Training in progress, step 10000

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bfd24e8b882a313de234580db68819b96c5e17fda2cc398b7c6fc6c08aa02608
+oid sha256:70fcbc74e2643dd1fe96e9941f38709168bc70e9d0f45f0f0038d777222301a0
 size 151061672

wandb/run-20250217_214618-7bygcjmf/files/output.log

@@ -1695,5 +1695,197 @@ Retrying in 1s [Retry 1/5].
 [INFO|feature_extraction_utils.py:437] 2025-02-17 22:45:12,126 >> Feature extractor saved in ./preprocessor_config.json
 /home/tknika/xezpeleta/whisper/.venv/lib/python3.12/site-packages/datasets/formatting/torch_formatter.py:87: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
   return torch.tensor(value, **{**default_dtype, **self.torch_tensor_kwargs})
- 90%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏            | 9026/10000 [59:01<08:47,  1.85it/s]
+                                                                                                                                                                             
 {'loss': 0.1113, 'grad_norm': 3.3851749897003174, 'learning_rate': 4.083333333333333e-06, 'epoch': 0.9}
+{'loss': 0.1345, 'grad_norm': 3.7882187366485596, 'learning_rate': 3.9791666666666665e-06, 'epoch': 0.91}
+{'loss': 0.1739, 'grad_norm': 4.723577976226807, 'learning_rate': 3.874999999999999e-06, 'epoch': 0.91}
+{'loss': 0.173, 'grad_norm': 3.7890665531158447, 'learning_rate': 3.770833333333333e-06, 'epoch': 0.91}
+{'loss': 0.1381, 'grad_norm': 3.224851608276367, 'learning_rate': 3.6666666666666666e-06, 'epoch': 0.91}
+{'loss': 0.1291, 'grad_norm': 2.4666051864624023, 'learning_rate': 3.5624999999999998e-06, 'epoch': 0.92}
+{'loss': 0.1249, 'grad_norm': 3.090360403060913, 'learning_rate': 3.458333333333333e-06, 'epoch': 0.92}
+{'loss': 0.1292, 'grad_norm': 4.420276641845703, 'learning_rate': 3.354166666666666e-06, 'epoch': 0.92}
+{'loss': 0.1508, 'grad_norm': 4.0348663330078125, 'learning_rate': 3.25e-06, 'epoch': 0.92}
+{'loss': 0.1749, 'grad_norm': 4.426786422729492, 'learning_rate': 3.145833333333333e-06, 'epoch': 0.93}
+{'loss': 0.1932, 'grad_norm': 3.2825982570648193, 'learning_rate': 3.041666666666666e-06, 'epoch': 0.93}
+{'loss': 0.1182, 'grad_norm': 2.628441572189331, 'learning_rate': 2.9375e-06, 'epoch': 0.93}
+{'loss': 0.1196, 'grad_norm': 2.464374303817749, 'learning_rate': 2.833333333333333e-06, 'epoch': 0.93}
+{'loss': 0.0968, 'grad_norm': 2.491689682006836, 'learning_rate': 2.7291666666666662e-06, 'epoch': 0.94}
+{'loss': 0.1251, 'grad_norm': 3.4698681831359863, 'learning_rate': 2.625e-06, 'epoch': 0.94}
+{'loss': 0.1507, 'grad_norm': 3.9469587802886963, 'learning_rate': 2.520833333333333e-06, 'epoch': 0.94}
+{'loss': 0.169, 'grad_norm': 3.5354082584381104, 'learning_rate': 2.4166666666666663e-06, 'epoch': 0.94}
+{'loss': 0.2939, 'grad_norm': 2.7507169246673584, 'learning_rate': 2.3125e-06, 'epoch': 0.94}
+{'loss': 0.1602, 'grad_norm': 3.2346575260162354, 'learning_rate': 2.208333333333333e-06, 'epoch': 0.95}
+{'loss': 0.134, 'grad_norm': 4.097684860229492, 'learning_rate': 2.1041666666666667e-06, 'epoch': 0.95}
+{'loss': 0.1553, 'grad_norm': 4.1434855461120605, 'learning_rate': 2e-06, 'epoch': 0.95}
+{'loss': 0.171, 'grad_norm': 3.8769595623016357, 'learning_rate': 1.8958333333333331e-06, 'epoch': 0.95}
+{'loss': 0.1935, 'grad_norm': 4.308934211730957, 'learning_rate': 1.7916666666666665e-06, 'epoch': 0.96}
+{'loss': 0.1496, 'grad_norm': 3.0821316242218018, 'learning_rate': 1.6874999999999997e-06, 'epoch': 0.96}
+{'loss': 0.1223, 'grad_norm': 4.086252689361572, 'learning_rate': 1.5833333333333331e-06, 'epoch': 0.96}
+{'loss': 0.0933, 'grad_norm': 2.3455817699432373, 'learning_rate': 1.4791666666666663e-06, 'epoch': 0.96}
+{'loss': 0.0959, 'grad_norm': 2.302103042602539, 'learning_rate': 1.375e-06, 'epoch': 0.97}
+{'loss': 0.0963, 'grad_norm': 2.8981809616088867, 'learning_rate': 1.2708333333333334e-06, 'epoch': 0.97}
+{'loss': 0.096, 'grad_norm': 1.89694082736969, 'learning_rate': 1.1666666666666666e-06, 'epoch': 0.97}
+{'loss': 0.1005, 'grad_norm': 2.363546371459961, 'learning_rate': 1.0624999999999998e-06, 'epoch': 0.97}
+{'loss': 0.11, 'grad_norm': 2.9393630027770996, 'learning_rate': 9.583333333333334e-07, 'epoch': 0.98}
+{'loss': 0.153, 'grad_norm': 2.824051856994629, 'learning_rate': 8.541666666666666e-07, 'epoch': 0.98}
+{'loss': 0.1647, 'grad_norm': 3.817542552947998, 'learning_rate': 7.499999999999999e-07, 'epoch': 0.98}
+{'loss': 0.1499, 'grad_norm': 3.506502151489258, 'learning_rate': 6.458333333333332e-07, 'epoch': 0.98}
+{'loss': 0.148, 'grad_norm': 2.9996325969696045, 'learning_rate': 5.416666666666666e-07, 'epoch': 0.99}
+{'loss': 0.2303, 'grad_norm': 6.828978538513184, 'learning_rate': 4.375e-07, 'epoch': 0.99}
+{'loss': 0.195, 'grad_norm': 2.7112486362457275, 'learning_rate': 3.333333333333333e-07, 'epoch': 0.99}
+{'loss': 0.106, 'grad_norm': 3.023608684539795, 'learning_rate': 2.2916666666666666e-07, 'epoch': 0.99}
+{'loss': 0.1, 'grad_norm': 2.6858067512512207, 'learning_rate': 1.25e-07, 'epoch': 1.0}
+{'loss': 0.106, 'grad_norm': 2.3674516677856445, 'learning_rate': 2.083333333333333e-08, 'epoch': 1.0}
+***** Running Evaluation *****
+[INFO|trainer.py:4262] 2025-02-17 22:50:41,500 >>   Num examples: Unknown
+[INFO|trainer.py:4263] 2025-02-17 22:50:41,501 >>   Batch size = 16
