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
0.5415	0.4920	200	0.5600	0.1618	0.4943
0.5541	0.9840	400	0.5432	0.1611	0.4941
0.5282	1.4748	600	0.5389	0.1541	0.4669
0.5117	1.9668	800	0.5340	0.1549	0.4759
0.4838	2.4576	1000	0.5371	0.1541	0.4689
0.4974	2.9496	1200	0.5348	0.1527	0.4646
0.4507	3.4403	1400	0.5267	0.1503	0.4536
0.4644	3.9323	1600	0.5283	0.1508	0.4592
0.4739	4.4231	1800	0.5228	0.1496	0.4553
0.4471	4.9151	2000	0.5232	0.1493	0.4536

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

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