End of training

README.md

@@ -2,6 +2,8 @@
 license: apache-2.0
 base_model: facebook/wav2vec2-large-xlsr-53
 tags:
+- automatic-speech-recognition
+- DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv
 - generated_from_trainer
 metrics:
 - wer
@@ -15,7 +17,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # wav2vec2-xlsr-53-ft-btb-ccv-cy
 
-This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on an unknown dataset.
+This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the DEWIBRYNJONES/BANC-TRAWSGRIFIADAU-BANGOR-CLEAN-WITH-CCV - DEFAULT dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.3888
 - Wer: 0.2885

all_results.json

@@ -1,15 +1,15 @@
 {
-    "epoch": 1.9359272164150214,
-    "eval_loss": 0.4060601592063904,
-    "eval_runtime": 154.5004,
+    "epoch": 3.8714672861014323,
+    "eval_loss": 0.38884833455085754,
+    "eval_runtime": 156.3404,
     "eval_samples": 5656,
-    "eval_samples_per_second": 36.608,
-    "eval_steps_per_second": 4.576,
-    "eval_wer": 0.31914108263388485,
-    "total_flos": 1.1256772526671323e+19,
-    "train_loss": 2.751509769715472e-05,
-    "train_runtime": 1.6872,
+    "eval_samples_per_second": 36.177,
+    "eval_steps_per_second": 4.522,
+    "eval_wer": 0.2884883888879973,
+    "total_flos": 2.249387574100498e+19,
+    "train_loss": 0.15996522521972656,
+    "train_runtime": 19346.8732,
     "train_samples": 41326,
-    "train_samples_per_second": 47416.17,
-    "train_steps_per_second": 5927.021
+    "train_samples_per_second": 8.27,
+    "train_steps_per_second": 1.034
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 1.9359272164150214,
-    "eval_loss": 0.4060601592063904,
-    "eval_runtime": 154.5004,
+    "epoch": 3.8714672861014323,
+    "eval_loss": 0.38884833455085754,
+    "eval_runtime": 156.3404,
     "eval_samples": 5656,
-    "eval_samples_per_second": 36.608,
-    "eval_steps_per_second": 4.576,
-    "eval_wer": 0.31914108263388485
+    "eval_samples_per_second": 36.177,
+    "eval_steps_per_second": 4.522,
+    "eval_wer": 0.2884883888879973
 }

runs/Jun03_08-01-01_3405f00559a4/events.out.tfevents.1717417641.3405f00559a4.205.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3bb01bd656926a3c98192f80006878026b873d0a450b78eaf06d5fd7b92bb8f
+size 412

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 1.9359272164150214,
-    "total_flos": 1.1256772526671323e+19,
-    "train_loss": 2.751509769715472e-05,
-    "train_runtime": 1.6872,
+    "epoch": 3.8714672861014323,
+    "total_flos": 2.249387574100498e+19,
+    "train_loss": 0.15996522521972656,
+    "train_runtime": 19346.8732,
     "train_samples": 41326,
-    "train_samples_per_second": 47416.17,
-    "train_steps_per_second": 5927.021
+    "train_samples_per_second": 8.27,
+    "train_steps_per_second": 1.034
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 1.9359272164150214,
+  "epoch": 3.8714672861014323,
   "eval_steps": 100,
-  "global_step": 10001,
+  "global_step": 20000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -1049,21 +1049,1061 @@
       "step": 10000
     },
     {
-      "epoch": 1.9359272164150214,
-      "step": 10001,
-      "total_flos": 1.1256772526671323e+19,
-      "train_loss": 2.751509769715472e-05,
-      "train_runtime": 1.6872,
-      "train_samples_per_second": 47416.17,
-      "train_steps_per_second": 5927.021
+      "epoch": 1.9550909794812235,
+      "eval_loss": 0.45229342579841614,
+      "eval_runtime": 154.2948,
+      "eval_samples_per_second": 36.657,
+      "eval_steps_per_second": 4.582,
+      "eval_wer": 0.3406140167867632,
+      "step": 10100
+    },
+    {
+      "epoch": 1.9744483159117305,
+      "eval_loss": 0.4579542577266693,
+      "eval_runtime": 151.5074,
+      "eval_samples_per_second": 37.331,
+      "eval_steps_per_second": 4.666,
+      "eval_wer": 0.3517837941936416,
+      "step": 10200
+    },
+    {
+      "epoch": 1.9938056523422376,
+      "eval_loss": 0.46043792366981506,
+      "eval_runtime": 151.4438,
+      "eval_samples_per_second": 37.347,
+      "eval_steps_per_second": 4.668,
+      "eval_wer": 0.35115790149411824,
+      "step": 10300
+    },
+    {
+      "epoch": 2.013162988772745,
+      "eval_loss": 0.46549099683761597,
+      "eval_runtime": 151.5994,
