learning_rate: 0.0004
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Use 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: 132
num_epochs: 50
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
4.5146	2.2727	200	2.5284	1.0
1.4558	4.5455	400	0.4417	0.6139
0.253	6.8182	600	0.2623	0.4254
0.1064	9.0909	800	0.3151	0.4013
0.0715	11.3636	1000	0.2867	0.3744
0.0575	13.6364	1200	0.3176	0.3584
0.0445	15.9091	1400	0.3323	0.3432
0.029	18.1818	1600	0.3549	0.3467
0.0291	20.4545	1800	0.2824	0.3458
0.0196	22.7273	2000	0.3145	0.3414
0.0205	25.0	2200	0.3013	0.3396
0.0168	27.2727	2400	0.3207	0.3387
0.0136	29.5455	2600	0.3306	0.3333
0.0109	31.8182	2800	0.3197	0.3324
0.0077	34.0909	3000	0.3511	0.3378
0.0082	36.3636	3200	0.3408	0.3342
0.0088	38.6364	3400	0.3354	0.3369
0.0066	40.9091	3600	0.3246	0.3360
0.0035	43.1818	3800	0.3295	0.3369
0.0044	45.4545	4000	0.3220	0.3369
0.0049	47.7273	4200	0.3196	0.3378
0.0029	50.0	4400	0.3211	0.3369

susmitabhatt/xlsr-a-nomi-ag

xlsr-a-nomi-ag