End of training

README.md

@@ -16,7 +16,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # whisper-large-v3-ft-btbn-ca
 
-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/banc-trawsgrifiadau-bangor train main, cymen-arfor/15awr train+dev main dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.4127
 - Wer: 0.2775

all_results.json

@@ -1,15 +1,15 @@
 {
-    "epoch": 3.1938677738741617,
-    "eval_loss": 0.4074931740760803,
-    "eval_runtime": 1638.5657,
+    "epoch": 3.2776138970829236,
+    "eval_loss": 0.41265004873275757,
+    "eval_runtime": 1622.4,
     "eval_samples": 3901,
-    "eval_samples_per_second": 2.381,
-    "eval_steps_per_second": 0.149,
-    "eval_wer": 0.2823846789481359,
+    "eval_samples_per_second": 2.404,
+    "eval_steps_per_second": 0.15,
+    "eval_wer": 0.2775239391435376,
     "total_flos": 5.435895365546803e+20,
-    "train_loss": 0.7105956988334656,
-    "train_runtime": 59674.9515,
-    "train_samples": 50095,
-    "train_samples_per_second": 2.681,
-    "train_steps_per_second": 0.084
+    "train_loss": 0.7020168427467346,
+    "train_runtime": 60500.8955,
+    "train_samples": 48815,
+    "train_samples_per_second": 2.645,
+    "train_steps_per_second": 0.083
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 3.1938677738741617,
-    "eval_loss": 0.4074931740760803,
-    "eval_runtime": 1638.5657,
+    "epoch": 3.2776138970829236,
+    "eval_loss": 0.41265004873275757,
+    "eval_runtime": 1622.4,
     "eval_samples": 3901,
-    "eval_samples_per_second": 2.381,
-    "eval_steps_per_second": 0.149,
-    "eval_wer": 0.2823846789481359
+    "eval_samples_per_second": 2.404,
+    "eval_steps_per_second": 0.15,
+    "eval_wer": 0.2775239391435376
 }

runs/Nov05_10-53-57_4b49612c0fe9/events.out.tfevents.1730867197.4b49612c0fe9.120.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fc93920f892f94adc6a0dff156b3517607841e74b6ad78f7db93b3119777e16
+size 406

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 3.1938677738741617,
+    "epoch": 3.2776138970829236,
     "total_flos": 5.435895365546803e+20,
-    "train_loss": 0.7105956988334656,
-    "train_runtime": 59674.9515,
-    "train_samples": 50095,
-    "train_samples_per_second": 2.681,
-    "train_steps_per_second": 0.084
+    "train_loss": 0.7020168427467346,
+    "train_runtime": 60500.8955,
+    "train_samples": 48815,
+    "train_samples_per_second": 2.645,
+    "train_steps_per_second": 0.083
 }

trainer_state.json

@@ -1,7 +1,7 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 3.1938677738741617,
+  "epoch": 3.2776138970829236,
   "eval_steps": 1000,
   "global_step": 5000,
   "is_hyper_param_search": false,
@@ -9,1458 +9,1458 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.015969338869370808,
-      "grad_norm": 17.92350196838379,
+      "epoch": 0.01638806948541462,
+      "grad_norm": 16.339120864868164,
       "learning_rate": 4.800000000000001e-07,
-      "loss": 3.223,
+      "loss": 3.195,
       "step": 25
     },
     {
-      "epoch": 0.031938677738741615,
-      "grad_norm": 11.101584434509277,
+      "epoch": 0.03277613897082924,
+      "grad_norm": 11.416489601135254,
       "learning_rate": 9.800000000000001e-07,
-      "loss": 2.5321,
+      "loss": 2.5931,
       "step": 50
     },
     {
-      "epoch": 0.047908016608112426,
-      "grad_norm": 9.936028480529785,
+      "epoch": 0.049164208456243856,
+      "grad_norm": 10.20141887664795,
       "learning_rate": 1.48e-06,
-      "loss": 2.0032,
+      "loss": 2.0124,
       "step": 75
     },
     {
-      "epoch": 0.06387735547748323,
-      "grad_norm": 10.770648956298828,
+      "epoch": 0.06555227794165848,
+      "grad_norm": 10.408208847045898,
       "learning_rate": 1.98e-06,
-      "loss": 1.6134,
+      "loss": 1.6783,
       "step": 100
     },
     {
-      "epoch": 0.07984669434685404,
-      "grad_norm": 10.68463134765625,
+      "epoch": 0.0819403474270731,
+      "grad_norm": 10.501668930053711,
       "learning_rate": 2.4800000000000004e-06,
-      "loss": 1.5694,
+      "loss": 1.6138,
       "step": 125
     },
     {
-      "epoch": 0.09581603321622485,
-      "grad_norm": 10.635124206542969,
+      "epoch": 0.09832841691248771,
+      "grad_norm": 9.28870964050293,
       "learning_rate": 2.9800000000000003e-06,
-      "loss": 1.4934,
+      "loss": 1.5721,
       "step": 150
     },
     {
-      "epoch": 0.11178537208559565,
-      "grad_norm": 9.225995063781738,
+      "epoch": 0.11471648639790233,
+      "grad_norm": 10.96129322052002,
       "learning_rate": 3.48e-06,
-      "loss": 1.518,
+      "loss": 1.4486,
       "step": 175
     },
     {
-      "epoch": 0.12775471095496646,
-      "grad_norm": 9.05696964263916,
+      "epoch": 0.13110455588331696,
+      "grad_norm": 9.153674125671387,
       "learning_rate": 3.980000000000001e-06,
-      "loss": 1.4258,
+      "loss": 1.3586,
       "step": 200
     },
     {
-      "epoch": 0.14372404982433729,
-      "grad_norm": 9.599803924560547,
+      "epoch": 0.14749262536873156,
+      "grad_norm": 9.216712951660156,
       "learning_rate": 4.48e-06,
-      "loss": 1.422,
+      "loss": 1.3824,
       "step": 225
     },
     {
-      "epoch": 0.15969338869370808,
-      "grad_norm": 9.442488670349121,
+      "epoch": 0.1638806948541462,
+      "grad_norm": 9.344610214233398,
       "learning_rate": 4.980000000000001e-06,
-      "loss": 1.3208,
+      "loss": 1.3523,
       "step": 250
     },
     {
-      "epoch": 0.17566272756307888,
-      "grad_norm": 10.746581077575684,
+      "epoch": 0.1802687643395608,
+      "grad_norm": 9.565411567687988,
       "learning_rate": 5.480000000000001e-06,
-      "loss": 1.3997,
+      "loss": 1.3112,
       "step": 275
     },
     {
-      "epoch": 0.1916320664324497,
-      "grad_norm": 9.297645568847656,
+      "epoch": 0.19665683382497542,
+      "grad_norm": 9.21192455291748,
       "learning_rate": 5.98e-06,
-      "loss": 1.3343,
+      "loss": 1.2953,
       "step": 300
     },
     {
-      "epoch": 0.2076014053018205,
-      "grad_norm": 10.026762962341309,
+      "epoch": 0.21304490331039003,
+      "grad_norm": 9.407176971435547,
       "learning_rate": 6.480000000000001e-06,
-      "loss": 1.3353,
+      "loss": 1.297,
       "step": 325
     },
     {
-      "epoch": 0.2235707441711913,
-      "grad_norm": 9.392842292785645,
+      "epoch": 0.22943297279580466,
+      "grad_norm": 11.401097297668457,
       "learning_rate": 6.98e-06,
-      "loss": 1.3451,
+      "loss": 1.2525,
       "step": 350
     },
     {
-      "epoch": 0.23954008304056212,
-      "grad_norm": 7.376865386962891,
+      "epoch": 0.24582104228121926,
+      "grad_norm": 7.8178181648254395,
       "learning_rate": 7.48e-06,
-      "loss": 1.2645,
+      "loss": 1.2158,
       "step": 375
     },
     {
-      "epoch": 0.2555094219099329,
-      "grad_norm": 10.001627922058105,
+      "epoch": 0.2622091117666339,
+      "grad_norm": 10.073293685913086,
       "learning_rate": 7.980000000000002e-06,
-      "loss": 1.271,
+      "loss": 1.2654,
       "step": 400
     },
     {
-      "epoch": 0.2714787607793037,
-      "grad_norm": 7.699875354766846,
+      "epoch": 0.2785971812520485,
+      "grad_norm": 9.110590934753418,
       "learning_rate": 8.48e-06,
-      "loss": 1.2116,
+      "loss": 1.2629,
       "step": 425
     },
     {
-      "epoch": 0.28744809964867457,
-      "grad_norm": 8.733208656311035,
+      "epoch": 0.2949852507374631,
+      "grad_norm": 9.107154846191406,
       "learning_rate": 8.98e-06,
-      "loss": 1.2222,
+      "loss": 1.2484,
       "step": 450
     },
     {
-      "epoch": 0.30341743851804537,
-      "grad_norm": 8.286600112915039,