+      "eval_samples_per_second": 37.309,
+      "eval_steps_per_second": 4.664,
+      "eval_wer": 0.3552181797756415,
+      "step": 10400
+    },
+    {
+      "epoch": 2.032520325203252,
+      "grad_norm": 0.703632652759552,
+      "learning_rate": 0.0001463076923076923,
+      "loss": 0.3624,
+      "step": 10500
+    },
+    {
+      "epoch": 2.032520325203252,
+      "eval_loss": 0.4670031666755676,
+      "eval_runtime": 151.5063,
+      "eval_samples_per_second": 37.332,
+      "eval_steps_per_second": 4.666,
+      "eval_wer": 0.35144677504774435,
+      "step": 10500
+    },
+    {
+      "epoch": 2.051877661633759,
+      "eval_loss": 0.459250271320343,
+      "eval_runtime": 153.0971,
+      "eval_samples_per_second": 36.944,
+      "eval_steps_per_second": 4.618,
+      "eval_wer": 0.3628251833544639,
+      "step": 10600
+    },
+    {
+      "epoch": 2.0712349980642664,
+      "eval_loss": 0.46061432361602783,
+      "eval_runtime": 152.0732,
+      "eval_samples_per_second": 37.193,
+      "eval_steps_per_second": 4.649,
+      "eval_wer": 0.3545922870761182,
+      "step": 10700
+    },
+    {
+      "epoch": 2.0905923344947737,
+      "eval_loss": 0.46500489115715027,
+      "eval_runtime": 151.985,
+      "eval_samples_per_second": 37.214,
+      "eval_steps_per_second": 4.652,
+      "eval_wer": 0.35905377862656673,
+      "step": 10800
+    },
+    {
+      "epoch": 2.1099496709252805,
+      "eval_loss": 0.46085453033447266,
+      "eval_runtime": 152.4835,
+      "eval_samples_per_second": 37.093,
+      "eval_steps_per_second": 4.637,
+      "eval_wer": 0.35483301503747333,
+      "step": 10900
+    },
+    {
+      "epoch": 2.129307007355788,
+      "grad_norm": 0.5008242726325989,
+      "learning_rate": 0.00013863076923076922,
+      "loss": 0.3755,
+      "step": 11000
+    },
+    {
+      "epoch": 2.129307007355788,
+      "eval_loss": 0.4708138406276703,
+      "eval_runtime": 152.3457,
+      "eval_samples_per_second": 37.126,
+      "eval_steps_per_second": 4.641,
+      "eval_wer": 0.35573173275986586,
+      "step": 11000
+    },
+    {
+      "epoch": 2.148664343786295,
+      "eval_loss": 0.4649392366409302,
+      "eval_runtime": 152.7087,
+      "eval_samples_per_second": 37.038,
+      "eval_steps_per_second": 4.63,
+      "eval_wer": 0.3548009179759593,
+      "step": 11100
+    },
+    {
+      "epoch": 2.168021680216802,
+      "eval_loss": 0.4624271094799042,
+      "eval_runtime": 153.153,
+      "eval_samples_per_second": 36.93,
+      "eval_steps_per_second": 4.616,
+      "eval_wer": 0.355956412190464,
+      "step": 11200
+    },
+    {
+      "epoch": 2.1873790166473093,
+      "eval_loss": 0.45822229981422424,
+      "eval_runtime": 156.1964,
+      "eval_samples_per_second": 36.211,
+      "eval_steps_per_second": 4.526,
+      "eval_wer": 0.35229734717786587,
+      "step": 11300
+    },
+    {
+      "epoch": 2.2067363530778166,
+      "eval_loss": 0.466250479221344,
+      "eval_runtime": 152.6707,
+      "eval_samples_per_second": 37.047,
+      "eval_steps_per_second": 4.631,
+      "eval_wer": 0.3586044197653705,
+      "step": 11400
+    },
+    {
+      "epoch": 2.226093689508324,
+      "grad_norm": 0.9631055593490601,
+      "learning_rate": 0.00013093846153846151,
+      "loss": 0.3891,
+      "step": 11500
+    },
+    {
+      "epoch": 2.226093689508324,
+      "eval_loss": 0.46153655648231506,
+      "eval_runtime": 153.1909,
+      "eval_samples_per_second": 36.921,
+      "eval_steps_per_second": 4.615,
+      "eval_wer": 0.3552181797756415,
+      "step": 11500
+    },
+    {
+      "epoch": 2.2454510259388307,
+      "eval_loss": 0.4631531238555908,
+      "eval_runtime": 152.9395,
+      "eval_samples_per_second": 36.982,
+      "eval_steps_per_second": 4.623,
+      "eval_wer": 0.35886119625748264,
+      "step": 11600
+    },
+    {
+      "epoch": 2.264808362369338,
+      "eval_loss": 0.4495234191417694,
+      "eval_runtime": 153.0237,
+      "eval_samples_per_second": 36.962,
+      "eval_steps_per_second": 4.62,
+      "eval_wer": 0.3425398404776043,
+      "step": 11700
+    },
+    {
+      "epoch": 2.2841656987998453,
+      "eval_loss": 0.462666779756546,
+      "eval_runtime": 152.4714,
