0.0004,
       "step": 3775
     },
     {
       "epoch": 15.261044176706827,
-      "grad_norm": 0.015428266488015652,
+      "grad_norm": 0.016020847484469414,
       "learning_rate": 2.666666666666667e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3800
     },
     {
       "epoch": 15.36144578313253,
-      "grad_norm": 0.011229559779167175,
+      "grad_norm": 0.013202512636780739,
       "learning_rate": 2.6111111111111113e-06,
       "loss": 0.0003,
       "step": 3825
     },
     {
       "epoch": 15.461847389558233,
-      "grad_norm": 0.009039835073053837,
+      "grad_norm": 0.01130182109773159,
       "learning_rate": 2.5555555555555557e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3850
     },
     {
       "epoch": 15.562248995983936,
-      "grad_norm": 0.0124340346083045,
+      "grad_norm": 0.016152923926711082,
       "learning_rate": 2.5e-06,
-      "loss": 0.0003,
+      "loss": 0.0005,
       "step": 3875
     },
     {
       "epoch": 15.662650602409638,
-      "grad_norm": 0.009659750387072563,
+      "grad_norm": 0.014990521594882011,
       "learning_rate": 2.4444444444444447e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3900
     },
     {
       "epoch": 15.763052208835342,
-      "grad_norm": 0.013714014552533627,
+      "grad_norm": 0.021178970113396645,
       "learning_rate": 2.388888888888889e-06,
       "loss": 0.0003,
       "step": 3925
     },
     {
       "epoch": 15.863453815261044,
-      "grad_norm": 0.00784530583769083,
+      "grad_norm": 0.01082056388258934,
       "learning_rate": 2.3333333333333336e-06,
       "loss": 0.0003,
       "step": 3950
     },
     {
       "epoch": 15.963855421686747,
-      "grad_norm": 0.010499561205506325,
+      "grad_norm": 0.009652756154537201,
       "learning_rate": 2.277777777777778e-06,
-      "loss": 0.0003,
+      "loss": 0.0004,
       "step": 3975
     },
     {
       "epoch": 16.06425702811245,
-      "grad_norm": 0.01295757107436657,
+      "grad_norm": 0.013466122560203075,
       "learning_rate": 2.222222222222222e-06,
-      "loss": 0.0003,
+      "loss": 0.0005,
       "step": 4000
     },
     {
       "epoch": 16.06425702811245,
-      "eval_loss": 0.4106931984424591,
-      "eval_runtime": 2194.2451,
-      "eval_samples_per_second": 2.452,
-      "eval_steps_per_second": 0.154,
-      "eval_wer": 0.19143902342041433,
+      "eval_loss": 0.43511074781417847,
+      "eval_runtime": 2099.2694,
+      "eval_samples_per_second": 2.563,
+      "eval_steps_per_second": 0.161,
+      "eval_wer": 0.20501809217066286,
       "step": 4000
     },
     {
       "epoch": 16.164658634538153,
-      "grad_norm": 0.010303654707968235,
+      "grad_norm": 0.010581804439425468,
       "learning_rate": 2.166666666666667e-06,
       "loss": 0.0003,
       "step": 4025
     },
     {
       "epoch": 16.265060240963855,
-      "grad_norm": 0.008325839415192604,
+      "grad_norm": 0.009896391071379185,
       "learning_rate": 2.1111111111111114e-06,
       "loss": 0.0003,
       "step": 4050
     },
     {
       "epoch": 16.365461847389557,
-      "grad_norm": 0.00958819966763258,
+      "grad_norm": 0.01030905544757843,
       "learning_rate": 2.0555555555555555e-06,
       "loss": 0.0003,
       "step": 4075
     },
     {
       "epoch": 16.46586345381526,
-      "grad_norm": 0.008696039207279682,
+      "grad_norm": 0.009518243372440338,
       "learning_rate": 2.0000000000000003e-06,
       "loss": 0.0003,
       "step": 4100
     },
     {
       "epoch": 16.566265060240966,
-      "grad_norm": 0.010361140593886375,
+      "grad_norm": 0.013147858902812004,
       "learning_rate": 1.944444444444445e-06,
       "loss": 0.0003,
       "step": 4125
     },
     {
       "epoch": 16.666666666666668,
-      "grad_norm": 0.008520281873643398,
+      "grad_norm": 0.012813227251172066,
       "learning_rate": 1.888888888888889e-06,
       "loss": 0.0003,
       "step": 4150
     },
     {
       "epoch": 16.76706827309237,
-      "grad_norm": 0.012094419449567795,
+      "grad_norm": 0.012356545776128769,
       "learning_rate": 1.8333333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0003,
       "step": 4175
     },
     {
       "epoch": 16.867469879518072,
