learning_rate: 0.0003
train_batch_size: 2
eval_batch_size: 6
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 4
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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
0.7498	0.5731	200	1.1147	0.2298	0.5904
0.8273	1.1461	400	1.0470	0.2311	0.5910
0.8044	1.7192	600	0.9994	0.2265	0.5781
0.7486	2.2923	800	0.9808	0.2230	0.5678
0.6732	2.8653	1000	0.9598	0.2286	0.5797
0.6591	3.4384	1200	0.9658	0.2231	0.5656
0.6232	4.0115	1400	1.0106	0.2220	0.5613
0.6856	4.5845	1600	0.9892	0.2188	0.5550

ctaguchi/ssc-kcn-mms-model-mix-adapt-max-lowlr

ssc-kcn-mms-model-mix-adapt-max-lowlr