+      "eval_samples_per_second": 37.095,
+      "eval_steps_per_second": 4.637,
+      "eval_wer": 0.34942466017236123,
+      "step": 11800
+    },
+    {
+      "epoch": 2.303523035230352,
+      "eval_loss": 0.4550352096557617,
+      "eval_runtime": 152.8072,
+      "eval_samples_per_second": 37.014,
+      "eval_steps_per_second": 4.627,
+      "eval_wer": 0.3451717995217538,
+      "step": 11900
+    },
+    {
+      "epoch": 2.3228803716608595,
+      "grad_norm": 0.7961182594299316,
+      "learning_rate": 0.00012324615384615384,
+      "loss": 0.3946,
+      "step": 12000
+    },
+    {
+      "epoch": 2.3228803716608595,
+      "eval_loss": 0.44988927245140076,
+      "eval_runtime": 152.9644,
+      "eval_samples_per_second": 36.976,
+      "eval_steps_per_second": 4.622,
+      "eval_wer": 0.3462310025517164,
+      "step": 12000
+    },
+    {
+      "epoch": 2.3422377080913668,
+      "eval_loss": 0.4501667320728302,
+      "eval_runtime": 153.061,
+      "eval_samples_per_second": 36.953,
+      "eval_steps_per_second": 4.619,
+      "eval_wer": 0.341978141901109,
+      "step": 12100
+    },
+    {
+      "epoch": 2.3615950445218736,
+      "eval_loss": 0.4580215513706207,
+      "eval_runtime": 153.2108,
+      "eval_samples_per_second": 36.916,
+      "eval_steps_per_second": 4.615,
+      "eval_wer": 0.3412399094862865,
+      "step": 12200
+    },
+    {
+      "epoch": 2.380952380952381,
+      "eval_loss": 0.4506891667842865,
+      "eval_runtime": 153.6611,
+      "eval_samples_per_second": 36.808,
+      "eval_steps_per_second": 4.601,
+      "eval_wer": 0.34339041260772574,
+      "step": 12300
+    },
+    {
+      "epoch": 2.4003097173828882,
+      "eval_loss": 0.44618555903434753,
+      "eval_runtime": 153.273,
+      "eval_samples_per_second": 36.901,
+      "eval_steps_per_second": 4.613,
+      "eval_wer": 0.34475453772207154,
+      "step": 12400
+    },
+    {
+      "epoch": 2.419667053813395,
+      "grad_norm": 0.828158974647522,
+      "learning_rate": 0.00011556923076923076,
+      "loss": 0.3824,
+      "step": 12500
+    },
+    {
+      "epoch": 2.419667053813395,
+      "eval_loss": 0.44126543402671814,
+      "eval_runtime": 153.3979,
+      "eval_samples_per_second": 36.871,
+      "eval_steps_per_second": 4.609,
+      "eval_wer": 0.34127200654780054,
+      "step": 12500
+    },
+    {
+      "epoch": 2.4390243902439024,
+      "eval_loss": 0.44880929589271545,
+      "eval_runtime": 153.7143,
+      "eval_samples_per_second": 36.796,
+      "eval_steps_per_second": 4.599,
+      "eval_wer": 0.3443212273916323,
+      "step": 12600
+    },
+    {
+      "epoch": 2.4583817266744097,
+      "eval_loss": 0.44148463010787964,
+      "eval_runtime": 153.647,
+      "eval_samples_per_second": 36.812,
+      "eval_steps_per_second": 4.601,
+      "eval_wer": 0.3431657331771276,
+      "step": 12700
+    },
+    {
+      "epoch": 2.4777390631049165,
+      "eval_loss": 0.44202256202697754,
+      "eval_runtime": 153.5743,
+      "eval_samples_per_second": 36.829,
+      "eval_steps_per_second": 4.604,
+      "eval_wer": 0.34093498740190337,
+      "step": 12800
+    },
+    {
+      "epoch": 2.497096399535424,
+      "eval_loss": 0.4379221200942993,
+      "eval_runtime": 153.5736,
+      "eval_samples_per_second": 36.829,
+      "eval_steps_per_second": 4.604,
+      "eval_wer": 0.3361204281748006,
+      "step": 12900
+    },
+    {
+      "epoch": 2.516453735965931,
+      "grad_norm": 1.2163615226745605,
+      "learning_rate": 0.00010787692307692307,
+      "loss": 0.372,
+      "step": 13000
+    },
+    {
+      "epoch": 2.516453735965931,
+      "eval_loss": 0.43855318427085876,
+      "eval_runtime": 153.5476,
+      "eval_samples_per_second": 36.835,
+      "eval_steps_per_second": 4.604,
+      "eval_wer": 0.3334884691306511,
+      "step": 13000
+    },
+    {
+      "epoch": 2.535811072396438,
+      "eval_loss": 0.44449883699417114,
+      "eval_runtime": 153.7016,
+      "eval_samples_per_second": 36.799,
+      "eval_steps_per_second": 4.6,
+      "eval_wer": 0.3397794931873987,
+      "step": 13100
+    },
+    {
+      "epoch": 2.5551684088269453,
+      "eval_loss": 0.4401286542415619,
+      "eval_runtime": 154.2488,
