learning_rate: 0.0004
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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
lr_scheduler_warmup_steps: 132
num_epochs: 30
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.9017	1.1364	100	0.3934	57.5514
0.6485	2.2727	200	0.3226	93.0295
0.3384	3.4091	300	0.2814	91.7784
0.2732	4.5455	400	0.2201	86.9526
0.2094	5.6818	500	0.2057	78.6416
0.1781	6.8182	600	0.1592	65.6836
0.1295	7.9545	700	0.1294	60.8579
0.092	9.0909	800	0.0881	42.8061
0.0625	10.2273	900	0.0956	53.0831
0.0663	11.3636	1000	0.0673	47.3637
0.037	12.5	1100	0.0712	42.9848
0.0282	13.6364	1200	0.1005	48.8829
0.0193	14.7727	1300	0.0705	37.8910
0.0124	15.9091	1400	0.0713	34.3164
0.0088	17.0455	1500	0.0650	30.4736
0.0114	18.1818	1600	0.0464	34.0483
0.008	19.3182	1700	0.0550	30.1162
0.0018	20.4545	1800	0.0623	25.5585
0.0005	21.5909	1900	0.0608	27.7033
0.0	22.7273	2000	0.0602	26.9884
0.0	23.8636	2100	0.0606	26.9884
0.0	25.0	2200	0.0608	26.8990
0.0	26.1364	2300	0.0610	26.6309
0.0	27.2727	2400	0.0611	26.4522
0.0	28.4091	2500	0.0612	26.3628
0.0	29.5455	2600	0.0613	26.3628

susmitabhatt/whisper-a-nomi

whisper-a-nomi