learning_rate: 5e-05
train_batch_size: 32
eval_batch_size: 32
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 64
optimizer: Use OptimizerNames.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
7.1475	3.5714	50	4.4300	0.8516
2.384	7.1429	100	0.6209	0.7782
0.3154	10.7143	150	0.2059	2.0570
0.1406	14.2857	200	0.1841	2.6379
0.0927	17.8571	250	0.1831	2.5492
0.062	21.4286	300	0.1891	2.6554
0.0423	25.0	350	0.1938	2.5195
0.0292	28.5714	400	0.1996	2.5295
0.0214	32.1429	450	0.2034	2.4541
0.0169	35.7143	500	0.2082	2.5956
0.0139	39.2857	550	0.2105	2.2852
0.0119	42.8571	600	0.2129	2.4300
0.0106	46.4286	650	0.2138	2.4355
0.01	50.0	700	0.2143	2.4047

tlstkr/git-base-clothes

git-base-clothes