-      "grad_norm": 0.00969509407877922,
+      "grad_norm": 0.0107422498986125,
       "learning_rate": 1.777777777777778e-06,
       "loss": 0.0003,
       "step": 4200
     },
     {
       "epoch": 16.967871485943775,
-      "grad_norm": 0.010107293725013733,
+      "grad_norm": 0.01414831355214119,
       "learning_rate": 1.7222222222222224e-06,
       "loss": 0.0003,
       "step": 4225
     },
     {
       "epoch": 17.068273092369477,
-      "grad_norm": 0.008444724604487419,
+      "grad_norm": 0.010546191595494747,
       "learning_rate": 1.6666666666666667e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4250
     },
     {
       "epoch": 17.16867469879518,
-      "grad_norm": 0.008527095429599285,
+      "grad_norm": 0.009328695945441723,
       "learning_rate": 1.6111111111111113e-06,
       "loss": 0.0003,
       "step": 4275
     },
     {
       "epoch": 17.269076305220885,
-      "grad_norm": 0.007767422124743462,
+      "grad_norm": 0.010023394599556923,
       "learning_rate": 1.5555555555555558e-06,
       "loss": 0.0003,
       "step": 4300
     },
     {
       "epoch": 17.369477911646587,
-      "grad_norm": 0.008298359811306,
+      "grad_norm": 0.009099387563765049,
       "learning_rate": 1.5e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4325
     },
     {
       "epoch": 17.46987951807229,
-      "grad_norm": 0.00869645643979311,
+      "grad_norm": 0.010183015838265419,
       "learning_rate": 1.4444444444444445e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4350
     },
     {
       "epoch": 17.570281124497992,
-      "grad_norm": 0.009767497889697552,
+      "grad_norm": 0.010759984143078327,
       "learning_rate": 1.3888888888888892e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4375
     },
     {
       "epoch": 17.670682730923694,
-      "grad_norm": 0.009676006622612476,
+      "grad_norm": 0.013764439150691032,
       "learning_rate": 1.3333333333333334e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4400
     },
     {
       "epoch": 17.771084337349397,
-      "grad_norm": 0.010664808563888073,
+      "grad_norm": 0.010277372784912586,
       "learning_rate": 1.2777777777777779e-06,
       "loss": 0.0003,
       "step": 4425
     },
     {
       "epoch": 17.8714859437751,
-      "grad_norm": 0.011917660012841225,
+      "grad_norm": 0.012754006311297417,
       "learning_rate": 1.2222222222222223e-06,
       "loss": 0.0003,
       "step": 4450
     },
     {
       "epoch": 17.971887550200805,
-      "grad_norm": 0.008082253858447075,
+      "grad_norm": 0.009855114854872227,
       "learning_rate": 1.1666666666666668e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4475
     },
     {
       "epoch": 18.072289156626507,
-      "grad_norm": 0.008449643850326538,
+      "grad_norm": 0.008605270646512508,
       "learning_rate": 1.111111111111111e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4500
     },
     {
       "epoch": 18.17269076305221,
-      "grad_norm": 0.00893787294626236,
+      "grad_norm": 0.008842888288199902,
       "learning_rate": 1.0555555555555557e-06,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4525
     },
     {
       "epoch": 18.27309236947791,
-      "grad_norm": 0.007984068244695663,
+      "grad_norm": 0.008844111114740372,
       "learning_rate": 1.0000000000000002e-06,
       "loss": 0.0003,
       "step": 4550
     },
     {
       "epoch": 18.373493975903614,
-      "grad_norm": 0.008762707002460957,
+      "grad_norm": 0.009015376679599285,
       "learning_rate": 9.444444444444445e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4575
     },
     {
       "epoch": 18.473895582329316,
-      "grad_norm": 0.008354073390364647,
+      "grad_norm": 0.009200502187013626,
       "learning_rate": 8.88888888888889e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4600
     },
     {
       "epoch": 18.57429718875502,
-      "grad_norm": 0.010145510546863079,
+      "grad_norm": 0.009783760644495487,
       "learning_rate": 8.333333333333333e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4625
     },
     {
       "epoch": 18.674698795180724,
-      "grad_norm": 0.009271888993680477,
+      "grad_norm": 0.009740373119711876,