+[INFO|trainer_utils.py:837] 2025-02-17 22:50:49,139 >> The following columns in the evaluation set don't have a corresponding argument in `WhisperForConditionalGeneration.forward` and have been ignored: input_length. If input_length are not expected by `WhisperForConditionalGeneration.forward`,  you can safely ignore this message.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:49,289 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:49,700 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:50,260 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:50,765 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:51,202 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:51,674 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:52,064 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:52,511 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:52,931 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:53,377 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:53,816 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:54,278 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:54,861 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:55,228 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:55,726 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:56,208 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:56,590 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:57,081 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:57,498 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:57,847 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:58,291 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:58,663 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:59,094 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:59,482 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:50:59,839 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:00,206 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:00,624 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:01,080 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:01,344 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:01,705 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:02,111 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:02,422 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:02,815 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:03,195 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:03,554 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:03,948 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:04,238 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:04,649 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:05,029 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:05,968 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:06,257 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:06,562 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:06,831 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:07,120 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:07,394 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:07,716 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:08,005 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:08,277 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:08,641 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:08,941 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:09,274 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:09,583 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:09,850 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:10,148 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:10,442 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:10,775 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:11,038 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:11,309 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:11,631 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:11,902 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:12,219 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:13,046 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:13,359 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:13,656 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:13,913 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:14,153 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:14,453 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:14,748 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:15,050 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:15,339 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:15,630 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:15,966 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:16,256 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:16,601 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:16,900 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:17,204 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:17,467 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:17,755 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:18,059 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:18,341 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:18,614 