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
1.4784	0.9880	72	0.6310	0.0033	0.4579
0.3876	1.9880	144	0.3217	0.0033	0.1994
0.1931	2.9880	216	0.2909	0.0033	0.1876
0.0977	3.9880	288	0.3108	0.0033	0.1857
0.0531	4.9880	360	0.3348	0.0033	0.1885
0.0403	5.9880	432	0.3479	0.0033	0.2027
0.0303	6.9880	504	0.3298	0.0033	0.1897
0.0198	7.9880	576	0.3237	0.0033	0.1816
0.0119	8.9880	648	0.3407	0.0033	0.1820
0.009	9.9880	720	0.3427	0.0033	0.1807
0.0067	10.9880	792	0.3401	0.0033	0.2080
0.0058	11.9880	864	0.3441	0.0033	0.1603
0.0032	12.9880	936	0.3481	0.0033	0.1479
0.0027	13.9880	1008	0.3527	0.0033	0.1510
0.0014	14.9880	1080	0.3529	0.0033	0.1596
0.001	15.9880	1152	0.3683	0.0033	0.1516
0.0016	16.9880	1224	0.3571	0.0033	0.1548
0.0006	17.9880	1296	0.3681	0.0033	0.1543
0.0004	18.9880	1368	0.3683	0.0033	0.1605
0.0003	19.9880	1440	0.3687	0.0033	0.1595
0.0003	20.9880	1512	0.3695	0.0033	0.1590
0.0002	21.9880	1584	0.3658	0.0033	0.1587
0.0002	22.9880	1656	0.3726	0.0033	0.1581
0.0002	23.9880	1728	0.3596	0.0033	0.1581

dhaval0108/ap-Ev25J1npPqT6HIfz2iXjhN

ap-Ev25J1npPqT6HIfz2iXjhN