learning_rate: 1e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
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: 3
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Sacrebleu
1.7235	0.2834	500	1.0139	36.7530
1.1783	0.5669	1000	0.7031	42.9010
0.9599	0.8503	1500	0.5852	46.7438
0.7981	1.1338	2000	0.5062	49.4535
0.6947	1.4172	2500	0.4498	51.3923
0.6415	1.7007	3000	0.4203	52.4852
0.61	1.9841	3500	0.4040	53.0134
0.5536	2.2676	4000	0.3863	53.8282
0.5308	2.5510	4500	0.3710	54.2751
0.5218	2.8345	5000	0.3689	54.4066

tminhtri1910/bartpho-vietnamese-correction

bartpho-vietnamese-correction