       "learning_rate": 7.777777777777779e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4650
     },
     {
       "epoch": 18.775100401606426,
-      "grad_norm": 0.007404220290482044,
+      "grad_norm": 0.009242956526577473,
       "learning_rate": 7.222222222222222e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4675
     },
     {
       "epoch": 18.87550200803213,
-      "grad_norm": 0.00828209612518549,
+      "grad_norm": 0.009656492620706558,
       "learning_rate": 6.666666666666667e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4700
     },
     {
       "epoch": 18.97590361445783,
-      "grad_norm": 0.008047865703701973,
+      "grad_norm": 0.009480450302362442,
       "learning_rate": 6.111111111111112e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4725
     },
     {
       "epoch": 19.076305220883533,
-      "grad_norm": 0.00727940583601594,
+      "grad_norm": 0.00849216990172863,
       "learning_rate": 5.555555555555555e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4750
     },
     {
       "epoch": 19.176706827309236,
-      "grad_norm": 0.008462085388600826,
+      "grad_norm": 0.00895814411342144,
       "learning_rate": 5.000000000000001e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4775
     },
     {
       "epoch": 19.27710843373494,
-      "grad_norm": 0.008368249051272869,
+      "grad_norm": 0.008503418415784836,
       "learning_rate": 4.444444444444445e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4800
     },
     {
       "epoch": 19.377510040160644,
-      "grad_norm": 0.007355119101703167,
+      "grad_norm": 0.009040705859661102,
       "learning_rate": 3.8888888888888895e-07,
       "loss": 0.0002,
       "step": 4825
     },
     {
       "epoch": 19.477911646586346,
-      "grad_norm": 0.00911016296595335,
+      "grad_norm": 0.009361708536744118,
       "learning_rate": 3.3333333333333335e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4850
     },
     {
       "epoch": 19.57831325301205,
-      "grad_norm": 0.00813527312129736,
+      "grad_norm": 0.009007732383906841,
       "learning_rate": 2.7777777777777776e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4875
     },
     {
       "epoch": 19.67871485943775,
-      "grad_norm": 0.007565053179860115,
+      "grad_norm": 0.008576186373829842,
       "learning_rate": 2.2222222222222224e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4900
     },
     {
       "epoch": 19.779116465863453,
-      "grad_norm": 0.008976846002042294,
+      "grad_norm": 0.009484385140240192,
       "learning_rate": 1.6666666666666668e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4925
     },
     {
       "epoch": 19.879518072289155,
-      "grad_norm": 0.007831977680325508,
+      "grad_norm": 0.007250281050801277,
       "learning_rate": 1.1111111111111112e-07,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4950
     },
     {
       "epoch": 19.97991967871486,
-      "grad_norm": 0.0071573760360479355,
+      "grad_norm": 0.00842439103871584,
       "learning_rate": 5.555555555555556e-08,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 4975
     },
     {
       "epoch": 20.080321285140563,
-      "grad_norm": 0.008769778534770012,
+      "grad_norm": 0.00960406195372343,
       "learning_rate": 0.0,
-      "loss": 0.0002,
+      "loss": 0.0003,
       "step": 5000
     },
     {
       "epoch": 20.080321285140563,
-      "eval_loss": 0.41853219270706177,
-      "eval_runtime": 2165.4621,
-      "eval_samples_per_second": 2.485,
-      "eval_steps_per_second": 0.156,
-      "eval_wer": 0.19108115630840192,
+      "eval_loss": 0.44258150458335876,
+      "eval_runtime": 2114.4157,
+      "eval_samples_per_second": 2.545,
+      "eval_steps_per_second": 0.159,
+      "eval_wer": 0.2026720744363593,
       "step": 5000
     },
     {
       "epoch": 20.080321285140563,
       "step": 5000,
       "total_flos": 5.435997290496e+20,
-      "train_loss": 0.05159526972509921,
-      "train_runtime": 59448.8268,
-      "train_samples_per_second": 2.691,
-      "train_steps_per_second": 0.084
+      "train_loss": 0.0536484617867507,
+      "train_runtime": 58113.0051,
+      "train_samples_per_second": 2.753,
+      "train_steps_per_second": 0.086
     }
   ],
   "logging_steps": 25,