+      "epoch": 0.3113733202228777,
+      "grad_norm": 8.568262100219727,
       "learning_rate": 9.48e-06,
-      "loss": 1.2134,
+      "loss": 1.2081,
       "step": 475
     },
     {
-      "epoch": 0.31938677738741617,
-      "grad_norm": 7.408142566680908,
+      "epoch": 0.3277613897082924,
+      "grad_norm": 7.736518383026123,
       "learning_rate": 9.980000000000001e-06,
-      "loss": 1.1662,
+      "loss": 1.234,
       "step": 500
     },
     {
-      "epoch": 0.33535611625678696,
-      "grad_norm": 9.274364471435547,
+      "epoch": 0.344149459193707,
+      "grad_norm": 8.957286834716797,
       "learning_rate": 9.946666666666667e-06,
-      "loss": 1.2752,
+      "loss": 1.1519,
       "step": 525
     },
     {
-      "epoch": 0.35132545512615776,
-      "grad_norm": 8.4090576171875,
+      "epoch": 0.3605375286791216,
+      "grad_norm": 8.355584144592285,
       "learning_rate": 9.891111111111113e-06,
-      "loss": 1.2137,
+      "loss": 1.1517,
       "step": 550
     },
     {
-      "epoch": 0.3672947939955286,
-      "grad_norm": 8.74751091003418,
+      "epoch": 0.3769255981645362,
+      "grad_norm": 6.338685512542725,
       "learning_rate": 9.835555555555556e-06,
-      "loss": 1.1794,
+      "loss": 1.1563,
       "step": 575
     },
     {
-      "epoch": 0.3832641328648994,
-      "grad_norm": 7.037300109863281,
+      "epoch": 0.39331366764995085,
+      "grad_norm": 8.323567390441895,
       "learning_rate": 9.780000000000001e-06,
-      "loss": 1.1806,
+      "loss": 1.1948,
       "step": 600
     },
     {
-      "epoch": 0.3992334717342702,
-      "grad_norm": 8.758708953857422,
+      "epoch": 0.40970173713536545,
+      "grad_norm": 8.510113716125488,
       "learning_rate": 9.724444444444445e-06,
-      "loss": 1.0997,
+      "loss": 1.1357,
       "step": 625
     },
     {
-      "epoch": 0.415202810603641,
-      "grad_norm": 7.583318710327148,
+      "epoch": 0.42608980662078005,
+      "grad_norm": 8.770384788513184,
       "learning_rate": 9.66888888888889e-06,
-      "loss": 1.1122,
+      "loss": 1.1083,
       "step": 650
     },
     {
-      "epoch": 0.4311721494730118,
-      "grad_norm": 7.173635959625244,
+      "epoch": 0.4424778761061947,
+      "grad_norm": 6.591556072235107,
       "learning_rate": 9.613333333333335e-06,
-      "loss": 1.1167,
+      "loss": 1.1361,
       "step": 675
     },
     {
-      "epoch": 0.4471414883423826,
-      "grad_norm": 6.702845096588135,
+      "epoch": 0.4588659455916093,
+      "grad_norm": 9.191658973693848,
       "learning_rate": 9.557777777777777e-06,
-      "loss": 1.0873,
+      "loss": 1.1053,
       "step": 700
     },
     {
-      "epoch": 0.46311082721175345,
-      "grad_norm": 7.966395854949951,
+      "epoch": 0.4752540150770239,
+      "grad_norm": 7.101362705230713,
       "learning_rate": 9.502222222222223e-06,
-      "loss": 1.0764,
+      "loss": 1.1253,
       "step": 725
     },
     {
-      "epoch": 0.47908016608112425,
-      "grad_norm": 7.768787860870361,
+      "epoch": 0.4916420845624385,
+      "grad_norm": 7.415987491607666,
       "learning_rate": 9.446666666666667e-06,
-      "loss": 1.0985,
+      "loss": 1.0605,
       "step": 750
     },
     {
-      "epoch": 0.49504950495049505,
-      "grad_norm": 8.754780769348145,
+      "epoch": 0.5080301540478531,
+      "grad_norm": 7.0226616859436035,
       "learning_rate": 9.391111111111111e-06,
-      "loss": 1.0646,
+      "loss": 1.0165,
       "step": 775
     },
     {
-      "epoch": 0.5110188438198658,
-      "grad_norm": 7.156813144683838,
+      "epoch": 0.5244182235332678,
+      "grad_norm": 8.283411026000977,
       "learning_rate": 9.335555555555557e-06,
-      "loss": 1.0149,
+      "loss": 1.0527,
       "step": 800
     },
     {
-      "epoch": 0.5269881826892366,
-      "grad_norm": 6.864979267120361,
+      "epoch": 0.5408062930186824,
+      "grad_norm": 7.523382186889648,
       "learning_rate": 9.280000000000001e-06,
-      "loss": 1.0514,
+      "loss": 1.0575,
       "step": 825
     },
     {
-      "epoch": 0.5429575215586074,
-      "grad_norm": 6.696143627166748,
+      "epoch": 0.557194362504097,
+      "grad_norm": 8.657249450683594,
       "learning_rate": 9.224444444444445e-06,
-      "loss": 1.0376,
+      "loss": 1.0674,
       "step": 850
     },
     {
-      "epoch": 0.5589268604279782,
-      "grad_norm": 6.3973469734191895,
+      "epoch": 0.5735824319895116,
+      "grad_norm": 6.097987174987793,
       "learning_rate": 9.168888888888889e-06,
-      "loss": 0.992,
+      "loss": 0.9898,
       "step": 875
     },
     {
-      "epoch": 0.5748961992973491,
-      "grad_norm": 6.48210334777832,
+      "epoch": 0.5899705014749262,
+      "grad_norm": 8.815755844116211,
       "learning_rate": 9.113333333333335e-06,
-      "loss": 1.0463,
+      "loss": 1.0372,
       "step": 900
     },
     {
-      "epoch": 0.5908655381667199,
-      "grad_norm": 7.711927890777588,
+      "epoch": 0.6063585709603408,
+      "grad_norm": 6.628593921661377,
       "learning_rate": 9.057777777777779e-06,
-      "loss": 1.0023,
+      "loss": 1.052,
       "step": 925
     },
     {
-      "epoch": 0.6068348770360907,
-      "grad_norm": 7.2077250480651855,
+      "epoch": 0.6227466404457555,
+      "grad_norm": 7.880568027496338,
       "learning_rate": 9.002222222222223e-06,
-      "loss": 1.0283,
+      "loss": 1.0182,
       "step": 950
     },
     {
-      "epoch": 0.6228042159054615,
-      "grad_norm": 7.416881561279297,
+      "epoch": 0.63913470993117,
+      "grad_norm": 7.334752559661865,
       "learning_rate": 8.946666666666669e-06,
-      "loss": 1.0249,
+      "loss": 1.0596,
       "step": 975
     },
     {
-      "epoch": 0.6387735547748323,
-      "grad_norm": 6.398010730743408,
+      "epoch": 0.6555227794165848,
+      "grad_norm": 6.15729284286499,
       "learning_rate": 8.891111111111111e-06,
-      "loss": 0.9546,
+      "loss": 0.9964,
       "step": 1000
     },
     {
-      "epoch": 0.6387735547748323,
-      "eval_loss": 0.4855521023273468,
-      "eval_runtime": 1667.9567,
-      "eval_samples_per_second": 2.339,
-      "eval_steps_per_second": 0.146,
-      "eval_wer": 0.3615418266660186,
+      "epoch": 0.6555227794165848,
+      "eval_loss": 0.4889208674430847,
+      "eval_runtime": 1659.2481,
+      "eval_samples_per_second": 2.351,
+      "eval_steps_per_second": 0.147,
+      "eval_wer": 0.36632965537354784,
       "step": 1000
     },
     {
-      "epoch": 0.6547428936442031,
-      "grad_norm": 5.443120956420898,
+      "epoch": 0.6719108489019994,
+      "grad_norm": 7.4670538902282715,
       "learning_rate": 8.835555555555557e-06,
-      "loss": 1.013,
+      "loss": 1.0353,
       "step": 1025
     },
     {
-      "epoch": 0.6707122325135739,
-      "grad_norm": 7.361939907073975,
+      "epoch": 0.688298918387414,
+      "grad_norm": 6.965098857879639,
       "learning_rate": 8.78e-06,
-      "loss": 0.946,
+      "loss": 0.9819,
       "step": 1050
     },
     {
-      "epoch": 0.6866815713829447,
-      "grad_norm": 6.80987024307251,
+      "epoch": 0.7046869878728286,
+      "grad_norm": 6.953728675842285,
       "learning_rate": 8.724444444444445e-06,
-      "loss": 0.9945,
+      "loss": 0.9931,
       "step": 1075
     },
     {
-      "epoch": 0.7026509102523155,
-      "grad_norm": 6.040420055389404,
+      "epoch": 0.7210750573582432,
+      "grad_norm": 7.410843849182129,
       "learning_rate": 8.66888888888889e-06,
-      "loss": 1.0078,
+      "loss": 1.0368,
       "step": 1100
     },
     {
-      "epoch": 0.7186202491216863,
-      "grad_norm": 7.904597282409668,
