learning_rate: 0.001
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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
1.3274	0.3265	200	1.3596	0.3034	0.7994
1.1117	0.6531	400	1.3266	0.2698	0.7272
1.0661	0.9796	600	1.2708	0.2526	0.6822
1.0701	1.3053	800	1.3483	0.2587	0.6955
0.9163	1.6318	1000	1.3488	0.2502	0.6549
0.9005	1.9584	1200	1.3639	0.2487	0.6600
0.8301	2.2841	1400	1.3388	0.2447	0.6339
0.8878	2.6106	1600	1.3057	0.2348	0.6285
0.8899	2.9371	1800	1.3062	0.2408	0.6295
0.8035	3.2629	2000	1.3022	0.2413	0.6372
0.8172	3.5894	2200	1.3599	0.2319	0.6133
0.7303	3.9159	2400	1.3201	0.2340	0.6187
0.793	4.2416	2600	1.3509	0.2305	0.6113
0.7757	4.5682	2800	1.3349	0.2367	0.6153
0.7047	4.8947	3000	1.3308	0.2323	0.6059

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

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