End of training

README.md

@@ -16,9 +16,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model was trained from scratch on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 3.2885
-- Cer: 0.5196
-- Wer: 0.8050
+- Loss: 1.9947
+- Cer: 0.4133
+- Wer: 0.6195
 
 ## Model description
 

all_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 5.6,
-    "eval_cer": 0.39621716426834613,
-    "eval_loss": 2.00065016746521,
-    "eval_runtime": 157.5611,
+    "epoch": 9.6,
+    "eval_cer": 0.4133216406903974,
+    "eval_loss": 1.9947007894515991,
+    "eval_runtime": 158.2803,
     "eval_samples": 3136,
-    "eval_samples_per_second": 19.903,
-    "eval_steps_per_second": 1.244,
-    "eval_wer": 0.6068801160501502,
-    "total_flos": 2.158464150901847e+19,
-    "train_loss": 1.3790143143790108,
-    "train_runtime": 12492.843,
+    "eval_samples_per_second": 19.813,
+    "eval_steps_per_second": 1.238,
+    "eval_wer": 0.6194798466480157,
+    "total_flos": 3.700768773245485e+19,
+    "train_loss": 2.825458660195271,
+    "train_runtime": 12931.7591,
     "train_samples": 20000,
-    "train_samples_per_second": 8.965,
-    "train_steps_per_second": 0.28
+    "train_samples_per_second": 14.847,
+    "train_steps_per_second": 0.464
 }

eval_results.json

@@ -1,10 +1,10 @@
 {
-    "epoch": 5.6,
-    "eval_cer": 0.39621716426834613,
-    "eval_loss": 2.00065016746521,
-    "eval_runtime": 157.5611,
+    "epoch": 9.6,
+    "eval_cer": 0.4133216406903974,
+    "eval_loss": 1.9947007894515991,
+    "eval_runtime": 158.2803,
     "eval_samples": 3136,
-    "eval_samples_per_second": 19.903,
-    "eval_steps_per_second": 1.244,
-    "eval_wer": 0.6068801160501502
+    "eval_samples_per_second": 19.813,
+    "eval_steps_per_second": 1.238,
+    "eval_wer": 0.6194798466480157
 }

indicwav2vec_trainwtags_MUCS_warmup2000_s300shuff100_2212172.out

@@ -7586,3 +7586,127 @@ last prediction string <>धलता<>धवारा अधुवाधित
 
                                                 
 
 
                                                 