+      "eval_samples_per_second": 36.668,
+      "eval_steps_per_second": 4.584,
+      "eval_wer": 0.3392819887339314,
+      "step": 13200
+    },
+    {
+      "epoch": 2.5745257452574526,
+      "eval_loss": 0.437770813703537,
+      "eval_runtime": 153.8927,
+      "eval_samples_per_second": 36.753,
+      "eval_steps_per_second": 4.594,
+      "eval_wer": 0.335077273675595,
+      "step": 13300
+    },
+    {
+      "epoch": 2.59388308168796,
+      "eval_loss": 0.4315861463546753,
+      "eval_runtime": 153.7886,
+      "eval_samples_per_second": 36.778,
+      "eval_steps_per_second": 4.597,
+      "eval_wer": 0.33517356486013705,
+      "step": 13400
+    },
+    {
+      "epoch": 2.6132404181184667,
+      "grad_norm": 1.084632158279419,
+      "learning_rate": 0.0001002,
+      "loss": 0.3521,
+      "step": 13500
+    },
+    {
+      "epoch": 2.6132404181184667,
+      "eval_loss": 0.43864014744758606,
+      "eval_runtime": 153.9711,
+      "eval_samples_per_second": 36.734,
+      "eval_steps_per_second": 4.592,
+      "eval_wer": 0.33398597358411836,
+      "step": 13500
+    },
+    {
+      "epoch": 2.632597754548974,
+      "eval_loss": 0.43551018834114075,
+      "eval_runtime": 154.3017,
+      "eval_samples_per_second": 36.655,
+      "eval_steps_per_second": 4.582,
+      "eval_wer": 0.33096884980180064,
+      "step": 13600
+    },
+    {
+      "epoch": 2.6519550909794813,
+      "eval_loss": 0.4325660765171051,
+      "eval_runtime": 154.4812,
+      "eval_samples_per_second": 36.613,
+      "eval_steps_per_second": 4.577,
+      "eval_wer": 0.3343871868530436,
+      "step": 13700
+    },
+    {
+      "epoch": 2.6713124274099886,
+      "eval_loss": 0.4263465404510498,
+      "eval_runtime": 154.0733,
+      "eval_samples_per_second": 36.71,
+      "eval_steps_per_second": 4.589,
+      "eval_wer": 0.32629872735151094,
+      "step": 13800
+    },
+    {
+      "epoch": 2.6906697638404955,
+      "eval_loss": 0.42636117339134216,
+      "eval_runtime": 154.1615,
+      "eval_samples_per_second": 36.689,
+      "eval_steps_per_second": 4.586,
+      "eval_wer": 0.32353838006130536,
+      "step": 13900
+    },
+    {
+      "epoch": 2.710027100271003,
+      "grad_norm": 1.1979655027389526,
+      "learning_rate": 9.25076923076923e-05,
+      "loss": 0.3592,
+      "step": 14000
+    },
+    {
+      "epoch": 2.710027100271003,
+      "eval_loss": 0.4322036802768707,
+      "eval_runtime": 154.5242,
+      "eval_samples_per_second": 36.603,
+      "eval_steps_per_second": 4.575,
+      "eval_wer": 0.3299738408948661,
+      "step": 14000
+    },
+    {
+      "epoch": 2.72938443670151,
+      "eval_loss": 0.4294193983078003,
+      "eval_runtime": 154.4329,
+      "eval_samples_per_second": 36.624,
+      "eval_steps_per_second": 4.578,
+      "eval_wer": 0.3261542905746979,
+      "step": 14100
+    },
+    {
+      "epoch": 2.748741773132017,
+      "eval_loss": 0.43099814653396606,
+      "eval_runtime": 154.4209,
+      "eval_samples_per_second": 36.627,
+      "eval_steps_per_second": 4.578,
+      "eval_wer": 0.32329765209995026,
+      "step": 14200
+    },
+    {
+      "epoch": 2.7680991095625243,
+      "eval_loss": 0.42700281739234924,
+      "eval_runtime": 155.4008,
+      "eval_samples_per_second": 36.396,
+      "eval_steps_per_second": 4.55,
+      "eval_wer": 0.3268122803357353,
+      "step": 14300
+    },
+    {
+      "epoch": 2.7874564459930316,
+      "eval_loss": 0.4209098219871521,
+      "eval_runtime": 156.5271,
+      "eval_samples_per_second": 36.134,
+      "eval_steps_per_second": 4.517,
+      "eval_wer": 0.3254321066906325,
+      "step": 14400
+    },
+    {
+      "epoch": 2.8068137824235384,
+      "grad_norm": 0.6974443793296814,
+      "learning_rate": 8.48153846153846e-05,
+      "loss": 0.3459,
+      "step": 14500
+    },
+    {
+      "epoch": 2.8068137824235384,
+      "eval_loss": 0.42542555928230286,
+      "eval_runtime": 157.9392,
+      "eval_samples_per_second": 35.811,
+      "eval_steps_per_second": 4.476,
+      "eval_wer": 0.32729373625844554,
+      "step": 14500
+    },
+    {
+      "epoch": 2.8261711188540457,
+      "eval_loss": 0.42783817648887634,
+      "eval_runtime": 155.0217,
+      "eval_samples_per_second": 36.485,
