whisper-a-clp-ls-35

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

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: 35
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	40	0.1519	181.5514
No log	2.0	80	0.1651	50.5241
1.2204	3.0	120	0.3615	112.1593
1.2204	4.0	160	0.3251	62.8931
0.1484	5.0	200	0.2501	50.3145
0.1484	6.0	240	0.1648	61.2159
0.1484	7.0	280	0.1351	30.8176
0.0823	8.0	320	0.0712	30.1887
0.0823	9.0	360	0.1608	57.0231
0.0528	10.0	400	0.0652	21.5933
0.0528	11.0	440	0.0569	19.9161
0.0528	12.0	480	0.0706	22.6415
0.0436	13.0	520	0.0498	14.6751
0.0436	14.0	560	0.0493	22.4319
0.0374	15.0	600	0.0946	28.0922
0.0374	16.0	640	0.0930	29.5597
0.0374	17.0	680	0.0532	24.3187
0.0249	18.0	720	0.0639	22.2222
0.0249	19.0	760	0.0355	11.3208
0.0151	20.0	800	0.0608	14.2558
0.0151	21.0	840	0.0612	15.7233
0.0151	22.0	880	0.0632	13.8365
0.0124	23.0	920	0.0436	15.3040
0.0124	24.0	960	0.0415	14.4654
0.0084	25.0	1000	0.0474	17.6101
0.0084	26.0	1040	0.0284	10.2725
0.0084	27.0	1080	0.0291	7.9665
0.0058	28.0	1120	0.0275	7.9665
0.0058	29.0	1160	0.0302	7.9665
0.005	30.0	1200	0.0278	7.9665
0.005	31.0	1240	0.0277	8.1761
0.005	32.0	1280	0.0289	8.1761
0.0035	33.0	1320	0.0284	7.3375
0.0035	34.0	1360	0.0283	7.3375
0.0035	34.1266	1365	0.0283	7.3375