                                                     
+Printing predictions for a few samples:
+Sample 1:
+  Reference: लिबर ऑफिस <cs> impress </cs> में एक प्रस्तुति <cs> document </cs> बनाना और बुनियादी <cs> formatting </cs> के इस <cs> spoken  tutorial </cs> में आपका स्वागत है
+######
+
+
+  Prediction: > लिbeरॉफe <s> enp्रe</cs> में एक <्स्धुध<cs>धो्ुमn </c> बनधानाअर<बुधियाध <s> फर्मmेटen </cs> के इस <s>्ोकn <>्uुटोर </cs> में अपका सधवाधँहंह
+
+
+
+Sample 2:
+  Reference: इस <cs> tutorial </cs> में हम <cs> impress  window </cs> के भागों के बारे में सीखेंगे और कैसे स्लाइड इन्सर्ट करें और कॉपी करें फॉन्ट तथा फॉन्ट को फॉर्मेट करना सीखेंगे
+######
+
+
+  Prediction: इस<cs>्ुटोaल </s> में<हम<> em्res> विnधो </s> के धागों केबारे मं सीखेंगेउर<कैसे <cs> लae </s> इnser्</cs> करेंअर <cs> cॉ </cs> कर>n </c>तधा <cs> on</cs> को <s> ॉर्मa</cs> करधा सीखेंगे
+
+
+
+Sample 3:
+  Reference: यहाँ हम अपने ऑपरेटिंग सिस्टम के रूप में gnu/linux और लिबरऑफिस वर्जन <cs> 334 </cs> का उपयोग कर रहे हैं
+######
+
+
+  Prediction: धहाँ हम अपने <cs> ॉp्रेen <>सिs्टeम </cs> के रधुमें <s>जnuउ<s> लिnc्/अउर<cs> लिber>ॉफe<s> वer्जon <</cs> का उप्योगकरधहंं
+
+
+
+Sample 4:
+  Reference: चलिए अपनी प्रस्तुति प्रेजैटेशन <cs> sample  impress  open </cs> करते हैं जिसे पिछले <cs> tutorial </cs> में बनाया था
+######
+
+
+  Prediction: चलिएअपधी >्रस्धुतधि<ape <> ene<उn </cs> करते हैंजिसेधल्हं
+
+
+
+Sample 5:
+  Reference: चलिए देखते हैं कि <cs> screen </cs> पर क्या क्या है
+######
+
+
+  Prediction: अचलिएदेखते हैंकि <cs>्क्रीn </cs> पर क्ा क्ा है
+
+
+
+last Reference string यह स्क्रिप्ट लता द्वारा अनुवादित है आईआईटी मुंबई की ओर से मैं रवि कुमार अब आपसे विदा लेता हूँहमसे जुड़ने के लिए धन्यवाद
+
+
+last prediction string <>धलता<>धवारा अधुवाधितहअउट>मुnंब> की उरसेमैं रविकुमारअब अपसे विधा लेता हूँहमसेजुधने के लिए धध्यवाद
+{'eval_loss': 3.2885050773620605, 'eval_cer': 0.5195918204647826, 'eval_wer': 0.8050150243498083, 'eval_runtime': 157.437, 'eval_samples_per_second': 19.919, 'eval_steps_per_second': 1.245, 'epoch': 9.6}
+{'train_runtime': 12931.7591, 'train_samples_per_second': 14.847, 'train_steps_per_second': 0.464, 'train_loss': 2.825458660195271, 'epoch': 9.6}
+***** train metrics *****
+  epoch                    =           9.6
+  total_flos               = 34466095019GF
+  train_loss               =        2.8255
+  train_runtime            =    3:35:31.75
+  train_samples            =         20000
+  train_samples_per_second =        14.847
+  train_steps_per_second   =         0.464
+08/28/2024 06:04:48 - INFO - __main__ - *** Evaluate ***
+/scratch/work/palp3/myenv/lib/python3.11/site-packages/transformers/models/wav2vec2/processing_wav2vec2.py:157: UserWarning: `as_target_processor` is deprecated and will be removed in v5 of Transformers. You can process your labels by using the argument `text` of the regular `__call__` method (either in the same call as your audio inputs, or in a separate call.
+  warnings.warn(
+
  0%|          | 0/196 [00:00<?, ?it/s]
  1%|          | 2/196 [00:00<01:02,  3.08it/s]
  2%|▏         | 3/196 [00:01<01:30,  2.13it/s]
  2%|▏         | 4/196 [00:02<01:51,  1.72it/s]
  3%|▎         | 5/196 [00:02<02:04,  1.53it/s]
  3%|▎         | 6/196 [00:03<02:21,  1.34it/s]
  4%|▎         | 7/196 [00:04<02:24,  1.31it/s]
  4%|▍         | 8/196 [00:05<02:34,  1.22it/s]
  5%|▍         | 9/196 [00:07<03:23,  1.09s/it]
  5%|▌         | 10/196 [00:08<03:55,  1.27s/it]
  6%|▌         | 11/196 [00:10<04:30,  1.46s/it]
  6%|▌         | 12/196 [00:12<04:35,  1.50s/it]
  7%|▋         | 13/196 [00:13<03:59,  1.31s/it]
  7%|▋         | 14/196 [00:13<03:23,  1.12s/it]
  8%|▊         | 15/196 [00:14<02:53,  1.04it/s]
  8%|▊         | 16/196 [00:15<02:43,  1.10it/s]