+      "eval_steps_per_second": 4.561,
+      "eval_wer": 0.3231532153231372,
+      "step": 14600
+    },
+    {
+      "epoch": 2.845528455284553,
+      "eval_loss": 0.4212438464164734,
+      "eval_runtime": 154.853,
+      "eval_samples_per_second": 36.525,
+      "eval_steps_per_second": 4.566,
+      "eval_wer": 0.3215002166551652,
+      "step": 14700
+    },
+    {
+      "epoch": 2.86488579171506,
+      "eval_loss": 0.4169256389141083,
+      "eval_runtime": 154.5142,
+      "eval_samples_per_second": 36.605,
+      "eval_steps_per_second": 4.576,
+      "eval_wer": 0.31928551941069794,
+      "step": 14800
+    },
+    {
+      "epoch": 2.884243128145567,
+      "eval_loss": 0.42132049798965454,
+      "eval_runtime": 154.8091,
+      "eval_samples_per_second": 36.535,
+      "eval_steps_per_second": 4.567,
+      "eval_wer": 0.3195262473720531,
+      "step": 14900
+    },
+    {
+      "epoch": 2.9036004645760745,
+      "grad_norm": 1.099702000617981,
+      "learning_rate": 7.713846153846152e-05,
+      "loss": 0.3483,
+      "step": 15000
+    },
+    {
+      "epoch": 2.9036004645760745,
+      "eval_loss": 0.41696369647979736,
+      "eval_runtime": 155.3223,
+      "eval_samples_per_second": 36.415,
+      "eval_steps_per_second": 4.552,
+      "eval_wer": 0.31652517212049236,
+      "step": 15000
+    },
+    {
+      "epoch": 2.9229578010065813,
+      "eval_loss": 0.41230952739715576,
+      "eval_runtime": 154.9181,
+      "eval_samples_per_second": 36.51,
+      "eval_steps_per_second": 4.564,
+      "eval_wer": 0.31418208662996905,
+      "step": 15100
+    },
+    {
+      "epoch": 2.9423151374370886,
+      "eval_loss": 0.4116990566253662,
+      "eval_runtime": 154.97,
+      "eval_samples_per_second": 36.497,
+      "eval_steps_per_second": 4.562,
+      "eval_wer": 0.31337966009211854,
+      "step": 15200
+    },
+    {
+      "epoch": 2.961672473867596,
+      "eval_loss": 0.410386323928833,
+      "eval_runtime": 155.0232,
+      "eval_samples_per_second": 36.485,
+      "eval_steps_per_second": 4.561,
+      "eval_wer": 0.31158222464733354,
+      "step": 15300
+    },
+    {
+      "epoch": 2.9810298102981028,
+      "eval_loss": 0.41244322061538696,
+      "eval_runtime": 154.4682,
+      "eval_samples_per_second": 36.616,
+      "eval_steps_per_second": 4.577,
+      "eval_wer": 0.31419813516072603,
+      "step": 15400
+    },
+    {
+      "epoch": 3.00038714672861,
+      "grad_norm": 0.725528359413147,
+      "learning_rate": 6.946153846153845e-05,
+      "loss": 0.3501,
+      "step": 15500
+    },
+    {
+      "epoch": 3.00038714672861,
+      "eval_loss": 0.40684688091278076,
+      "eval_runtime": 154.5863,
+      "eval_samples_per_second": 36.588,
+      "eval_steps_per_second": 4.573,
+      "eval_wer": 0.31272167033108117,
+      "step": 15500
+    },
+    {
+      "epoch": 3.0197444831591174,
+      "eval_loss": 0.4200752079486847,
+      "eval_runtime": 154.5821,
+      "eval_samples_per_second": 36.589,
+      "eval_steps_per_second": 4.574,
+      "eval_wer": 0.3087416347033429,
+      "step": 15600
+    },
+    {
+      "epoch": 3.0391018195896247,
+      "eval_loss": 0.4186869263648987,
+      "eval_runtime": 154.7417,
+      "eval_samples_per_second": 36.551,
+      "eval_steps_per_second": 4.569,
+      "eval_wer": 0.3137808733610438,
+      "step": 15700
+    },
+    {
+      "epoch": 3.0584591560201315,
+      "eval_loss": 0.41133585572242737,
+      "eval_runtime": 155.21,
+      "eval_samples_per_second": 36.441,
+      "eval_steps_per_second": 4.555,
+      "eval_wer": 0.31092423488629617,
+      "step": 15800
+    },
+    {
+      "epoch": 3.077816492450639,
+      "eval_loss": 0.4191639721393585,
+      "eval_runtime": 155.276,
+      "eval_samples_per_second": 36.425,
+      "eval_steps_per_second": 4.553,
+      "eval_wer": 0.30851695527274475,
+      "step": 15900
+    },
+    {
+      "epoch": 3.097173828881146,
+      "grad_norm": 0.5114701390266418,
+      "learning_rate": 6.176923076923076e-05,
+      "loss": 0.2754,
+      "step": 16000
+    },
+    {
+      "epoch": 3.097173828881146,
+      "eval_loss": 0.4161028265953064,
+      "eval_runtime": 154.597,
+      "eval_samples_per_second": 36.585,