+      "epoch": 0.7374631268436578,
+      "grad_norm": 6.275593280792236,
       "learning_rate": 8.613333333333333e-06,
-      "loss": 0.9707,
+      "loss": 0.974,
       "step": 1125
     },
     {
-      "epoch": 0.7345895879910572,
-      "grad_norm": 6.9557204246521,
+      "epoch": 0.7538511963290724,
+      "grad_norm": 5.739853382110596,
       "learning_rate": 8.557777777777778e-06,
-      "loss": 0.9712,
+      "loss": 0.9568,
       "step": 1150
     },
     {
-      "epoch": 0.750558926860428,
-      "grad_norm": 5.4510178565979,
+      "epoch": 0.7702392658144871,
+      "grad_norm": 7.921228408813477,
       "learning_rate": 8.502222222222223e-06,
-      "loss": 0.9659,
+      "loss": 0.9109,
       "step": 1175
     },
     {
-      "epoch": 0.7665282657297988,
-      "grad_norm": 5.72000789642334,
+      "epoch": 0.7866273352999017,
+      "grad_norm": 6.905638694763184,
       "learning_rate": 8.446666666666668e-06,
-      "loss": 0.9368,
+      "loss": 0.9741,
       "step": 1200
     },
     {
-      "epoch": 0.7824976045991696,
-      "grad_norm": 5.816184043884277,
+      "epoch": 0.8030154047853163,
+      "grad_norm": 7.002967834472656,
       "learning_rate": 8.391111111111112e-06,
-      "loss": 0.9089,
+      "loss": 0.9512,
       "step": 1225
     },
     {
-      "epoch": 0.7984669434685404,
-      "grad_norm": 6.033445835113525,
+      "epoch": 0.8194034742707309,
+      "grad_norm": 8.190713882446289,
       "learning_rate": 8.335555555555556e-06,
-      "loss": 0.9502,
+      "loss": 0.9474,
       "step": 1250
     },
     {
-      "epoch": 0.8144362823379112,
-      "grad_norm": 5.273908615112305,
+      "epoch": 0.8357915437561455,
+      "grad_norm": 6.305277347564697,
       "learning_rate": 8.28e-06,
-      "loss": 0.9151,
+      "loss": 0.9251,
       "step": 1275
     },
     {
-      "epoch": 0.830405621207282,
-      "grad_norm": 6.122651100158691,
+      "epoch": 0.8521796132415601,
+      "grad_norm": 6.228824615478516,
       "learning_rate": 8.224444444444444e-06,
-      "loss": 0.9517,
+      "loss": 0.9186,
       "step": 1300
     },
     {
-      "epoch": 0.8463749600766528,
-      "grad_norm": 6.673498630523682,
+      "epoch": 0.8685676827269747,
+      "grad_norm": 7.2379679679870605,
       "learning_rate": 8.16888888888889e-06,
-      "loss": 1.0318,
+      "loss": 0.9062,
       "step": 1325
     },
     {
-      "epoch": 0.8623442989460236,
-      "grad_norm": 7.197372913360596,
+      "epoch": 0.8849557522123894,
+      "grad_norm": 5.009719371795654,
       "learning_rate": 8.113333333333334e-06,
-      "loss": 0.937,
+      "loss": 0.9369,
       "step": 1350
     },
     {
-      "epoch": 0.8783136378153944,
-      "grad_norm": 6.718713283538818,
+      "epoch": 0.901343821697804,
+      "grad_norm": 6.123809814453125,
       "learning_rate": 8.057777777777778e-06,
-      "loss": 0.9325,
+      "loss": 0.9473,
       "step": 1375
     },
     {
-      "epoch": 0.8942829766847652,
-      "grad_norm": 7.090008735656738,
+      "epoch": 0.9177318911832186,
+      "grad_norm": 8.46691608428955,
       "learning_rate": 8.002222222222222e-06,
-      "loss": 0.9044,
+      "loss": 0.9325,
       "step": 1400
     },
     {
-      "epoch": 0.9102523155541361,
-      "grad_norm": 8.890588760375977,
+      "epoch": 0.9341199606686332,
+      "grad_norm": 6.441035747528076,
       "learning_rate": 7.946666666666666e-06,
-      "loss": 0.8935,
+      "loss": 0.9194,
       "step": 1425
     },
     {
-      "epoch": 0.9262216544235069,
-      "grad_norm": 5.521778583526611,
+      "epoch": 0.9505080301540478,
+      "grad_norm": 6.163531303405762,
       "learning_rate": 7.891111111111112e-06,
-      "loss": 0.8857,
+      "loss": 0.8699,
       "step": 1450
     },
     {
-      "epoch": 0.9421909932928777,
-      "grad_norm": 6.829771041870117,
+      "epoch": 0.9668960996394624,
+      "grad_norm": 6.857112407684326,
       "learning_rate": 7.835555555555556e-06,
-      "loss": 0.9411,
+      "loss": 0.8956,
       "step": 1475
     },
     {
-      "epoch": 0.9581603321622485,
-      "grad_norm": 6.48971700668335,
+      "epoch": 0.983284169124877,
+      "grad_norm": 6.297347068786621,
       "learning_rate": 7.78e-06,
-      "loss": 0.8616,
+      "loss": 0.9163,
       "step": 1500
     },
     {
-      "epoch": 0.9741296710316193,
-      "grad_norm": 5.809508323669434,
+      "epoch": 0.9996722386102918,
+      "grad_norm": 6.199060440063477,
       "learning_rate": 7.724444444444446e-06,
-      "loss": 0.8976,
+      "loss": 0.8492,
       "step": 1525
     },
     {
-      "epoch": 0.9900990099009901,
-      "grad_norm": 6.918030738830566,
+      "epoch": 1.0160603080957062,
+      "grad_norm": 5.396669864654541,
       "learning_rate": 7.66888888888889e-06,
-      "loss": 0.9213,
+      "loss": 0.669,
       "step": 1550
     },
     {
-      "epoch": 1.006068348770361,
-      "grad_norm": 5.847545623779297,
+      "epoch": 1.0324483775811208,
+      "grad_norm": 4.675485134124756,
       "learning_rate": 7.613333333333334e-06,
-      "loss": 0.813,
+      "loss": 0.6436,
       "step": 1575
     },
     {
-      "epoch": 1.0220376876397317,
-      "grad_norm": 5.331517219543457,
+      "epoch": 1.0488364470665354,
+      "grad_norm": 4.590639591217041,
       "learning_rate": 7.557777777777779e-06,
-      "loss": 0.6469,
+      "loss": 0.656,
       "step": 1600
     },
     {
-      "epoch": 1.0380070265091026,
-      "grad_norm": 7.251556396484375,
+      "epoch": 1.0652245165519503,
+      "grad_norm": 5.911344528198242,
       "learning_rate": 7.502222222222223e-06,
-      "loss": 0.646,
+      "loss": 0.6784,
       "step": 1625
     },
     {
-      "epoch": 1.0539763653784733,
-      "grad_norm": 5.755091667175293,
+      "epoch": 1.0816125860373649,
+      "grad_norm": 5.509343147277832,
       "learning_rate": 7.446666666666668e-06,
-      "loss": 0.6496,
+      "loss": 0.694,
       "step": 1650
     },
     {
-      "epoch": 1.0699457042478442,
-      "grad_norm": 5.709854602813721,
+      "epoch": 1.0980006555227795,
+      "grad_norm": 5.67479944229126,
       "learning_rate": 7.3911111111111125e-06,
-      "loss": 0.6459,
+      "loss": 0.6891,
       "step": 1675
     },
     {
-      "epoch": 1.0859150431172149,
-      "grad_norm": 5.866008281707764,
+      "epoch": 1.114388725008194,
+      "grad_norm": 6.981929302215576,
       "learning_rate": 7.335555555555556e-06,
-      "loss": 0.6629,
+      "loss": 0.6854,
       "step": 1700
     },
     {
-      "epoch": 1.1018843819865858,
-      "grad_norm": 5.575136184692383,
+      "epoch": 1.1307767944936087,
+      "grad_norm": 5.943860054016113,
       "learning_rate": 7.280000000000001e-06,
-      "loss": 0.6724,
+      "loss": 0.6669,
       "step": 1725
     },
     {
-      "epoch": 1.1178537208559565,
-      "grad_norm": 5.722553730010986,
+      "epoch": 1.1471648639790233,
+      "grad_norm": 6.462680816650391,
       "learning_rate": 7.224444444444445e-06,
-      "loss": 0.6946,
+      "loss": 0.6775,
       "step": 1750
     },
     {
-      "epoch": 1.1338230597253274,
-      "grad_norm": 5.85053825378418,
+      "epoch": 1.1635529334644379,
+      "grad_norm": 5.6128621101379395,
       "learning_rate": 7.1688888888888895e-06,
-      "loss": 0.6676,
+      "loss": 0.6777,
       "step": 1775
     },
     {
-      "epoch": 1.1497923985946983,
-      "grad_norm": 6.432461261749268,
+      "epoch": 1.1799410029498525,
+      "grad_norm": 5.527766227722168,
       "learning_rate": 7.113333333333334e-06,
-      "loss": 0.6381,
+      "loss": 0.6723,
       "step": 1800
     },
     {
-      "epoch": 1.165761737464069,
