learning_rate: 3e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 8
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
lr_scheduler_warmup_steps: 400
num_epochs: 24
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer
1.769	0.9880	72	0.8737	0.0023	0.5230
0.5492	1.9880	144	0.4324	0.0023	0.2642
0.2924	2.9880	216	0.3604	0.0023	0.2240
0.1725	3.9880	288	0.3470	0.0023	0.2862
0.0972	4.9880	360	0.3579	0.0023	0.2654
0.0557	5.9880	432	0.3763	0.0023	0.2123
0.0358	6.9880	504	0.3877	0.0023	0.2779
0.0249	7.9880	576	0.3973	0.0023	0.3861
0.0197	8.9880	648	0.4276	0.0023	0.2077
0.0117	9.9880	720	0.4091	0.0023	0.2233
0.0088	10.9880	792	0.4025	0.0023	0.2202
0.0055	11.9880	864	0.4345	0.0023	0.2106
0.0051	12.9880	936	0.4417	0.0023	0.2220
0.0028	13.9880	1008	0.4082	0.0023	0.2356
0.0018	14.9880	1080	0.3936	0.0023	0.2155
0.0017	15.9880	1152	0.4010	0.0023	0.2065
0.0013	16.9880	1224	0.4107	0.0023	0.2013
0.0009	17.9880	1296	0.4117	0.0023	0.1957
0.0008	18.9880	1368	0.4086	0.0023	0.1949
0.0005	19.9880	1440	0.4157	0.0023	0.1947
0.0004	20.9880	1512	0.4176	0.0023	0.1941
0.0004	21.9880	1584	0.4206	0.0023	0.1977
0.0004	22.9880	1656	0.4127	0.0023	0.1969
0.0004	23.9880	1728	0.4256	0.0023	0.1969

dhaval0108/ap-BN69p5ihH8dL4VBCKc17cb

ap-BN69p5ihH8dL4VBCKc17cb