learning_rate: 0.0005
train_batch_size: 1
eval_batch_size: 6
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 2
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: 100
num_epochs: 5
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.8263	0.4145	200	0.6574	0.1727	0.4751
0.7999	0.8290	400	0.6652	0.1709	0.4466
0.7696	1.2425	600	0.7026	0.1696	0.4384
0.7621	1.6570	800	0.6841	0.1704	0.4351
0.7157	2.0705	1000	0.7128	0.1679	0.4307
0.7382	2.4850	1200	0.6641	0.1645	0.4313
0.7064	2.8995	1400	0.6524	0.1652	0.4247
0.6775	3.3130	1600	0.6693	0.1651	0.4234
0.6622	3.7275	1800	0.6772	0.1631	0.4188
0.6245	4.1409	2000	0.6559	0.1621	0.4169
0.6626	4.5554	2200	0.6580	0.1612	0.4143
0.6232	4.9699	2400	0.6585	0.1611	0.4147

ctaguchi/ssc-aln-mms-model-mix-adapt-max3-devtrain

ssc-aln-mms-model-mix-adapt-max3-devtrain