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
4.3717	2.8633	200	2.5118	1.0
1.1028	5.7194	400	0.3058	0.5279
0.1672	8.5755	600	0.2953	0.4243
0.0896	11.4317	800	0.2936	0.3663
0.0506	14.2878	1000	0.3016	0.3299
0.0354	17.1439	1200	0.3198	0.3481
0.0294	20.0	1400	0.2850	0.3242
0.0208	22.8633	1600	0.2837	0.3265
0.0151	25.7194	1800	0.2923	0.3231
0.0134	28.5755	2000	0.3153	0.3208
0.0151	31.4317	2200	0.3107	0.3231
0.0067	34.2878	2400	0.3222	0.3208
0.0066	37.1439	2600	0.3356	0.3208
0.0069	40.0	2800	0.3233	0.3231
0.0056	42.8633	3000	0.3418	0.3208
0.0051	45.7194	3200	0.3428	0.3197
0.0035	48.5755	3400	0.3484	0.3208

susmitabhatt/xlsr-a-no-ag

xlsr-a-no-ag