  9%|▊         | 17/196 [00:16<02:41,  1.11it/s]
  9%|▉         | 18/196 [00:17<03:04,  1.04s/it]
 10%|▉         | 19/196 [00:19<03:51,  1.31s/it]
 10%|█         | 20/196 [00:21<04:01,  1.37s/it]
 11%|█         | 21/196 [00:22<04:08,  1.42s/it]
 11%|█         | 22/196 [00:23<03:52,  1.34s/it]
 12%|█▏        | 23/196 [00:24<03:26,  1.20s/it]
 12%|█▏        | 24/196 [00:25<02:47,  1.03it/s]
 13%|█▎        | 25/196 [00:25<02:26,  1.17it/s]
 13%|█▎        | 26/196 [00:26<02:10,  1.31it/s]
 14%|█▍        | 27/196 [00:26<02:00,  1.40it/s]
 14%|█▍        | 28/196 [00:27<01:56,  1.45it/s]
 15%|█▍        | 29/196 [00:28<01:56,  1.43it/s]
 15%|█▌        | 30/196 [00:28<01:54,  1.44it/s]
 16%|█▌        | 31/196 [00:29<01:43,  1.59it/s]
 16%|█▋        | 32/196 [00:29<01:44,  1.57it/s]
 17%|█▋        | 33/196 [00:30<01:59,  1.37it/s]
 17%|█▋        | 34/196 [00:32<02:24,  1.12it/s]
 18%|█▊        | 35/196 [00:33<02:35,  1.03it/s]
 18%|█▊        | 36/196 [00:34<02:51,  1.07s/it]
 19%|█▉        | 37/196 [00:35<02:45,  1.04s/it]
 19%|█▉        | 38/196 [00:36<02:34,  1.03it/s]
 20%|█▉        | 39/196 [00:37<02:22,  1.10it/s]
 20%|██        | 40/196 [00:37<02:13,  1.17it/s]
 21%|██        | 41/196 [00:38<02:02,  1.27it/s]
 21%|██▏       | 42/196 [00:39<01:57,  1.31it/s]
 22%|██▏       | 43/196 [00:39<01:55,  1.33it/s]
 22%|██▏       | 44/196 [00:40<01:50,  1.38it/s]
 23%|██▎       | 45/196 [00:41<01:44,  1.45it/s]
 23%|██▎       | 46/196 [00:41<01:39,  1.51it/s]
 24%|██▍       | 47/196 [00:42<01:37,  1.52it/s]
 24%|██▍       | 48/196 [00:43<01:34,  1.57it/s]
 25%|██▌       | 49/196 [00:43<01:33,  1.58it/s]
 26%|██▌       | 50/196 [00:44<01:31,  1.60it/s]
 26%|██▌       | 51/196 [00:44<01:29,  1.62it/s]
 27%|██▋       | 52/196 [00:45<01:30,  1.59it/s]
 27%|██▋       | 53/196 [00:46<01:30,  1.58it/s]
 28%|██▊       | 54/196 [00:46<01:30,  1.57it/s]
 28%|██▊       | 55/196 [00:47<01:39,  1.42it/s]
 29%|██▊       | 56/196 [00:48<01:46,  1.32it/s]
 29%|██▉       | 57/196 [00:49<01:51,  1.24it/s]
 30%|██▉       | 58/196 [00:50<01:52,  1.22it/s]
 30%|███       | 59/196 [00:51<01:50,  1.24it/s]
 31%|███       | 60/196 [00:51<01:38,  1.38it/s]
 31%|███       | 61/196 [00:52<01:32,  1.45it/s]
 32%|███▏      | 62/196 [00:52<01:34,  1.42it/s]
 32%|███▏      | 63/196 [00:53<01:35,  1.39it/s]
 33%|███▎      | 64/196 [00:54<01:34,  1.39it/s]
 33%|███▎      | 65/196 [00:55<01:33,  1.40it/s]
 34%|███▎      | 66/196 [00:55<01:37,  1.33it/s]
 34%|███▍      | 67/196 [00:56<01:39,  1.29it/s]
 35%|███▍      | 68/196 [00:57<01:50,  1.15it/s]
 35%|███▌      | 69/196 [00:58<01:48,  1.17it/s]
 36%|███▌      | 70/196 [00:59<01:41,  1.24it/s]
 36%|███▌      | 71/196 [01:00<01:35,  1.31it/s]
 37%|███▋      | 72/196 [01:00<01:29,  1.38it/s]
 37%|███▋      | 73/196 [01:01<01:22,  1.49it/s]
 38%|███▊      | 74/196 [01:01<01:18,  1.56it/s]
 38%|███▊      | 75/196 [01:02<01:16,  1.59it/s]
 39%|███▉      | 76/196 [01:03<01:14,  1.62it/s]
 39%|███▉      | 77/196 [01:03<01:17,  1.54it/s]
 40%|███▉      | 78/196 [01:04<01:18,  1.50it/s]
 40%|████      | 79/196 [01:05<01:17,  1.52it/s]
 41%|████      | 80/196 [01:05<01:20,  1.43it/s]
 41%|████▏     | 81/196 [01:06<01:21,  1.41it/s]
 42%|████▏     | 82/196 [01:07<01:20,  1.42it/s]
 42%|████▏     | 83/196 [01:08<01:22,  1.37it/s]
 43%|████▎     | 84/196 [01:08<01:21,  1.37it/s]
 43%|████▎     | 85/196 [01:09<01:21,  1.36it/s]
 44%|████▍     | 86/196 [01:10<01:22,  1.33it/s]
 44%|████▍     | 87/196 [01:11<01:22,  1.32it/s]
 45%|████▍     | 88/196 [01:11<01:23,  1.29it/s]
 45%|████▌     | 89/196 [01:12<01:24,  1.26it/s]
 46%|████▌     | 90/196 [01:13<01:22,  1.28it/s]
 46%|████▋     | 91/196 [01:14<01:18,  1.35it/s]
