learning_rate: 5e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 32
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
num_epochs: 50
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer Score
5.2468	3.5398	50	4.5809	17.1528
0.7758	7.0796	100	0.4672	7.7810
0.0337	10.6195	150	0.0424	2.3531
0.0089	14.1593	200	0.0401	3.3039
0.0022	17.6991	250	0.0388	5.8557
0.0008	21.2389	300	0.0411	4.6740
0.0004	24.7788	350	0.0410	3.8676
0.0003	28.3186	400	0.0409	4.1766
0.0002	31.8584	450	0.0414	4.0136
0.0002	35.3982	500	0.0414	3.9779
0.0002	38.9381	550	0.0417	3.9542
0.0002	42.4779	600	0.0418	3.8913
0.0002	46.0177	650	0.0420	3.8183
0.0002	49.5575	700	0.0420	3.8081

osmanh/git-base-pokemon

git-base-pokemon