learning_rate: 3e-05
train_batch_size: 12
eval_batch_size: 48
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 48
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
num_epochs: 10
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Char Acc
1.1208	1.0	750	0.2115	0.3068	0.3585	0.6437
0.2822	2.0	1500	0.1456	0.2987	0.3327	0.6723
0.1715	3.0	2250	0.1287	0.2973	0.3267	0.6797
0.1166	4.0	3000	0.1225	0.2961	0.3220	0.6831
0.0827	5.0	3750	0.1225	0.2943	0.3180	0.6883
0.0602	6.0	4500	0.1199	0.2945	0.3177	0.6896
0.0455	7.0	5250	0.1221	0.2943	0.3159	0.6891
0.0346	8.0	6000	0.1228	0.2940	0.3150	0.6911
0.0274	9.0	6750	0.1223	0.2936	0.3146	0.6909
0.0232	10.0	7500	0.1218	0.2935	0.3139	0.6912

qthuan2604/correction

correction