 47%|████▋     | 92/196 [01:14<01:14,  1.39it/s]
 47%|████▋     | 93/196 [01:15<01:18,  1.32it/s]
 48%|████▊     | 94/196 [01:16<01:16,  1.33it/s]
 48%|████▊     | 95/196 [01:17<01:15,  1.34it/s]
 49%|████▉     | 96/196 [01:18<01:17,  1.29it/s]
 49%|████▉     | 97/196 [01:18<01:13,  1.34it/s]
 50%|████��     | 98/196 [01:19<01:14,  1.31it/s]
 51%|█████     | 99/196 [01:20<01:08,  1.41it/s]
 51%|█████     | 100/196 [01:20<01:00,  1.58it/s]
 52%|█████▏    | 101/196 [01:21<00:59,  1.60it/s]
 52%|█████▏    | 102/196 [01:21<01:02,  1.49it/s]
 53%|█████▎    | 103/196 [01:22<01:08,  1.35it/s]
 53%|█████▎    | 104/196 [01:23<01:15,  1.21it/s]
 54%|█████▎    | 105/196 [01:24<01:16,  1.20it/s]
 54%|█████▍    | 106/196 [01:25<01:14,  1.22it/s]
 55%|█████▍    | 107/196 [01:26<01:08,  1.29it/s]
 55%|█████▌    | 108/196 [01:26<01:01,  1.44it/s]
 56%|█████▌    | 109/196 [01:27<00:58,  1.50it/s]
 56%|█████▌    | 110/196 [01:27<00:56,  1.51it/s]
 57%|█████▋    | 111/196 [01:28<00:56,  1.50it/s]
 57%|█████▋    | 112/196 [01:29<00:58,  1.44it/s]
 58%|█████▊    | 113/196 [01:30<00:56,  1.46it/s]
 58%|█████▊    | 114/196 [01:30<00:52,  1.55it/s]
 59%|█████▊    | 115/196 [01:31<00:51,  1.57it/s]
 59%|█████▉    | 116/196 [01:31<00:50,  1.58it/s]
 60%|█████▉    | 117/196 [01:32<00:46,  1.69it/s]
 60%|██████    | 118/196 [01:32<00:42,  1.85it/s]
 61%|██████    | 119/196 [01:33<00:45,  1.71it/s]
 61%|██████    | 120/196 [01:34<00:46,  1.65it/s]
 62%|██████▏   | 121/196 [01:34<00:46,  1.62it/s]
 62%|██████▏   | 122/196 [01:35<00:47,  1.57it/s]
 63%|██████▎   | 123/196 [01:35<00:45,  1.61it/s]
 63%|██████▎   | 124/196 [01:36<00:46,  1.56it/s]
 64%|██████▍   | 125/196 [01:37<00:45,  1.54it/s]
 64%|██████▍   | 126/196 [01:38<00:51,  1.36it/s]
 65%|██████▍   | 127/196 [01:38<00:49,  1.40it/s]
 65%|██████▌   | 128/196 [01:39<00:46,  1.46it/s]
 66%|██████▌   | 129/196 [01:40<00:45,  1.48it/s]
 66%|██████▋   | 130/196 [01:40<00:45,  1.45it/s]
 67%|██████▋   | 131/196 [01:41<00:44,  1.45it/s]
 67%|██████▋   | 132/196 [01:42<00:41,  1.55it/s]
 68%|██████▊   | 133/196 [01:42<00:40,  1.56it/s]
 68%|██████▊   | 134/196 [01:43<00:40,  1.51it/s]
 69%|██████▉   | 135/196 [01:44<00:39,  1.54it/s]
 69%|██████▉   | 136/196 [01:44<00:38,  1.55it/s]
 70%|██████▉   | 137/196 [01:45<00:38,  1.55it/s]
 70%|███████   | 138/196 [01:46<00:37,  1.57it/s]
 71%|███████   | 139/196 [01:46<00:37,  1.52it/s]
 71%|███████▏  | 140/196 [01:47<00:35,  1.57it/s]
 72%|███████▏  | 141/196 [01:47<00:34,  1.60it/s]
 72%|███████▏  | 142/196 [01:48<00:34,  1.55it/s]
 73%|███████▎  | 143/196 [01:49<00:35,  1.49it/s]
 73%|███████▎  | 144/196 [01:49<00:33,  1.55it/s]
 74%|███████▍  | 145/196 [01:50<00:30,  1.67it/s]
 74%|███████▍  | 146/196 [01:50<00:28,  1.73it/s]
 75%|███████▌  | 147/196 [01:51<00:28,  1.73it/s]
 76%|███████▌  | 148/196 [01:52<00:28,  1.71it/s]
 76%|███████▌  | 149/196 [01:52<00:26,  1.80it/s]
 77%|███████▋  | 150/196 [01:53<00:27,  1.64it/s]
 77%|███████▋  | 151/196 [01:54<00:28,  1.59it/s]
 78%|███████▊  | 152/196 [01:54<00:27,  1.59it/s]
 78%|███████▊  | 153/196 [01:55<00:27,  1.58it/s]
 79%|███████▊  | 154/196 [01:55<00:26,  1.58it/s]
 79%|███████▉  | 155/196 [01:56<00:27,  1.48it/s]
 80%|███████▉  | 156/196 [01:57<00:29,  1.34it/s]
 80%|████████  | 157/196 [01:58<00:30,  1.29it/s]
 81%|████████  | 158/196 [01:58<00:26,  1.43it/s]
 81%|████████  | 159/196 [01:59<00:24,  1.52it/s]
 82%|████████▏ | 160/196 [02:00<00:23,  1.56it/s]
 82%|████████▏ | 161/196 [02:00<00:22,  1.53it/s]
 83%|████████▎ | 162/196 [02:01<00:21,  1.55it/s]
 83%|████████▎ | 163/196 [02:02<00:20,  1.58it/s]
 84%|████████▎ | 164/196 [02:02<00:20,  1.58it/s]
 84%|████████▍ | 165/196 [02:03<00:20,  1.53it/s]
