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
2.3038	0.5168	200	1.5889	0.1651	0.5510
3.1201	1.0336	400	1.4828	0.1601	0.5272
0.4433	1.5504	600	1.5661	0.1589	0.5287
0.8098	2.0672	800	1.5664	0.1591	0.5202
0.4093	2.5840	1000	1.5683	0.1572	0.5210
7.1871	3.1008	1200	1.4891	0.1553	0.5107
0.3387	3.6176	1400	1.6109	0.1546	0.5132
0.4889	4.1344	1600	1.5844	0.1527	0.5053
4.4192	4.6512	1800	1.5739	0.1541	0.5105

ctaguchi/ssc-el-CY-mms-model-mix-adapt-max3

ssc-el-CY-mms-model-mix-adapt-max3