-      "grad_norm": 6.030885696411133,
+      "epoch": 1.196329072435267,
+      "grad_norm": 5.95067024230957,
       "learning_rate": 7.057777777777778e-06,
-      "loss": 0.683,
+      "loss": 0.6477,
       "step": 1825
     },
     {
-      "epoch": 1.1817310763334399,
-      "grad_norm": 5.215100288391113,
+      "epoch": 1.2127171419206817,
+      "grad_norm": 6.410384654998779,
       "learning_rate": 7.0022222222222225e-06,
-      "loss": 0.6623,
+      "loss": 0.6597,
       "step": 1850
     },
     {
-      "epoch": 1.1977004152028106,
-      "grad_norm": 5.460089206695557,
+      "epoch": 1.2291052114060963,
+      "grad_norm": 5.738908767700195,
       "learning_rate": 6.946666666666667e-06,
-      "loss": 0.6519,
+      "loss": 0.6651,
       "step": 1875
     },
     {
-      "epoch": 1.2136697540721815,
-      "grad_norm": 5.882613182067871,
+      "epoch": 1.245493280891511,
+      "grad_norm": 6.21875,
       "learning_rate": 6.891111111111111e-06,
-      "loss": 0.6788,
+      "loss": 0.6647,
       "step": 1900
     },
     {
-      "epoch": 1.2296390929415522,
-      "grad_norm": 5.0330915451049805,
+      "epoch": 1.2618813503769255,
+      "grad_norm": 5.887701034545898,
       "learning_rate": 6.835555555555556e-06,
-      "loss": 0.6654,
+      "loss": 0.6963,
       "step": 1925
     },
     {
-      "epoch": 1.245608431810923,
-      "grad_norm": 4.624293804168701,
+      "epoch": 1.2782694198623403,
+      "grad_norm": 6.115960121154785,
       "learning_rate": 6.780000000000001e-06,
-      "loss": 0.6394,
+      "loss": 0.6637,
       "step": 1950
     },
     {
-      "epoch": 1.2615777706802938,
-      "grad_norm": 5.252276420593262,
+      "epoch": 1.2946574893477547,
+      "grad_norm": 5.413832664489746,
       "learning_rate": 6.724444444444444e-06,
-      "loss": 0.6522,
+      "loss": 0.6849,
       "step": 1975
     },
     {
-      "epoch": 1.2775471095496647,
-      "grad_norm": 5.393728733062744,
+      "epoch": 1.3110455588331695,
+      "grad_norm": 6.256831169128418,
       "learning_rate": 6.668888888888889e-06,
-      "loss": 0.6673,
+      "loss": 0.6606,
       "step": 2000
     },
     {
-      "epoch": 1.2775471095496647,
-      "eval_loss": 0.41851192712783813,
-      "eval_runtime": 1656.9911,
-      "eval_samples_per_second": 2.354,
+      "epoch": 1.3110455588331695,
+      "eval_loss": 0.42234352231025696,
+      "eval_runtime": 1655.8109,
+      "eval_samples_per_second": 2.356,
       "eval_steps_per_second": 0.147,
-      "eval_wer": 0.31497593933796725,
+      "eval_wer": 0.3116949399698634,
       "step": 2000
     },
     {
-      "epoch": 1.2935164484190356,
-      "grad_norm": 7.740047454833984,
+      "epoch": 1.3274336283185841,
+      "grad_norm": 5.587663650512695,
       "learning_rate": 6.613333333333334e-06,
-      "loss": 0.643,
+      "loss": 0.677,
       "step": 2025
     },
     {
-      "epoch": 1.3094857872884063,
-      "grad_norm": 4.754881381988525,
+      "epoch": 1.3438216978039987,
+      "grad_norm": 5.522030830383301,
       "learning_rate": 6.557777777777778e-06,
-      "loss": 0.6545,
+      "loss": 0.6378,
       "step": 2050
     },
     {
-      "epoch": 1.325455126157777,
-      "grad_norm": 5.5493693351745605,
+      "epoch": 1.3602097672894133,
+      "grad_norm": 5.9308953285217285,
       "learning_rate": 6.502222222222223e-06,
-      "loss": 0.6741,
+      "loss": 0.6468,
       "step": 2075
     },
     {
-      "epoch": 1.3414244650271478,
-      "grad_norm": 5.743963241577148,
+      "epoch": 1.376597836774828,
+      "grad_norm": 6.899292469024658,
       "learning_rate": 6.446666666666668e-06,
-      "loss": 0.65,
+      "loss": 0.6605,
       "step": 2100
     },
     {
-      "epoch": 1.3573938038965188,
-      "grad_norm": 6.479578495025635,
+      "epoch": 1.3929859062602425,
+      "grad_norm": 5.966159343719482,
       "learning_rate": 6.391111111111111e-06,
-      "loss": 0.6345,
+      "loss": 0.6554,
       "step": 2125
     },
     {
-      "epoch": 1.3733631427658894,
-      "grad_norm": 5.8687286376953125,
+      "epoch": 1.4093739757456571,
+      "grad_norm": 5.988006114959717,
       "learning_rate": 6.335555555555556e-06,
-      "loss": 0.6612,
+      "loss": 0.6532,
       "step": 2150
     },
     {
-      "epoch": 1.3893324816352604,
-      "grad_norm": 4.9471893310546875,
+      "epoch": 1.4257620452310718,
+      "grad_norm": 5.561163902282715,
       "learning_rate": 6.280000000000001e-06,
-      "loss": 0.6616,
+      "loss": 0.6513,
       "step": 2175
     },
     {
-      "epoch": 1.405301820504631,
-      "grad_norm": 4.968109130859375,
+      "epoch": 1.4421501147164864,
+      "grad_norm": 5.755290508270264,
       "learning_rate": 6.224444444444445e-06,
-      "loss": 0.6405,
+      "loss": 0.6308,
       "step": 2200
     },
     {
-      "epoch": 1.421271159374002,
-      "grad_norm": 6.063320159912109,
+      "epoch": 1.458538184201901,
+      "grad_norm": 5.094759464263916,
       "learning_rate": 6.16888888888889e-06,
-      "loss": 0.6615,
+      "loss": 0.629,
       "step": 2225
     },
     {
-      "epoch": 1.4372404982433729,
-      "grad_norm": 5.78763484954834,
+      "epoch": 1.4749262536873156,
+      "grad_norm": 5.547093391418457,
       "learning_rate": 6.113333333333333e-06,
-      "loss": 0.6308,
+      "loss": 0.6141,
       "step": 2250
     },
     {
-      "epoch": 1.4532098371127435,
-      "grad_norm": 5.714231014251709,
+      "epoch": 1.4913143231727304,
+      "grad_norm": 5.79092264175415,
       "learning_rate": 6.057777777777778e-06,
-      "loss": 0.6619,
+      "loss": 0.6252,
       "step": 2275
     },
     {
-      "epoch": 1.4691791759821142,
-      "grad_norm": 7.075171947479248,
+      "epoch": 1.5077023926581448,
+      "grad_norm": 5.5909271240234375,
       "learning_rate": 6.002222222222223e-06,
-      "loss": 0.6748,
+      "loss": 0.6533,
       "step": 2300
     },
     {
-      "epoch": 1.4851485148514851,
-      "grad_norm": 6.3147969245910645,
+      "epoch": 1.5240904621435596,
+      "grad_norm": 6.318563938140869,
       "learning_rate": 5.946666666666668e-06,
-      "loss": 0.6441,
+      "loss": 0.6474,
       "step": 2325
     },
     {
-      "epoch": 1.501117853720856,
-      "grad_norm": 6.079730033874512,
+      "epoch": 1.540478531628974,
+      "grad_norm": 6.217270851135254,
       "learning_rate": 5.891111111111112e-06,
-      "loss": 0.6303,
+      "loss": 0.6524,
       "step": 2350
     },
     {
-      "epoch": 1.5170871925902267,
-      "grad_norm": 5.132657527923584,
+      "epoch": 1.5568666011143888,
+      "grad_norm": 6.372344970703125,
       "learning_rate": 5.8355555555555565e-06,
-      "loss": 0.6718,
+      "loss": 0.6401,
       "step": 2375
     },
     {
-      "epoch": 1.5330565314595974,
-      "grad_norm": 5.566254138946533,
+      "epoch": 1.5732546705998034,
+      "grad_norm": 5.783897399902344,
       "learning_rate": 5.78e-06,
-      "loss": 0.6686,
+      "loss": 0.6285,
       "step": 2400
     },
     {
-      "epoch": 1.5490258703289683,
-      "grad_norm": 6.299161911010742,
+      "epoch": 1.589642740085218,
+      "grad_norm": 5.293729305267334,
       "learning_rate": 5.724444444444445e-06,
-      "loss": 0.6485,
+      "loss": 0.6294,
       "step": 2425
     },
     {
-      "epoch": 1.5649952091983392,
-      "grad_norm": 5.586335182189941,
+      "epoch": 1.6060308095706326,
+      "grad_norm": 5.974725246429443,
       "learning_rate": 5.6688888888888895e-06,
-      "loss": 0.6031,
+      "loss": 0.6373,
       "step": 2450
     },
     {
-      "epoch": 1.5809645480677101,
-      "grad_norm": 5.396686553955078,
