learning_rate: 0.001
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Use adamw_torch 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: 4
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
2.6853	0.25	400	0.2931	0.4680
0.7427	0.5	800	0.2290	0.3743
0.6983	0.75	1200	0.1836	0.3390
0.6328	1.0	1600	0.1743	0.3172
0.61	1.25	2000	0.1607	0.2915
0.5941	1.5	2400	0.1536	0.2996
0.545	1.75	2800	0.1470	0.2787
0.5534	2.0	3200	0.1513	0.2754
0.5245	2.25	3600	0.1409	0.2661
0.5206	2.5	4000	0.1386	0.2676
0.505	2.75	4400	0.1313	0.2545
0.4647	3.0	4800	0.1249	0.2398
0.4754	3.25	5200	0.1225	0.2395
0.4508	3.5	5600	0.1194	0.2369
0.4629	3.75	6000	0.1152	0.2282
0.4443	4.0	6400	0.1139	0.2315

MatricariaV/kbd-1s-data

kbd-1s-data