learning_rate: 5e-06
train_batch_size: 8
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Use 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: 200
num_epochs: 10

Training Loss	Epoch	Step	Validation Loss	Wer
0.5914	0.5038	500	0.2818	0.3268
0.6093	1.0071	1000	0.2764	0.3277
0.5349	1.5108	1500	0.2637	0.3196
0.4888	2.0141	2000	0.2613	0.3225
0.5106	2.5179	2500	0.2611	0.3178
0.4188	3.0212	3000	0.2532	0.3118
0.4391	3.5249	3500	0.2512	0.3090
0.3455	4.0282	4000	0.2460	0.3037
0.3495	4.5320	4500	0.2469	0.3026
0.3506	5.0353	5000	0.2452	0.3042
0.349	5.5390	5500	0.2474	0.2990
0.3143	6.0423	6000	0.2429	0.3000
0.2866	6.5461	6500	0.2443	0.2976
0.3469	7.0494	7000	0.2420	0.2980
0.3379	7.5531	7500	0.2409	0.2948
0.3262	8.0564	8000	0.2379	0.2939
0.2955	8.5602	8500	0.2393	0.2913
0.333	9.0635	9000	0.2396	0.2919
0.3258	9.5673	9500	0.2387	0.2914

SergeyAdamyan/outputs_finetune3

outputs_finetune3