+      "eval_steps_per_second": 4.573,
+      "eval_wer": 0.30901445972621205,
+      "step": 16000
+    },
+    {
+      "epoch": 3.116531165311653,
+      "eval_loss": 0.4183988571166992,
+      "eval_runtime": 155.0124,
+      "eval_samples_per_second": 36.487,
+      "eval_steps_per_second": 4.561,
+      "eval_wer": 0.307152830158399,
+      "step": 16100
+    },
+    {
+      "epoch": 3.1358885017421603,
+      "eval_loss": 0.4186756908893585,
+      "eval_runtime": 154.8535,
+      "eval_samples_per_second": 36.525,
+      "eval_steps_per_second": 4.566,
+      "eval_wer": 0.3060936271284364,
+      "step": 16200
+    },
+    {
+      "epoch": 3.1552458381726676,
+      "eval_loss": 0.4193824827671051,
+      "eval_runtime": 154.3195,
+      "eval_samples_per_second": 36.651,
+      "eval_steps_per_second": 4.581,
+      "eval_wer": 0.3059652388823803,
+      "step": 16300
+    },
+    {
+      "epoch": 3.1746031746031744,
+      "eval_loss": 0.40788766741752625,
+      "eval_runtime": 154.9673,
+      "eval_samples_per_second": 36.498,
+      "eval_steps_per_second": 4.562,
+      "eval_wer": 0.3038949784147261,
+      "step": 16400
+    },
+    {
+      "epoch": 3.1939605110336817,
+      "grad_norm": 0.5594165325164795,
+      "learning_rate": 5.4076923076923074e-05,
+      "loss": 0.2802,
+      "step": 16500
+    },
+    {
+      "epoch": 3.1939605110336817,
+      "eval_loss": 0.41461309790611267,
+      "eval_runtime": 154.8662,
+      "eval_samples_per_second": 36.522,
+      "eval_steps_per_second": 4.565,
+      "eval_wer": 0.30424804609138034,
+      "step": 16500
+    },
+    {
+      "epoch": 3.213317847464189,
+      "eval_loss": 0.4168522357940674,
+      "eval_runtime": 155.0374,
+      "eval_samples_per_second": 36.482,
+      "eval_steps_per_second": 4.56,
+      "eval_wer": 0.30116672818603457,
+      "step": 16600
+    },
+    {
+      "epoch": 3.2326751838946963,
+      "eval_loss": 0.40926745533943176,
+      "eval_runtime": 154.7423,
+      "eval_samples_per_second": 36.551,
+      "eval_steps_per_second": 4.569,
+      "eval_wer": 0.3023864165235673,
+      "step": 16700
+    },
+    {
+      "epoch": 3.252032520325203,
+      "eval_loss": 0.4115259051322937,
+      "eval_runtime": 154.7933,
+      "eval_samples_per_second": 36.539,
+      "eval_steps_per_second": 4.567,
+      "eval_wer": 0.3005408354865112,
+      "step": 16800
+    },
+    {
+      "epoch": 3.2713898567557105,
+      "eval_loss": 0.40197211503982544,
+      "eval_runtime": 155.5964,
+      "eval_samples_per_second": 36.35,
+      "eval_steps_per_second": 4.544,
+      "eval_wer": 0.30410360931456726,
+      "step": 16900
+    },
+    {
+      "epoch": 3.290747193186218,
+      "grad_norm": 1.4730154275894165,
+      "learning_rate": 4.6384615384615385e-05,
+      "loss": 0.2723,
+      "step": 17000
+    },
+    {
+      "epoch": 3.290747193186218,
+      "eval_loss": 0.4058869779109955,
+      "eval_runtime": 155.0898,
+      "eval_samples_per_second": 36.469,
+      "eval_steps_per_second": 4.559,
+      "eval_wer": 0.30442457992970745,
+      "step": 17000
+    },
+    {
+      "epoch": 3.3101045296167246,
+      "eval_loss": 0.40676185488700867,
+      "eval_runtime": 155.0576,
+      "eval_samples_per_second": 36.477,
+      "eval_steps_per_second": 4.56,
+      "eval_wer": 0.3013753590858757,
+      "step": 17100
+    },
+    {
+      "epoch": 3.329461866047232,
+      "eval_loss": 0.40653425455093384,
+      "eval_runtime": 155.8377,
+      "eval_samples_per_second": 36.294,
+      "eval_steps_per_second": 4.537,
+      "eval_wer": 0.30878978029561394,
+      "step": 17200
+    },
+    {
+      "epoch": 3.3488192024777392,
+      "eval_loss": 0.4082197844982147,
+      "eval_runtime": 155.7924,
+      "eval_samples_per_second": 36.305,
+      "eval_steps_per_second": 4.538,
+      "eval_wer": 0.3010543884707355,
+      "step": 17300
+    },
+    {
+      "epoch": 3.368176538908246,
+      "eval_loss": 0.4083554446697235,
+      "eval_runtime": 155.6775,
+      "eval_samples_per_second": 36.332,
+      "eval_steps_per_second": 4.541,
+      "eval_wer": 0.3007494663863523,
+      "step": 17400
+    },
+    {
+      "epoch": 3.3875338753387534,
+      "grad_norm": 0.5211097598075867,
