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
1.4716	1.0	96	0.6226	0.0038	0.2860
0.426	2.0	192	0.4039	0.0038	0.2032
0.2686	3.0	288	0.3802	0.0038	0.1906
0.1853	4.0	384	0.3837	0.0038	0.1951
0.1408	5.0	480	0.3908	0.0038	0.1987
0.1117	6.0	576	0.4120	0.0038	0.2120
0.0883	7.0	672	0.4322	0.0038	0.2033
0.0762	8.0	768	0.4580	0.0038	0.1822
0.0622	9.0	864	0.4785	0.0038	0.1845
0.0526	10.0	960	0.4701	0.0038	0.1722
0.0366	11.0	1056	0.4882	0.0038	0.1855
0.032	12.0	1152	0.5433	0.0038	0.1777
0.0268	13.0	1248	0.5267	0.0038	0.1801
0.0238	14.0	1344	0.5164	0.0038	0.1700
0.0165	15.0	1440	0.5553	0.0038	0.1905
0.0181	16.0	1536	0.5468	0.0038	0.1886
0.0143	17.0	1632	0.5834	0.0038	0.1824
0.0137	18.0	1728	0.5722	0.0038	0.1644
0.0069	19.0	1824	0.5759	0.0038	0.1816
0.0073	20.0	1920	0.5830	0.0038	0.1656
0.0032	21.0	2016	0.5990	0.0038	0.1679
0.0018	22.0	2112	0.6166	0.0038	0.1703
0.0012	23.0	2208	0.6260	0.0038	0.1712
0.0011	23.7510	2280	0.6257	0.0038	0.1723

dhaval0108/whisper-base-ce-dhc-combined

ap-HKK7gNmNqOuDPrXSI0yYra