 85%|████████▍ | 166/196 [02:04<00:19,  1.56it/s]
 85%|████████▌ | 167/196 [02:04<00:18,  1.60it/s]
 86%|████████▌ | 168/196 [02:05<00:16,  1.68it/s]
 86%|████████▌ | 169/196 [02:05<00:16,  1.59it/s]
 87%|████████▋ | 170/196 [02:06<00:17,  1.51it/s]
 87%|████████▋ | 171/196 [02:07<00:16,  1.52it/s]
 88%|████████▊ | 172/196 [02:07<00:16,  1.49it/s]
 88%|████████▊ | 173/196 [02:08<00:15,  1.49it/s]
 89%|████████▉ | 174/196 [02:09<00:15,  1.42it/s]
 89%|████████▉ | 175/196 [02:10<00:18,  1.16it/s]
 90%|████████▉ | 176/196 [02:12<00:25,  1.28s/it]
 90%|█████████ | 177/196 [02:14<00:27,  1.45s/it]
 91%|█████████ | 178/196 [02:16<00:29,  1.62s/it]
 91%|█████████▏| 179/196 [02:18<00:27,  1.63s/it]
 92%|█████████▏| 180/196 [02:19<00:21,  1.34s/it]
 92%|█████████▏| 181/196 [02:19<00:17,  1.14s/it]
 93%|█████████▎| 182/196 [02:20<00:13,  1.01it/s]
 93%|█████████▎| 183/196 [02:21<00:12,  1.02it/s]
 94%|█████████▍| 184/196 [02:21<00:10,  1.13it/s]
 94%|█████████▍| 185/196 [02:22<00:09,  1.17it/s]
 95%|█████████▍| 186/196 [02:23<00:08,  1.16it/s]
 95%|█████████▌| 187/196 [02:24<00:07,  1.26it/s]
 96%|█████████▌| 188/196 [02:24<00:06,  1.33it/s]
 96%|█████████▋| 189/196 [02:25<00:05,  1.36it/s]
 97%|█████████▋| 190/196 [02:26<00:04,  1.44it/s]
 97%|█████████▋| 191/196 [02:26<00:03,  1.52it/s]
 98%|█████████▊| 192/196 [02:27<00:02,  1.50it/s]
 98%|█████████▊| 193/196 [02:28<00:02,  1.48it/s]
 99%|█████████▉| 194/196 [02:28<00:01,  1.51it/s]
 99%|█████████▉| 195/196 [02:29<00:00,  1.56it/s]
+Printing predictions for a few samples:
+Sample 1:
+  Reference: लिबर ऑफिस <cs> impress </cs> में एक प्रस्तुति <cs> document </cs> बनाना और बुनियादी <cs> formatting </cs> के इस <cs> spoken  tutorial </cs> में आपका स्वागत है
+######
+
+
+  Prediction: s> liber  oफis  impres </cs> में एक प्रस्तुति > doकumnt /cs> बनाना और बुनियादी c> foरmatig </cs> के इस > pोken  tयutोरil /cs> में आपका सवाैहैहैह
+
+
+
+Sample 2:
+  Reference: इस <cs> tutorial </cs> में हम <cs> impress  window </cs> के भागों के बारे में सीखेंगे और कैसे स्लाइड इन्सर्ट करें और कॉपी करें फॉन्ट तथा फॉन्ट को फॉर्मेट करना सीखेंगे
+######
+
+
+  Prediction: इस <cs> tutoral </cs> में हम <s> eimpres < winडो </cs> के भागों के बारे में सीखेंगे और कैसे <cs> slide < insert </cs> करें और <cs> cope </c> cs> forn </cs> तथा <cs> font </cs> को <cs> foरmat </cs> करना सीखेंगे
+
+
+
+Sample 3:
+  Reference: यहाँ हम अपने ऑपरेटिंग सिस्टम के रूप में gnu/linux और लिबरऑफिस वर्जन <cs> 334 </cs> का उपयोग कर रहे हैं
+######
+
+
+  Prediction: यहाँ हम अपने <cs> oprेting  सistem </cs> के रूप में <cs> gnu  linx </cs> और cs> liber  ofis vergons </cs>s> </c> का उपयोग कर रह हैंं
+
+
+
+Sample 4:
+  Reference: चलिए अपनी प्रस्तुति प्रेजैटेशन <cs> sample  impress  open </cs> करते हैं जिसे पिछले <cs> tutorial </cs> में बनाया था
+######
+
+
+  Prediction: चलिए अपनी प्रस्तुति <cs> sampl  impcs  open </cs> करते हैं जिसे पिछले ं
+
+
+
+Sample 5:
+  Reference: चलिए देखते हैं कि <cs> screen </cs> पर क्या क्या है
+######
+
+
+  Prediction: चलिए देखते हैं कि sक्रीन </cs> पर क्या क्या है
+
+
+
+last Reference string यह स्क्रिप्ट लता द्वारा अनुवादित है आईआईटी मुंबई की ओर से मैं रवि कुमार अब आपसे विदा लेता हूँहमसे जुड़ने के लिए धन्यवाद
+
+
+last prediction string >लता द्वारा अनुवादित है आईआईटी मुमंबई की ओर से मैं रविकुमार अब आपसे विदा लेता हूँ हमसे जु़ने के लिए धन्यवाद
+***** eval metrics *****
+  epoch                   =        9.6
+  eval_cer                =     0.4133
+  eval_loss               =     1.9947
+  eval_runtime            = 0:02:38.28
+  eval_samples            =       3136
+  eval_samples_per_second =     19.813
+  eval_steps_per_second   =      1.238
+  eval_wer                =     0.6195
+

