learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 15
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
1.3249	0.6667	500	1.2381	53.9007
1.1851	1.3333	1000	1.1940	52.5907
1.1473	2.0	1500	1.1305	51.1462
1.1077	2.6667	2000	1.0999	49.8800
1.0843	3.3333	2500	1.0811	49.7477
1.0685	4.0	3000	1.0695	49.1241
1.0577	4.6667	3500	1.0631	48.4613
1.049	5.3333	4000	1.0608	48.7416
1.0439	6.0	4500	1.0461	48.2796
1.0381	6.6667	5000	1.0335	47.8040
1.0151	7.3333	5500	1.0227	46.4223
1.0012	8.0	6000	1.0136	47.0257
0.9982	8.6667	6500	1.0060	46.7341
0.9893	9.3333	7000	0.9997	45.5958
0.9866	10.0	7500	0.9938	46.2104
0.98	10.6667	8000	0.9892	45.6171
0.9707	11.3333	8500	0.9853	44.5785
0.9714	12.0	9000	0.9822	44.7299
0.9717	12.6667	9500	0.9798	44.6211
0.9618	13.3333	10000	0.9779	44.2331
0.9618	14.0	10500	0.9767	43.8529
0.9664	14.6667	11000	0.9761	43.7093

stealavie/whisper_finetuned_vi

whisper_finetuned_vi