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
1.1957	1.0	126	0.5566	0.0026	0.3097
0.3192	2.0	252	0.3838	0.0026	0.2868
0.1776	3.0	378	0.3642	0.0026	0.2321
0.1	4.0	504	0.3987	0.0026	0.2116
0.054	5.0	630	0.3833	0.0026	0.2825
0.0315	6.0	756	0.3903	0.0026	0.2350
0.018	7.0	882	0.3841	0.0026	0.2146
0.0131	8.0	1008	0.4031	0.0026	0.2207
0.007	9.0	1134	0.4072	0.0026	0.2043
0.0045	10.0	1260	0.4174	0.0026	0.2653
0.0056	11.0	1386	0.4117	0.0026	0.2158
0.0027	12.0	1512	0.4218	0.0026	0.2044
0.003	13.0	1638	0.4229	0.0026	0.1991
0.0017	14.0	1764	0.4195	0.0026	0.2032
0.0013	15.0	1890	0.4194	0.0026	0.1940
0.0012	16.0	2016	0.3966	0.0026	0.1935
0.0008	17.0	2142	0.4154	0.0026	0.1870
0.0006	18.0	2268	0.4141	0.0026	0.1841
0.0006	19.0	2394	0.4256	0.0026	0.1944
0.0004	20.0	2520	0.4270	0.0026	0.1863
0.0003	21.0	2646	0.4312	0.0026	0.1853
0.0003	22.0	2772	0.4264	0.0026	0.1859
0.0002	23.0	2898	0.4270	0.0026	0.1870
0.0002	24.0	3024	0.4282	0.0026	0.1864

dhaval0108/whisper-tiny-ce-dhc-combined-final

ap-qQomjOc6PHQk3VKKJyUtqd