|
|
This job can be monitored from: https://job.c3se.chalmers.se/alvis/4073855 |
|
|
Using d_ll_araelmt1k config... |
|
|
Loading new model openai/whisper-large-v2 (in 8bit for PEFT)... |
|
|
trainable params: 10,485,760 || all params: 1,553,790,720 || trainable |
|
|
{'asc': {'languages': 'ara', 'limit': [1000]}, 'multipa': {'languages': ['el', 'mt'], 'limit': [1000]}} |
|
|
Loading ['ara'] (limit: [1000]) from asc-train. |
|
|
Resampling audio... |
|
|
Samples asc train: 1000 |
|
|
Loading ['el', 'mt'] (limit: [1000, 1000]) from multipa-train. |
|
|
Resampling audio... |
|
|
Samples multipa train: 2000 |
|
|
Creating input values and labels... |
|
|
{'asc': {'languages': 'ara', 'limit': [150]}, 'multipa': {'languages': ['el', 'mt'], 'limit': [150]}} |
|
|
Loading ['ara'] (limit: [150]) from asc-dev. |
|
|
Resampling audio... |
|
|
Samples asc dev: 150 |
|
|
Loading ['el', 'mt'] (limit: [150, 150]) from multipa-dev. |
|
|
Resampling audio... |
|
|
Samples multipa validation: 300 |
|
|
Creating input values and labels... |
|
|
-------------------------------------------------------------------------------- |
|
|
Start fine-tuning... |
|
|
{'loss': 3.5833, 'grad_norm': 2.3249564170837402, 'learning_rate': 1e-05, 'epoch': 0.0} |
|
|
{'loss': 1.4683, 'grad_norm': 0.4263363778591156, 'learning_rate': 0.00094, 'epoch': 0.1} |
|
|
{'loss': 0.4344, 'grad_norm': 0.5972639322280884, 'learning_rate': 0.0008952380952380953, 'epoch': 1.1} |
|
|
{'eval_loss': 0.3692754805088043, 'eval_runtime': 71.2145, 'eval_samples_per_second': 6.319, 'eval_steps_per_second': 0.8, 'epoch': 1.1} |
|
|
{'loss': 0.2858, 'grad_norm': 0.6023200750350952, 'learning_rate': 0.0007833333333333334, 'epoch': 2.1} |
|
|
{'loss': 0.1875, 'grad_norm': 0.4565546214580536, 'learning_rate': 0.0006714285714285714, 'epoch': 3.1} |
|
|
{'eval_loss': 0.3102695643901825, 'eval_runtime': 69.8732, 'eval_samples_per_second': 6.44, 'eval_steps_per_second': 0.816, 'epoch': 3.1} |
|
|
{'loss': 0.1162, 'grad_norm': 0.43287548422813416, 'learning_rate': 0.0005595238095238096, 'epoch': 4.1} |
|
|
{'loss': 0.0717, 'grad_norm': 0.24221746623516083, 'learning_rate': 0.00044761904761904766, 'epoch': 5.1} |
|
|
{'eval_loss': 0.30600059032440186, 'eval_runtime': 70.275, 'eval_samples_per_second': 6.403, 'eval_steps_per_second': 0.811, 'epoch': 5.1} |
|
|
{'loss': 0.0363, 'grad_norm': 0.2153695970773697, 'learning_rate': 0.0003357142857142857, 'epoch': 6.1} |
|
|
{'loss': 0.0202, 'grad_norm': 0.14185360074043274, 'learning_rate': 0.00022380952380952383, 'epoch': 7.1} |
|
|
{'eval_loss': 0.32697898149490356, 'eval_runtime': 69.9983, 'eval_samples_per_second': 6.429, 'eval_steps_per_second': 0.814, 'epoch': 7.1} |
|
|
{'loss': 0.0134, 'grad_norm': 0.061574213206768036, 'learning_rate': 0.00011190476190476191, 'epoch': 8.1} |
|
|
{'loss': 0.0101, 'grad_norm': 0.06894426047801971, 'learning_rate': 0.0, 'epoch': 9.1} |
|
|
{'eval_loss': 0.34040552377700806, 'eval_runtime': 70.2499, 'eval_samples_per_second': 6.406, 'eval_steps_per_second': 0.811, 'epoch': 9.1} |
|
|
{'train_runtime': 6384.8945, 'train_samples_per_second': 4.711, 'train_steps_per_second': 0.147, 'train_loss': 0.26664772788260843, 'epoch': 9.1} |
|
|
----------------------------------- COMPLETE 6385.079031229019 ----------------------------------- |
|
|
|
