learning_rate: 0.0001
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: 30
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	2	1.9291	0.1538
No log	2.0	4	1.9180	0.1538
No log	3.0	6	1.7831	0.5385
No log	4.0	8	1.5618	0.6154
No log	5.0	10	1.4707	0.8462
No log	6.0	12	1.3069	0.7692
No log	7.0	14	1.1631	0.9231
No log	8.0	16	1.0261	1.0
No log	9.0	18	0.9246	1.0
No log	10.0	20	0.8220	1.0
No log	11.0	22	0.7281	1.0
No log	12.0	24	0.6651	1.0
No log	13.0	26	0.6156	0.9231
No log	14.0	28	0.5512	1.0
No log	15.0	30	0.5101	1.0
No log	16.0	32	0.4734	1.0
No log	17.0	34	0.4246	1.0
No log	18.0	36	0.3925	1.0
No log	19.0	38	0.3662	1.0
No log	20.0	40	0.3428	1.0
No log	21.0	42	0.3286	1.0
No log	22.0	44	0.3136	1.0
No log	23.0	46	0.3008	1.0
No log	24.0	48	0.2900	1.0
No log	25.0	50	0.2813	1.0
No log	26.0	52	0.2751	1.0
No log	27.0	54	0.2708	1.0
No log	28.0	56	0.2675	1.0
No log	29.0	58	0.2652	1.0
No log	30.0	60	0.2642	1.0

gharam12/korean_kws2

korean_kws2