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
0.3723	0.4678	200	1.1080	0.2455	0.7857
0.4093	0.9357	400	0.9789	0.2540	0.7605
0.3411	1.4023	600	1.0196	0.2296	0.7190
0.3298	1.8702	800	0.9632	0.2361	0.7264
0.2142	2.3368	1000	1.0632	0.2261	0.7062
0.2039	2.8047	1200	1.0578	0.2262	0.7279
0.1415	3.2713	1400	1.0888	0.2200	0.6962
0.1275	3.7392	1600	1.0863	0.2196	0.7045
0.1145	4.2058	1800	1.1901	0.2152	0.6977
0.0968	4.6737	2000	1.1667	0.2158	0.6987

ctaguchi/ssc-lke-xlsr300m-model-mix-adapt-max-lowlr

ssc-lke-xlsr300m-model-mix-adapt-max-lowlr