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:18,890 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:19,186 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:19,501 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:19,816 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:20,098 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:20,399 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:20,748 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:21,055 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:21,409 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:21,716 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:22,013 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:22,281 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:22,563 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:22,905 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:23,215 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:23,516 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:23,821 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:24,114 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:24,423 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:24,730 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:25,055 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:25,344 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:25,664 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:26,010 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:26,314 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:26,628 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:26,963 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:27,242 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:27,518 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:27,828 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:28,136 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:28,405 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:28,673 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:28,963 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:29,264 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:29,582 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:29,893 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:30,179 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:30,474 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:30,780 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:31,049 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:31,359 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:31,668 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:31,951 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:32,228 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:32,508 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:32,812 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:33,105 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:33,400 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:33,658 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+[INFO|generation_whisper.py:1844] 2025-02-17 22:51:33,894 >> Increase max_length from 225 to 228 since input is conditioned on previous segment.
+100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10000/10000 [1:05:22<00:00,  3.05it/s][INFO|trainer.py:3942] 2025-02-17 22:51:41,918 >> Saving model checkpoint to ./checkpoint-10000
+{'eval_loss': 0.2991788983345032, 'eval_wer': 15.72403477610545, 'eval_runtime': 60.4166, 'eval_samples_per_second': 34.825, 'eval_steps_per_second': 2.185, 'epoch': 1.0}
+[INFO|configuration_utils.py:423] 2025-02-17 22:51:41,919 >> Configuration saved in ./checkpoint-10000/config.json
+[INFO|configuration_utils.py:909] 2025-02-17 22:51:41,920 >> Configuration saved in ./checkpoint-10000/generation_config.json
+[INFO|modeling_utils.py:3040] 2025-02-17 22:51:42,130 >> Model weights saved in ./checkpoint-10000/model.safetensors
+[INFO|feature_extraction_utils.py:437] 2025-02-17 22:51:42,131 >> Feature extractor saved in ./checkpoint-10000/preprocessor_config.json
+[INFO|feature_extraction_utils.py:437] 2025-02-17 22:51:42,830 >> Feature extractor saved in ./preprocessor_config.json
+[INFO|trainer.py:2657] 2025-02-17 22:51:42,831 >>
+
+Training completed. Do not forget to share your model on huggingface.co/models =)
+
+
+[INFO|trainer.py:2895] 2025-02-17 22:51:42,832 >> Loading best model from ./checkpoint-8000 (score: 14.985509956062447).
+[WARNING|trainer.py:3032] 2025-02-17 22:51:42,895 >> There were missing keys in the checkpoint model loaded: ['proj_out.weight'].
+100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10000/10000 [1:05:23<00:00,  2.55it/s]
+{'train_runtime': 3924.8092, 'train_samples_per_second': 81.533, 'train_steps_per_second': 2.548, 'train_loss': 0.24547564173936845, 'epoch': 1.0}
+[INFO|trainer.py:4723] 2025-02-17 22:51:42,905 >> Waiting for the current checkpoint push to be finished, this might take a couple of minutes.

wandb/run-20250217_214618-7bygcjmf/run-7bygcjmf.wandb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dba80738c9117c0ef4fa22838cc5baab13cd7ac7f3c7efb51aeea05f9cb79184
-size 4325376
+oid sha256:4b8a57613cf10725c6769a6e7ad801d52e9a486bfd327b974ae5e778d6cdd2a6
+size 4947968