c_ll_arafix_slurm.out

This job can be monitored from: https://job.c3se.chalmers.se/alvis/4059586
Using c_ll_comb91k_10_arafix config...
Loading new model openai/whisper-large-v2 (in 8bit for PEFT)...
trainable params: 10,485,760 || all params: 1,553,790,720 || trainable%: 0.6749
Loading ['ara'] (limit: [1000]) from asc-train.
Resampling audio...
Samples asc train: 1000
Loading ['ja', 'pl', 'mt', 'hu', 'fi', 'el', 'ta'] (limit: [1000, 1000, 1000, 1000, 1000, 1000, 1000]) from multipa-train.
Resampling audio...
Samples multipa train: 7000
Loading ['cmn'] (limit: [1000]) from thchs-train.
Resampling audio...
Samples thchs train: 1000
Creating input values and labels...
Loading ['ara'] (limit: [50]) from asc-dev.
Resampling audio...
Samples asc dev: 50
Loading ['ja', 'pl', 'mt', 'hu', 'fi', 'el', 'ta'] (limit: [50, 50, 50, 50, 50, 50, 50]) from multipa-dev.
Resampling audio...
Samples multipa validation: 350
Loading ['cmn'] (limit: [50]) from thchs-dev.
Resampling audio...
Samples thchs dev: 50
Creating input values and labels...
--------------------------------------------------------------------------------
Start fine-tuning...
{'loss': 2.6917, 'grad_norm': 1.791174054145813, 'learning_rate': 1e-05, 'epoch': 0.0}
{'loss': 0.7537, 'grad_norm': 0.42204609513282776, 'learning_rate': 0.0008610687022900763, 'epoch': 1.1}
{'eval_loss': 0.5796585083007812, 'eval_runtime': 102.9977, 'eval_samples_per_second': 4.359, 'eval_steps_per_second': 0.553, 'epoch': 1.1}
{'loss': 0.2638, 'grad_norm': 0.2676902711391449, 'learning_rate': 0.0006458015267175574, 'epoch': 3.1}
{'eval_loss': 0.4017384648323059, 'eval_runtime': 100.6276, 'eval_samples_per_second': 4.462, 'eval_steps_per_second': 0.566, 'epoch': 3.1}
{'loss': 0.1532, 'grad_norm': 0.27857616543769836, 'learning_rate': 0.00043053435114503817, 'epoch': 5.1}
{'eval_loss': 0.40539106726646423, 'eval_runtime': 108.1949, 'eval_samples_per_second': 4.15, 'eval_steps_per_second': 0.527, 'epoch': 5.1}
{'loss': 0.0909, 'grad_norm': 0.25622501969337463, 'learning_rate': 0.00021526717557251909, 'epoch': 7.1}
{'eval_loss': 0.4510815143585205, 'eval_runtime': 101.1112, 'eval_samples_per_second': 4.441, 'eval_steps_per_second': 0.564, 'epoch': 7.1}
{'loss': 0.0535, 'grad_norm': 0.22360892593860626, 'learning_rate': 0.0, 'epoch': 9.1}
{'eval_loss': 0.4732421040534973, 'eval_runtime': 100.5466, 'eval_samples_per_second': 4.466, 'eval_steps_per_second': 0.567, 'epoch': 9.1}
{'train_runtime': 25095.8699, 'train_samples_per_second': 3.596, 'train_steps_per_second': 0.056, 'train_loss': 0.2643802025639419, 'epoch': 9.1}
----------------------------------- COMPLETE 25096.07261276245 -----------------------------------