+      "epoch": 1.6224188790560472,
+      "grad_norm": 5.62349271774292,
       "learning_rate": 5.613333333333334e-06,
-      "loss": 0.6407,
+      "loss": 0.7165,
       "step": 2475
     },
     {
-      "epoch": 1.5969338869370808,
-      "grad_norm": 6.483650207519531,
+      "epoch": 1.6388069485414618,
+      "grad_norm": 6.215331077575684,
       "learning_rate": 5.557777777777778e-06,
-      "loss": 0.6703,
+      "loss": 0.6337,
       "step": 2500
     },
     {
-      "epoch": 1.6129032258064515,
-      "grad_norm": 5.946717739105225,
+      "epoch": 1.6551950180268764,
+      "grad_norm": 6.165288925170898,
       "learning_rate": 5.5022222222222224e-06,
-      "loss": 0.6079,
+      "loss": 0.6335,
       "step": 2525
     },
     {
-      "epoch": 1.6288725646758224,
-      "grad_norm": 5.817110061645508,
+      "epoch": 1.671583087512291,
+      "grad_norm": 5.34564208984375,
       "learning_rate": 5.4466666666666665e-06,
-      "loss": 0.6275,
+      "loss": 0.6482,
       "step": 2550
     },
     {
-      "epoch": 1.6448419035451933,
-      "grad_norm": 5.888713836669922,
+      "epoch": 1.6879711569977056,
+      "grad_norm": 6.15928316116333,
       "learning_rate": 5.391111111111111e-06,
-      "loss": 0.64,
+      "loss": 0.631,
       "step": 2575
     },
     {
-      "epoch": 1.660811242414564,
-      "grad_norm": 5.296940803527832,
+      "epoch": 1.7043592264831204,
+      "grad_norm": 6.502092361450195,
       "learning_rate": 5.335555555555556e-06,
-      "loss": 0.6278,
+      "loss": 0.6239,
       "step": 2600
     },
     {
-      "epoch": 1.6767805812839347,
-      "grad_norm": 5.4542388916015625,
+      "epoch": 1.7207472959685348,
+      "grad_norm": 5.569223403930664,
       "learning_rate": 5.28e-06,
-      "loss": 0.6236,
+      "loss": 0.6203,
       "step": 2625
     },
     {
-      "epoch": 1.6927499201533056,
-      "grad_norm": 5.980891227722168,
+      "epoch": 1.7371353654539496,
+      "grad_norm": 6.04316520690918,
       "learning_rate": 5.224444444444445e-06,
-      "loss": 0.6607,
+      "loss": 0.631,
       "step": 2650
     },
     {
-      "epoch": 1.7087192590226765,
-      "grad_norm": 7.3297014236450195,
+      "epoch": 1.753523434939364,
+      "grad_norm": 5.956827163696289,
       "learning_rate": 5.168888888888889e-06,
-      "loss": 0.673,
+      "loss": 0.6421,
       "step": 2675
     },
     {
-      "epoch": 1.7246885978920474,
-      "grad_norm": 4.739117622375488,
+      "epoch": 1.7699115044247788,
+      "grad_norm": 5.976485252380371,
       "learning_rate": 5.113333333333333e-06,
-      "loss": 0.5893,
+      "loss": 0.6041,
       "step": 2700
     },
     {
-      "epoch": 1.7406579367614181,
-      "grad_norm": 4.754851818084717,
+      "epoch": 1.7862995739101932,
+      "grad_norm": 5.431225776672363,
       "learning_rate": 5.057777777777778e-06,
-      "loss": 0.6459,
+      "loss": 0.6521,
       "step": 2725
     },
     {
-      "epoch": 1.7566272756307888,
-      "grad_norm": 4.72427225112915,
+      "epoch": 1.802687643395608,
+      "grad_norm": 5.089391708374023,
       "learning_rate": 5.002222222222223e-06,
-      "loss": 0.6093,
+      "loss": 0.608,
       "step": 2750
     },
     {
-      "epoch": 1.7725966145001597,
-      "grad_norm": 5.871730327606201,
+      "epoch": 1.8190757128810227,
+      "grad_norm": 7.186110496520996,
       "learning_rate": 4.946666666666667e-06,
-      "loss": 0.6088,
+      "loss": 0.638,
       "step": 2775
     },
     {
-      "epoch": 1.7885659533695306,
-      "grad_norm": 5.640221118927002,
+      "epoch": 1.8354637823664373,
+      "grad_norm": 5.833242416381836,
       "learning_rate": 4.891111111111111e-06,
-      "loss": 0.6214,
+      "loss": 0.6396,
       "step": 2800
     },
     {
-      "epoch": 1.8045352922389013,
-      "grad_norm": 5.749293327331543,
+      "epoch": 1.8518518518518519,
+      "grad_norm": 4.990212440490723,
       "learning_rate": 4.835555555555556e-06,
-      "loss": 0.6105,
+      "loss": 0.6231,
       "step": 2825
     },
     {
-      "epoch": 1.820504631108272,
-      "grad_norm": 6.414817810058594,
+      "epoch": 1.8682399213372665,
+      "grad_norm": 5.418929100036621,
       "learning_rate": 4.78e-06,
-      "loss": 0.6523,
+      "loss": 0.6365,
       "step": 2850
     },
     {
-      "epoch": 1.836473969977643,
-      "grad_norm": 5.655066967010498,
+      "epoch": 1.884627990822681,
+      "grad_norm": 4.335667610168457,
       "learning_rate": 4.724444444444445e-06,
-      "loss": 0.6348,
+      "loss": 0.6426,
       "step": 2875
     },
     {
-      "epoch": 1.8524433088470138,
-      "grad_norm": 5.395172119140625,
+      "epoch": 1.9010160603080957,
+      "grad_norm": 4.153316020965576,
       "learning_rate": 4.66888888888889e-06,
-      "loss": 0.634,
+      "loss": 0.6197,
       "step": 2900
     },
     {
-      "epoch": 1.8684126477163845,
-      "grad_norm": 4.894425868988037,
+      "epoch": 1.9174041297935103,
+      "grad_norm": 4.748465538024902,
       "learning_rate": 4.613333333333334e-06,
-      "loss": 0.595,
+      "loss": 0.6105,
       "step": 2925
     },
     {
-      "epoch": 1.8843819865857554,
-      "grad_norm": 5.686845779418945,
+      "epoch": 1.9337921992789249,
+      "grad_norm": 4.694342136383057,
       "learning_rate": 4.557777777777778e-06,
-      "loss": 0.6113,
+      "loss": 0.6158,
       "step": 2950
     },
     {
-      "epoch": 1.900351325455126,
-      "grad_norm": 5.444372177124023,
+      "epoch": 1.9501802687643397,
+      "grad_norm": 5.531050205230713,
       "learning_rate": 4.502222222222223e-06,
-      "loss": 0.6277,
+      "loss": 0.6089,
       "step": 2975
     },
     {
-      "epoch": 1.916320664324497,
-      "grad_norm": 5.224650859832764,
+      "epoch": 1.966568338249754,
+      "grad_norm": 5.279643535614014,
       "learning_rate": 4.446666666666667e-06,
-      "loss": 0.6601,
+      "loss": 0.6065,
       "step": 3000
     },
     {
-      "epoch": 1.916320664324497,
-      "eval_loss": 0.389792263507843,
-      "eval_runtime": 1666.4367,
-      "eval_samples_per_second": 2.341,
+      "epoch": 1.966568338249754,
+      "eval_loss": 0.3859391212463379,
+      "eval_runtime": 1670.3478,
+      "eval_samples_per_second": 2.335,
       "eval_steps_per_second": 0.146,
-      "eval_wer": 0.2912798327905507,
+      "eval_wer": 0.2873426335488261,
       "step": 3000
     },
     {
-      "epoch": 1.932290003193868,
-      "grad_norm": 4.736299514770508,
+      "epoch": 1.982956407735169,
+      "grad_norm": 4.542531967163086,
       "learning_rate": 4.391111111111112e-06,
-      "loss": 0.6351,
+      "loss": 0.6178,
       "step": 3025
     },
     {
-      "epoch": 1.9482593420632386,
-      "grad_norm": 4.797982692718506,
+      "epoch": 1.9993444772205833,
+      "grad_norm": 5.153913974761963,
       "learning_rate": 4.3355555555555565e-06,
-      "loss": 0.6017,
+      "loss": 0.6124,
       "step": 3050
     },
     {
-      "epoch": 1.9642286809326093,
-      "grad_norm": 5.4142022132873535,
+      "epoch": 2.015732546705998,
+      "grad_norm": 4.631136417388916,
       "learning_rate": 4.2800000000000005e-06,
-      "loss": 0.6124,
+      "loss": 0.4396,
       "step": 3075
     },
     {
-      "epoch": 1.9801980198019802,
-      "grad_norm": 5.374582767486572,
+      "epoch": 2.0321206161914125,
+      "grad_norm": 5.14950704574585,
       "learning_rate": 4.2244444444444446e-06,
-      "loss": 0.6243,
+      "loss": 0.3907,
       "step": 3100
     },
     {
-      "epoch": 1.996167358671351,
-      "grad_norm": 4.480958938598633,
+      "epoch": 2.0485086856768273,