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d6e542b15e77cd1bb8a7693945d939d7b6324a0178a2a54a151c7870a9a3a70
+oid sha256:11123f516d513f049ccf8eca7a64f45fe94183f0c16009563846a3cba3ca71e4
 size 1262426580

predictionswtags_indicw2v_ad0_3_hd_02_featd_0_2_lr6e-4_warmup2000_s300_shuff100.txt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a83a46fa419ec5d9c1de83cff6b5473d12a65029127b92cbbd408597d51997f5
-size 30012168
+oid sha256:b80776faf14765af4462d6d31aaa0c48e2ef5a0992cf1aa97bdf95d9d8e25c7c
+size 30974098

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 5.6,
-    "total_flos": 2.158464150901847e+19,
-    "train_loss": 1.3790143143790108,
-    "train_runtime": 12492.843,
+    "epoch": 9.6,
+    "total_flos": 3.700768773245485e+19,
+    "train_loss": 2.825458660195271,
+    "train_runtime": 12931.7591,
     "train_samples": 20000,
-    "train_samples_per_second": 8.965,
-    "train_steps_per_second": 0.28
+    "train_samples_per_second": 14.847,
+    "train_steps_per_second": 0.464
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:360e5454f79c3f727fc27756fff70b6813ee1d83db4d96e1406670ae6b1e3e02
+oid sha256:83f52cedbc94f4a9afc031c9289677879e1c4f30bc20bbb9df0588e106379758
 size 5496