learning_rate: 0.001
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
0.4961	0.4790	200	0.8817	0.1778	0.5910
0.4919	0.9581	400	0.8375	0.1776	0.5852
0.4513	1.4359	600	0.8275	0.1816	0.6014
0.4387	1.9150	800	0.8396	0.1791	0.5979
0.4225	2.3928	1000	0.8080	0.1786	0.5884
0.393	2.8719	1200	0.8259	0.1780	0.5956
0.4362	3.3497	1400	0.8368	0.1772	0.5747
0.366	3.8287	1600	0.8345	0.1763	0.5775
0.366	4.3066	1800	0.8136	0.1769	0.5831
0.354	4.7856	2000	0.8158	0.1745	0.5777

ctaguchi/ssc-tob-mms-model-mix-adapt-max2

ssc-tob-mms-model-mix-adapt-max2