+      "grad_norm": 4.144317150115967,
       "learning_rate": 4.168888888888889e-06,
-      "loss": 0.5981,
+      "loss": 0.39,
       "step": 3125
     },
     {
-      "epoch": 2.012136697540722,
-      "grad_norm": 5.1775736808776855,
+      "epoch": 2.0648967551622417,
+      "grad_norm": 4.183619499206543,
       "learning_rate": 4.1133333333333335e-06,
-      "loss": 0.4607,
+      "loss": 0.4086,
       "step": 3150
     },
     {
-      "epoch": 2.0281060364100925,
-      "grad_norm": 4.18350887298584,
+      "epoch": 2.0812848246476565,
+      "grad_norm": 5.38789176940918,
       "learning_rate": 4.057777777777778e-06,
-      "loss": 0.3907,
+      "loss": 0.3987,
       "step": 3175
     },
     {
-      "epoch": 2.0440753752794634,
-      "grad_norm": 4.902673721313477,
+      "epoch": 2.097672894133071,
+      "grad_norm": 5.071357727050781,
       "learning_rate": 4.002222222222222e-06,
-      "loss": 0.4004,
+      "loss": 0.4209,
       "step": 3200
     },
     {
-      "epoch": 2.0600447141488343,
-      "grad_norm": 4.528185844421387,
+      "epoch": 2.1140609636184857,
+      "grad_norm": 5.261667728424072,
       "learning_rate": 3.946666666666667e-06,
-      "loss": 0.3936,
+      "loss": 0.4061,
       "step": 3225
     },
     {
-      "epoch": 2.076014053018205,
-      "grad_norm": 5.0421061515808105,
+      "epoch": 2.1304490331039005,
+      "grad_norm": 4.820098876953125,
       "learning_rate": 3.891111111111111e-06,
-      "loss": 0.4242,
+      "loss": 0.4132,
       "step": 3250
     },
     {
-      "epoch": 2.0919833918875756,
-      "grad_norm": 4.861361026763916,
+      "epoch": 2.146837102589315,
+      "grad_norm": 5.199064254760742,
       "learning_rate": 3.835555555555555e-06,
-      "loss": 0.3916,
+      "loss": 0.41,
       "step": 3275
     },
     {
-      "epoch": 2.1079527307569466,
-      "grad_norm": 6.367924213409424,
+      "epoch": 2.1632251720747298,
+      "grad_norm": 4.427770614624023,
       "learning_rate": 3.7800000000000002e-06,
-      "loss": 0.413,
+      "loss": 0.3957,
       "step": 3300
     },
     {
-      "epoch": 2.1239220696263175,
-      "grad_norm": 3.6156017780303955,
+      "epoch": 2.179613241560144,
+      "grad_norm": 4.931752681732178,
       "learning_rate": 3.724444444444445e-06,
-      "loss": 0.4103,
+      "loss": 0.4051,
       "step": 3325
     },
     {
-      "epoch": 2.1398914084956884,
-      "grad_norm": 5.132371425628662,
+      "epoch": 2.196001311045559,
+      "grad_norm": 4.410216331481934,
       "learning_rate": 3.668888888888889e-06,
-      "loss": 0.407,
+      "loss": 0.3908,
       "step": 3350
     },
     {
-      "epoch": 2.1558607473650593,
-      "grad_norm": 4.446967601776123,
+      "epoch": 2.2123893805309733,
+      "grad_norm": 5.322948455810547,
       "learning_rate": 3.6133333333333336e-06,
-      "loss": 0.4082,
+      "loss": 0.3829,
       "step": 3375
     },
     {
-      "epoch": 2.1718300862344297,
-      "grad_norm": 5.18086576461792,
+      "epoch": 2.228777450016388,
+      "grad_norm": 4.69309663772583,
       "learning_rate": 3.5577777777777785e-06,
-      "loss": 0.4053,
+      "loss": 0.3768,
       "step": 3400
     },
     {
-      "epoch": 2.1877994251038007,
-      "grad_norm": 4.955840587615967,
+      "epoch": 2.2451655195018025,
+      "grad_norm": 5.363163471221924,
       "learning_rate": 3.5022222222222225e-06,
-      "loss": 0.3995,
+      "loss": 0.4241,
       "step": 3425
     },
     {
-      "epoch": 2.2037687639731716,
-      "grad_norm": 4.59506368637085,
+      "epoch": 2.2615535889872174,
+      "grad_norm": 5.2487688064575195,
       "learning_rate": 3.446666666666667e-06,
-      "loss": 0.4064,
+      "loss": 0.4001,
       "step": 3450
     },
     {
-      "epoch": 2.2197381028425425,
-      "grad_norm": 5.527169227600098,
+      "epoch": 2.2779416584726317,
+      "grad_norm": 4.444194316864014,
       "learning_rate": 3.391111111111111e-06,
-      "loss": 0.3946,
+      "loss": 0.4114,
       "step": 3475
     },
     {
-      "epoch": 2.235707441711913,
-      "grad_norm": 5.034458637237549,
+      "epoch": 2.2943297279580466,
+      "grad_norm": 4.2093892097473145,
       "learning_rate": 3.335555555555556e-06,
-      "loss": 0.4088,
+      "loss": 0.4056,
       "step": 3500
     },
     {
-      "epoch": 2.251676780581284,
-      "grad_norm": 4.814046382904053,
+      "epoch": 2.310717797443461,
+      "grad_norm": 4.929553508758545,
       "learning_rate": 3.2800000000000004e-06,
-      "loss": 0.4032,
+      "loss": 0.3952,
       "step": 3525
     },
     {
-      "epoch": 2.2676461194506548,
-      "grad_norm": 5.115776062011719,
+      "epoch": 2.3271058669288758,
+      "grad_norm": 5.331225395202637,
       "learning_rate": 3.2244444444444444e-06,
-      "loss": 0.4009,
+      "loss": 0.381,
       "step": 3550
     },
     {
-      "epoch": 2.2836154583200257,
-      "grad_norm": 4.4629130363464355,
+      "epoch": 2.3434939364142906,
+      "grad_norm": 4.381256580352783,
       "learning_rate": 3.1688888888888893e-06,
-      "loss": 0.408,
+      "loss": 0.399,
       "step": 3575
     },
     {
-      "epoch": 2.2995847971893966,
-      "grad_norm": 4.914193153381348,
+      "epoch": 2.359882005899705,
+      "grad_norm": 4.569367408752441,
       "learning_rate": 3.1133333333333337e-06,
-      "loss": 0.3867,
+      "loss": 0.4195,
       "step": 3600
     },
     {
-      "epoch": 2.315554136058767,
-      "grad_norm": 4.358288288116455,
+      "epoch": 2.37627007538512,
+      "grad_norm": 5.157946586608887,
       "learning_rate": 3.0577777777777778e-06,
-      "loss": 0.4485,
+      "loss": 0.4796,
       "step": 3625
     },
     {
-      "epoch": 2.331523474928138,
-      "grad_norm": 4.8634161949157715,
+      "epoch": 2.392658144870534,
+      "grad_norm": 4.326797008514404,
       "learning_rate": 3.0022222222222227e-06,
-      "loss": 0.389,
+      "loss": 0.4094,
       "step": 3650
     },
     {
-      "epoch": 2.347492813797509,
-      "grad_norm": 5.283102989196777,
+      "epoch": 2.409046214355949,
+      "grad_norm": 5.1377081871032715,
       "learning_rate": 2.946666666666667e-06,
-      "loss": 0.395,
+      "loss": 0.3983,
       "step": 3675
     },
     {
-      "epoch": 2.3634621526668798,
-      "grad_norm": 4.393892765045166,
+      "epoch": 2.4254342838413634,
+      "grad_norm": 5.24452543258667,
       "learning_rate": 2.891111111111111e-06,
-      "loss": 0.4047,
+      "loss": 0.4243,
       "step": 3700
     },
     {
-      "epoch": 2.37943149153625,
-      "grad_norm": 5.246931552886963,
+      "epoch": 2.441822353326778,
+      "grad_norm": 5.172399997711182,
       "learning_rate": 2.835555555555556e-06,
-      "loss": 0.3957,
+      "loss": 0.3962,
       "step": 3725
     },
     {
-      "epoch": 2.395400830405621,
-      "grad_norm": 4.599190711975098,
+      "epoch": 2.4582104228121926,
+      "grad_norm": 4.068329811096191,
       "learning_rate": 2.7800000000000005e-06,
-      "loss": 0.3829,
+      "loss": 0.3857,
       "step": 3750
     },
     {
-      "epoch": 2.411370169274992,
-      "grad_norm": 5.136366367340088,
+      "epoch": 2.4745984922976074,
+      "grad_norm": 5.2074995040893555,
       "learning_rate": 2.7244444444444445e-06,
-      "loss": 0.4029,
+      "loss": 0.4035,
       "step": 3775
     },
     {
-      "epoch": 2.427339508144363,
-      "grad_norm": 4.448164463043213,
+      "epoch": 2.490986561783022,
+      "grad_norm": 6.278205871582031,
       "learning_rate": 2.6688888888888894e-06,
-      "loss": 0.3869,
+      "loss": 0.4125,
       "step": 3800
     },
     {
-      "epoch": 2.443308847013734,