+      "learning_rate": 3.87076923076923e-05,
+      "loss": 0.2557,
+      "step": 17500
+    },
+    {
+      "epoch": 3.3875338753387534,
+      "eval_loss": 0.4009736180305481,
+      "eval_runtime": 155.105,
+      "eval_samples_per_second": 36.466,
+      "eval_steps_per_second": 4.558,
+      "eval_wer": 0.29924090449519347,
+      "step": 17500
+    },
+    {
+      "epoch": 3.4068912117692607,
+      "eval_loss": 0.4061805009841919,
+      "eval_runtime": 154.8792,
+      "eval_samples_per_second": 36.519,
+      "eval_steps_per_second": 4.565,
+      "eval_wer": 0.2999951854407729,
+      "step": 17600
+    },
+    {
+      "epoch": 3.4262485481997675,
+      "eval_loss": 0.40264037251472473,
+      "eval_runtime": 155.9957,
+      "eval_samples_per_second": 36.257,
+      "eval_steps_per_second": 4.532,
+      "eval_wer": 0.2980533132191748,
+      "step": 17700
+    },
+    {
+      "epoch": 3.445605884630275,
+      "eval_loss": 0.40035372972488403,
+      "eval_runtime": 155.0928,
+      "eval_samples_per_second": 36.468,
+      "eval_steps_per_second": 4.559,
+      "eval_wer": 0.29893598241081026,
+      "step": 17800
+    },
+    {
+      "epoch": 3.464963221060782,
+      "eval_loss": 0.40443336963653564,
+      "eval_runtime": 154.9305,
+      "eval_samples_per_second": 36.507,
+      "eval_steps_per_second": 4.563,
+      "eval_wer": 0.29906437065686636,
+      "step": 17900
+    },
+    {
+      "epoch": 3.484320557491289,
+      "grad_norm": 0.7458967566490173,
+      "learning_rate": 3.101538461538461e-05,
+      "loss": 0.2578,
+      "step": 18000
+    },
+    {
+      "epoch": 3.484320557491289,
+      "eval_loss": 0.4003549814224243,
+      "eval_runtime": 155.7394,
+      "eval_samples_per_second": 36.317,
+      "eval_steps_per_second": 4.54,
+      "eval_wer": 0.29660894545104394,
+      "step": 18000
+    },
+    {
+      "epoch": 3.5036778939217963,
+      "eval_loss": 0.40592488646507263,
+      "eval_runtime": 159.2644,
+      "eval_samples_per_second": 35.513,
+      "eval_steps_per_second": 4.439,
+      "eval_wer": 0.29449053939111874,
+      "step": 18100
+    },
+    {
+      "epoch": 3.5230352303523036,
+      "eval_loss": 0.4014962613582611,
+      "eval_runtime": 155.6654,
+      "eval_samples_per_second": 36.334,
+      "eval_steps_per_second": 4.542,
+      "eval_wer": 0.29632007189741777,
+      "step": 18200
+    },
+    {
+      "epoch": 3.5423925667828104,
+      "eval_loss": 0.396659791469574,
+      "eval_runtime": 156.1536,
+      "eval_samples_per_second": 36.221,
+      "eval_steps_per_second": 4.528,
+      "eval_wer": 0.29585466450546455,
+      "step": 18300
+    },
+    {
+      "epoch": 3.5617499032133177,
+      "eval_loss": 0.4001907706260681,
+      "eval_runtime": 155.7578,
+      "eval_samples_per_second": 36.313,
+      "eval_steps_per_second": 4.539,
+      "eval_wer": 0.29412142318370754,
+      "step": 18400
+    },
+    {
+      "epoch": 3.581107239643825,
+      "grad_norm": 0.6122294664382935,
+      "learning_rate": 2.3338461538461535e-05,
+      "loss": 0.2508,
+      "step": 18500
+    },
+    {
+      "epoch": 3.581107239643825,
+      "eval_loss": 0.39826107025146484,
+      "eval_runtime": 155.467,
+      "eval_samples_per_second": 36.381,
+      "eval_steps_per_second": 4.548,
+      "eval_wer": 0.2945547335141468,
+      "step": 18500
+    },
+    {
+      "epoch": 3.600464576074332,
+      "eval_loss": 0.3958674967288971,
+      "eval_runtime": 155.8242,
+      "eval_samples_per_second": 36.297,
+      "eval_steps_per_second": 4.537,
+      "eval_wer": 0.29365601579175427,
+      "step": 18600
+    },
+    {
+      "epoch": 3.619821912504839,
+      "eval_loss": 0.3970955014228821,
+      "eval_runtime": 155.4329,
+      "eval_samples_per_second": 36.389,
+      "eval_steps_per_second": 4.549,
+      "eval_wer": 0.2942016658374926,
+      "step": 18700
+    },
+    {
+      "epoch": 3.6391792489353465,
+      "eval_loss": 0.3906669616699219,
+      "eval_runtime": 155.4929,
+      "eval_samples_per_second": 36.375,
+      "eval_steps_per_second": 4.547,
+      "eval_wer": 0.2923239877389225,
+      "step": 18800
+    },
+    {
+      "epoch": 3.658536585365854,
+      "eval_loss": 0.39506247639656067,
