learning_rate: 0.0003
train_batch_size: 1
eval_batch_size: 6
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 2
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
0.5899	0.2972	200	0.5503	0.1630	0.6628
0.4825	0.5944	400	0.5527	0.1606	0.6548
0.4697	0.8915	600	0.5579	0.1617	0.6577
0.5698	1.1887	800	0.5385	0.1584	0.6539
0.4574	1.4859	1000	0.5379	0.1577	0.6451
0.4811	1.7831	1200	0.5351	0.1562	0.6520
0.4618	2.0802	1400	0.5335	0.1554	0.6404
0.5116	2.3774	1600	0.5364	0.1558	0.6507
0.4292	2.6746	1800	0.5324	0.1553	0.6417
0.4965	2.9718	2000	0.5251	0.1566	0.6413
0.4014	3.2689	2200	0.5270	0.1544	0.6372
0.5158	3.5661	2400	0.5218	0.1543	0.6360
0.6268	3.8633	2600	0.5203	0.1542	0.6362
0.4544	4.1605	2800	0.5208	0.1539	0.6333
0.4955	4.4577	3000	0.5194	0.1538	0.6300
0.4367	4.7548	3200	0.5183	0.1525	0.6285

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

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