-      "grad_norm": 5.434977054595947,
+      "epoch": 2.5073746312684366,
+      "grad_norm": 5.124298572540283,
       "learning_rate": 2.6133333333333334e-06,
-      "loss": 0.3871,
+      "loss": 0.4153,
       "step": 3825
     },
     {
-      "epoch": 2.4592781858831043,
-      "grad_norm": 5.272222518920898,
+      "epoch": 2.523762700753851,
+      "grad_norm": 5.120030879974365,
       "learning_rate": 2.557777777777778e-06,
-      "loss": 0.388,
+      "loss": 0.4178,
       "step": 3850
     },
     {
-      "epoch": 2.4752475247524752,
-      "grad_norm": 5.163560390472412,
+      "epoch": 2.540150770239266,
+      "grad_norm": 5.407374382019043,
       "learning_rate": 2.5022222222222224e-06,
-      "loss": 0.4156,
+      "loss": 0.4044,
       "step": 3875
     },
     {
-      "epoch": 2.491216863621846,
-      "grad_norm": 4.091150283813477,
+      "epoch": 2.5565388397246807,
+      "grad_norm": 4.730270862579346,
       "learning_rate": 2.446666666666667e-06,
-      "loss": 0.4093,
+      "loss": 0.3733,
       "step": 3900
     },
     {
-      "epoch": 2.5071862024912166,
-      "grad_norm": 4.858378887176514,
+      "epoch": 2.572926909210095,
+      "grad_norm": 4.6310224533081055,
       "learning_rate": 2.3911111111111113e-06,
-      "loss": 0.4068,
+      "loss": 0.4064,
       "step": 3925
     },
     {
-      "epoch": 2.5231555413605875,
-      "grad_norm": 5.535120487213135,
+      "epoch": 2.5893149786955094,
+      "grad_norm": 3.4895918369293213,
       "learning_rate": 2.3355555555555557e-06,
-      "loss": 0.3992,
+      "loss": 0.3976,
       "step": 3950
     },
     {
-      "epoch": 2.5391248802299584,
-      "grad_norm": 4.70937442779541,
+      "epoch": 2.6057030481809242,
+      "grad_norm": 5.346146106719971,
       "learning_rate": 2.28e-06,
-      "loss": 0.4009,
+      "loss": 0.3833,
       "step": 3975
     },
     {
-      "epoch": 2.5550942190993293,
-      "grad_norm": 5.273413181304932,
+      "epoch": 2.622091117666339,
+      "grad_norm": 4.933919906616211,
       "learning_rate": 2.2244444444444447e-06,
-      "loss": 0.3887,
+      "loss": 0.3894,
       "step": 4000
     },
     {
-      "epoch": 2.5550942190993293,
-      "eval_loss": 0.3953465223312378,
-      "eval_runtime": 1634.3629,
-      "eval_samples_per_second": 2.387,
-      "eval_steps_per_second": 0.149,
-      "eval_wer": 0.2815826568803772,
+      "epoch": 2.622091117666339,
+      "eval_loss": 0.39615824818611145,
+      "eval_runtime": 1630.0413,
+      "eval_samples_per_second": 2.393,
+      "eval_steps_per_second": 0.15,
+      "eval_wer": 0.2787391240946872,
       "step": 4000
     },
     {
-      "epoch": 2.5710635579687002,
-      "grad_norm": 5.2189483642578125,
+      "epoch": 2.6384791871517534,
+      "grad_norm": 4.994978427886963,
       "learning_rate": 2.168888888888889e-06,
-      "loss": 0.3789,
+      "loss": 0.3979,
       "step": 4025
     },
     {
-      "epoch": 2.587032896838071,
-      "grad_norm": 5.335792064666748,
+      "epoch": 2.6548672566371683,
+      "grad_norm": 5.25327730178833,
       "learning_rate": 2.1133333333333336e-06,
-      "loss": 0.3911,
+      "loss": 0.3912,
       "step": 4050
     },
     {
-      "epoch": 2.6030022357074416,
-      "grad_norm": 4.469089508056641,
+      "epoch": 2.6712553261225827,
+      "grad_norm": 5.466716766357422,
       "learning_rate": 2.057777777777778e-06,
-      "loss": 0.388,
+      "loss": 0.3953,
       "step": 4075
     },
     {
-      "epoch": 2.6189715745768125,
-      "grad_norm": 4.454526424407959,
+      "epoch": 2.6876433956079975,
+      "grad_norm": 4.965219497680664,
       "learning_rate": 2.0022222222222225e-06,
-      "loss": 0.3829,
+      "loss": 0.3809,
       "step": 4100
     },
     {
-      "epoch": 2.6349409134461834,
-      "grad_norm": 5.536740303039551,
+      "epoch": 2.704031465093412,
+      "grad_norm": 5.104487895965576,
       "learning_rate": 1.9466666666666665e-06,
-      "loss": 0.3846,
+      "loss": 0.389,
       "step": 4125
     },
     {
-      "epoch": 2.650910252315554,
-      "grad_norm": 4.877845764160156,
+      "epoch": 2.7204195345788267,
+      "grad_norm": 5.978146076202393,
       "learning_rate": 1.8911111111111114e-06,
-      "loss": 0.4116,
+      "loss": 0.3995,
       "step": 4150
     },
     {
-      "epoch": 2.666879591184925,
-      "grad_norm": 5.3322014808654785,
+      "epoch": 2.736807604064241,
+      "grad_norm": 5.592274188995361,
       "learning_rate": 1.8355555555555557e-06,
-      "loss": 0.3857,
+      "loss": 0.3829,
       "step": 4175
     },
     {
-      "epoch": 2.6828489300542957,
-      "grad_norm": 4.9132866859436035,
+      "epoch": 2.753195673549656,
+      "grad_norm": 4.805532932281494,
       "learning_rate": 1.7800000000000001e-06,
-      "loss": 0.3922,
+      "loss": 0.387,
       "step": 4200
     },
     {
-      "epoch": 2.6988182689236666,
-      "grad_norm": 5.380650043487549,
+      "epoch": 2.7695837430350707,
+      "grad_norm": 4.371249198913574,
       "learning_rate": 1.7244444444444448e-06,
-      "loss": 0.3757,
+      "loss": 0.3969,
       "step": 4225
     },
     {
-      "epoch": 2.7147876077930375,
-      "grad_norm": 4.5387282371521,
+      "epoch": 2.785971812520485,
+      "grad_norm": 5.515538215637207,
       "learning_rate": 1.668888888888889e-06,
-      "loss": 0.3919,
+      "loss": 0.3804,
       "step": 4250
     },
     {
-      "epoch": 2.7307569466624084,
-      "grad_norm": 5.215519428253174,
+      "epoch": 2.8023598820058995,
+      "grad_norm": 4.387998104095459,
       "learning_rate": 1.6133333333333335e-06,
-      "loss": 0.4104,
+      "loss": 0.3976,
       "step": 4275
     },
     {
-      "epoch": 2.746726285531779,
-      "grad_norm": 6.126536846160889,
+      "epoch": 2.8187479514913143,
+      "grad_norm": 6.351314544677734,
       "learning_rate": 1.5577777777777777e-06,
-      "loss": 0.3792,
+      "loss": 0.423,
       "step": 4300
     },
     {
-      "epoch": 2.76269562440115,
-      "grad_norm": 5.707647323608398,
+      "epoch": 2.835136020976729,
+      "grad_norm": 5.132925987243652,
       "learning_rate": 1.5022222222222224e-06,
-      "loss": 0.4041,
+      "loss": 0.3954,
       "step": 4325
     },
     {
-      "epoch": 2.7786649632705207,
-      "grad_norm": 5.024244785308838,
+      "epoch": 2.8515240904621435,
+      "grad_norm": 4.940093517303467,
       "learning_rate": 1.4466666666666669e-06,
-      "loss": 0.4045,
+      "loss": 0.3874,
       "step": 4350
     },
     {
-      "epoch": 2.794634302139891,
-      "grad_norm": 5.240878105163574,
+      "epoch": 2.8679121599475583,
+      "grad_norm": 4.280588626861572,
       "learning_rate": 1.3911111111111111e-06,
-      "loss": 0.4022,
+      "loss": 0.3962,
       "step": 4375
     },
     {
-      "epoch": 2.810603641009262,
-      "grad_norm": 5.824533939361572,
+      "epoch": 2.8843002294329727,
+      "grad_norm": 4.708969593048096,
       "learning_rate": 1.3355555555555558e-06,
-      "loss": 0.3947,
+      "loss": 0.3874,
       "step": 4400
     },
     {
-      "epoch": 2.826572979878633,
-      "grad_norm": 4.228843688964844,
+      "epoch": 2.9006882989183875,
+      "grad_norm": 4.572214126586914,
       "learning_rate": 1.28e-06,
-      "loss": 0.3725,
+      "loss": 0.3618,
       "step": 4425
     },
     {
-      "epoch": 2.842542318748004,
-      "grad_norm": 4.704908847808838,
+      "epoch": 2.917076368403802,
+      "grad_norm": 5.374238967895508,
       "learning_rate": 1.2244444444444445e-06,
-      "loss": 0.4001,
+      "loss": 0.3789,
       "step": 4450
     },
     {
-      "epoch": 2.858511657617375,