+      "eval_runtime": 155.5246,
+      "eval_samples_per_second": 36.367,
+      "eval_steps_per_second": 4.546,
+      "eval_wer": 0.2903981640480814,
+      "step": 18900
+    },
+    {
+      "epoch": 3.6778939217963607,
+      "grad_norm": 0.33715635538101196,
+      "learning_rate": 1.5646153846153846e-05,
+      "loss": 0.2659,
+      "step": 19000
+    },
+    {
+      "epoch": 3.6778939217963607,
+      "eval_loss": 0.3892674744129181,
+      "eval_runtime": 155.5533,
+      "eval_samples_per_second": 36.361,
+      "eval_steps_per_second": 4.545,
+      "eval_wer": 0.29309431721525897,
+      "step": 19000
+    },
+    {
+      "epoch": 3.697251258226868,
+      "eval_loss": 0.39077267050743103,
+      "eval_runtime": 155.448,
+      "eval_samples_per_second": 36.385,
+      "eval_steps_per_second": 4.548,
+      "eval_wer": 0.2900771934329412,
+      "step": 19100
+    },
+    {
+      "epoch": 3.7166085946573753,
+      "eval_loss": 0.39407432079315186,
+      "eval_runtime": 155.4696,
+      "eval_samples_per_second": 36.38,
+      "eval_steps_per_second": 4.548,
+      "eval_wer": 0.2884241947649693,
+      "step": 19200
+    },
+    {
+      "epoch": 3.7359659310878826,
+      "eval_loss": 0.3924821615219116,
+      "eval_runtime": 155.4791,
+      "eval_samples_per_second": 36.378,
+      "eval_steps_per_second": 4.547,
+      "eval_wer": 0.2890019418722216,
+      "step": 19300
+    },
+    {
+      "epoch": 3.7553232675183894,
+      "eval_loss": 0.3916691243648529,
+      "eval_runtime": 155.9516,
+      "eval_samples_per_second": 36.268,
+      "eval_steps_per_second": 4.533,
+      "eval_wer": 0.2892908154258478,
+      "step": 19400
+    },
+    {
+      "epoch": 3.7746806039488967,
+      "grad_norm": 0.4647356867790222,
+      "learning_rate": 7.953846153846153e-06,
+      "loss": 0.2488,
+      "step": 19500
+    },
+    {
+      "epoch": 3.7746806039488967,
+      "eval_loss": 0.39043277502059937,
+      "eval_runtime": 155.3552,
+      "eval_samples_per_second": 36.407,
+      "eval_steps_per_second": 4.551,
+      "eval_wer": 0.2884562918264833,
+      "step": 19500
+    },
+    {
+      "epoch": 3.794037940379404,
+      "eval_loss": 0.39014604687690735,
+      "eval_runtime": 155.2137,
+      "eval_samples_per_second": 36.44,
+      "eval_steps_per_second": 4.555,
+      "eval_wer": 0.2887933109723805,
+      "step": 19600
+    },
+    {
+      "epoch": 3.813395276809911,
+      "eval_loss": 0.3883425295352936,
+      "eval_runtime": 155.5369,
+      "eval_samples_per_second": 36.364,
+      "eval_steps_per_second": 4.546,
+      "eval_wer": 0.28922662130281973,
+      "step": 19700
+    },
+    {
+      "epoch": 3.832752613240418,
+      "eval_loss": 0.38913780450820923,
+      "eval_runtime": 155.8958,
+      "eval_samples_per_second": 36.281,
+      "eval_steps_per_second": 4.535,
+      "eval_wer": 0.28903403893373564,
+      "step": 19800
+    },
+    {
+      "epoch": 3.8521099496709255,
+      "eval_loss": 0.3888201415538788,
+      "eval_runtime": 155.6372,
+      "eval_samples_per_second": 36.341,
+      "eval_steps_per_second": 4.543,
+      "eval_wer": 0.2888254080338945,
+      "step": 19900
+    },
+    {
+      "epoch": 3.8714672861014323,
+      "grad_norm": 0.3741956055164337,
+      "learning_rate": 2.615384615384615e-07,
+      "loss": 0.2602,
+      "step": 20000
+    },
+    {
+      "epoch": 3.8714672861014323,
+      "eval_loss": 0.38884833455085754,
+      "eval_runtime": 155.0772,
+      "eval_samples_per_second": 36.472,
+      "eval_steps_per_second": 4.559,
+      "eval_wer": 0.2884883888879973,
+      "step": 20000
+    },
+    {
+      "epoch": 3.8714672861014323,
+      "step": 20000,
+      "total_flos": 2.249387574100498e+19,
+      "train_loss": 0.15996522521972656,
+      "train_runtime": 19346.8732,
+      "train_samples_per_second": 8.27,
+      "train_steps_per_second": 1.034
     }
   ],
   "logging_steps": 500,
-  "max_steps": 10000,
+  "max_steps": 20000,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 2,
+  "num_train_epochs": 4,
   "save_steps": 400,
-  "total_flos": 1.1256772526671323e+19,
+  "total_flos": 2.249387574100498e+19,
   "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null