-      "grad_norm": 4.397124767303467,
+      "epoch": 2.9334644378892167,
+      "grad_norm": 4.603006362915039,
       "learning_rate": 1.168888888888889e-06,
-      "loss": 0.381,
+      "loss": 0.3744,
       "step": 4475
     },
     {
-      "epoch": 2.8744809964867457,
-      "grad_norm": 4.7757062911987305,
+      "epoch": 2.949852507374631,
+      "grad_norm": 5.514355659484863,
       "learning_rate": 1.1133333333333334e-06,
-      "loss": 0.3689,
+      "loss": 0.3705,
       "step": 4500
     },
     {
-      "epoch": 2.890450335356116,
-      "grad_norm": 4.855478763580322,
+      "epoch": 2.966240576860046,
+      "grad_norm": 5.96732234954834,
       "learning_rate": 1.0577777777777779e-06,
-      "loss": 0.373,
+      "loss": 0.3816,
       "step": 4525
     },
     {
-      "epoch": 2.906419674225487,
-      "grad_norm": 5.458982467651367,
+      "epoch": 2.9826286463454608,
+      "grad_norm": 4.869338035583496,
       "learning_rate": 1.0022222222222223e-06,
-      "loss": 0.3919,
+      "loss": 0.4033,
       "step": 4550
     },
     {
-      "epoch": 2.922389013094858,
-      "grad_norm": 3.9116828441619873,
+      "epoch": 2.999016715830875,
+      "grad_norm": 4.8514404296875,
       "learning_rate": 9.466666666666667e-07,
-      "loss": 0.4143,
+      "loss": 0.3862,
       "step": 4575
     },
     {
-      "epoch": 2.9383583519642285,
-      "grad_norm": 4.8182878494262695,
+      "epoch": 3.0154047853162895,
+      "grad_norm": 4.073346138000488,
       "learning_rate": 8.911111111111112e-07,
-      "loss": 0.3726,
+      "loss": 0.2642,
       "step": 4600
     },
     {
-      "epoch": 2.9543276908335994,
-      "grad_norm": 5.018452167510986,
+      "epoch": 3.0317928548017044,
+      "grad_norm": 3.8675925731658936,
       "learning_rate": 8.355555555555556e-07,
-      "loss": 0.4018,
+      "loss": 0.2484,
       "step": 4625
     },
     {
-      "epoch": 2.9702970297029703,
-      "grad_norm": 5.315720081329346,
+      "epoch": 3.048180924287119,
+      "grad_norm": 4.085783004760742,
       "learning_rate": 7.8e-07,
-      "loss": 0.3648,
+      "loss": 0.253,
       "step": 4650
     },
     {
-      "epoch": 2.986266368572341,
-      "grad_norm": 4.65333890914917,
-      "learning_rate": 7.266666666666668e-07,
-      "loss": 0.3865,
+      "epoch": 3.0645689937725336,
+      "grad_norm": 5.922070026397705,
+      "learning_rate": 7.244444444444446e-07,
+      "loss": 0.2572,
       "step": 4675
     },
     {
-      "epoch": 3.002235707441712,
-      "grad_norm": 2.9384920597076416,
-      "learning_rate": 6.711111111111111e-07,
-      "loss": 0.3793,
+      "epoch": 3.0809570632579484,
+      "grad_norm": 3.9240541458129883,
+      "learning_rate": 6.68888888888889e-07,
+      "loss": 0.2628,
       "step": 4700
     },
     {
-      "epoch": 3.0182050463110826,
-      "grad_norm": 4.572001934051514,
-      "learning_rate": 6.155555555555556e-07,
-      "loss": 0.2557,
+      "epoch": 3.0973451327433628,
+      "grad_norm": 3.6416871547698975,
+      "learning_rate": 6.133333333333333e-07,
+      "loss": 0.2762,
       "step": 4725
     },
     {
-      "epoch": 3.0341743851804535,
-      "grad_norm": 3.237912654876709,
-      "learning_rate": 5.6e-07,
-      "loss": 0.252,
+      "epoch": 3.1137332022287776,
+      "grad_norm": 3.966794729232788,
+      "learning_rate": 5.577777777777779e-07,
+      "loss": 0.2641,
       "step": 4750
     },
     {
-      "epoch": 3.0501437240498244,
-      "grad_norm": 3.8456506729125977,
-      "learning_rate": 5.044444444444445e-07,
-      "loss": 0.2556,
+      "epoch": 3.130121271714192,
+      "grad_norm": 4.7765045166015625,
+      "learning_rate": 5.022222222222222e-07,
+      "loss": 0.2581,
       "step": 4775
     },
     {
-      "epoch": 3.0661130629191953,
-      "grad_norm": 4.627549171447754,
-      "learning_rate": 4.488888888888889e-07,
-      "loss": 0.2618,
+      "epoch": 3.146509341199607,
+      "grad_norm": 4.1129608154296875,
+      "learning_rate": 4.466666666666667e-07,
+      "loss": 0.2715,
       "step": 4800
     },
     {
-      "epoch": 3.0820824017885657,
-      "grad_norm": 4.322127342224121,
-      "learning_rate": 3.9333333333333336e-07,
-      "loss": 0.2715,
+      "epoch": 3.162897410685021,
+      "grad_norm": 3.8220412731170654,
+      "learning_rate": 3.9111111111111115e-07,
+      "loss": 0.2639,
       "step": 4825
     },
     {
-      "epoch": 3.0980517406579366,
-      "grad_norm": 4.883068084716797,
-      "learning_rate": 3.3777777777777777e-07,
-      "loss": 0.2585,
+      "epoch": 3.179285480170436,
+      "grad_norm": 4.8853912353515625,
+      "learning_rate": 3.3555555555555556e-07,
+      "loss": 0.2666,
       "step": 4850
     },
     {
-      "epoch": 3.1140210795273076,
-      "grad_norm": 4.687808513641357,
-      "learning_rate": 2.822222222222222e-07,
-      "loss": 0.2593,
+      "epoch": 3.1956735496558504,
+      "grad_norm": 4.593063831329346,
+      "learning_rate": 2.8e-07,
+      "loss": 0.2639,
       "step": 4875
     },
     {
-      "epoch": 3.1299904183966785,
-      "grad_norm": 4.111570835113525,
-      "learning_rate": 2.266666666666667e-07,
-      "loss": 0.2645,
+      "epoch": 3.212061619141265,
+      "grad_norm": 3.6766130924224854,
+      "learning_rate": 2.2444444444444445e-07,
+      "loss": 0.2571,
       "step": 4900
     },
     {
-      "epoch": 3.1459597572660494,
-      "grad_norm": 4.777975082397461,
-      "learning_rate": 1.7111111111111114e-07,
-      "loss": 0.2594,
+      "epoch": 3.2284496886266796,
+      "grad_norm": 4.200336933135986,
+      "learning_rate": 1.6888888888888888e-07,
+      "loss": 0.2507,
       "step": 4925
     },
     {
-      "epoch": 3.16192909613542,
-      "grad_norm": 4.395523548126221,
-      "learning_rate": 1.1555555555555556e-07,
-      "loss": 0.2665,
+      "epoch": 3.2448377581120944,
+      "grad_norm": 4.104145050048828,
+      "learning_rate": 1.1333333333333336e-07,
+      "loss": 0.2527,
       "step": 4950
     },
     {
-      "epoch": 3.1778984350047907,
-      "grad_norm": 5.332032203674316,
-      "learning_rate": 6.000000000000001e-08,
-      "loss": 0.2772,
+      "epoch": 3.2612258275975092,
+      "grad_norm": 3.6322615146636963,
+      "learning_rate": 5.777777777777778e-08,
+      "loss": 0.2512,
       "step": 4975
     },
     {
-      "epoch": 3.1938677738741617,
-      "grad_norm": 4.1832733154296875,
-      "learning_rate": 4.444444444444445e-09,
-      "loss": 0.2637,
+      "epoch": 3.2776138970829236,
+      "grad_norm": 3.6525533199310303,
+      "learning_rate": 2.2222222222222225e-09,
+      "loss": 0.2478,
       "step": 5000
     },
     {
-      "epoch": 3.1938677738741617,
-      "eval_loss": 0.4074931740760803,
-      "eval_runtime": 1670.4337,
-      "eval_samples_per_second": 2.335,
-      "eval_steps_per_second": 0.146,
-      "eval_wer": 0.2823846789481359,
+      "epoch": 3.2776138970829236,
+      "eval_loss": 0.41265004873275757,
+      "eval_runtime": 1618.8893,
+      "eval_samples_per_second": 2.41,
+      "eval_steps_per_second": 0.151,
+      "eval_wer": 0.2775239391435376,
       "step": 5000
     },
     {
-      "epoch": 3.1938677738741617,
+      "epoch": 3.2776138970829236,
       "step": 5000,
       "total_flos": 5.435895365546803e+20,
-      "train_loss": 0.7105956988334656,
-      "train_runtime": 59674.9515,
-      "train_samples_per_second": 2.681,
-      "train_steps_per_second": 0.084
+      "train_loss": 0.7020168427467346,
+      "train_runtime": 60500.8955,
+      "train_samples_per_second": 2.645,
+      "train_steps_per_second": 0.083
     }
   ],
   "logging_steps": 25,