kuan2/whisper-small-en-book

Whisper Small Book

This model is a fine-tuned version of openai/whisper-small.en on the Book dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.0043
• Wer: 0.8502
• Cer: 0.1860

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.6055	0.0709	10	3.2795	20.7129	5.9391
3.5036	0.1418	20	2.9343	19.6532	5.7191
2.9344	0.2128	30	2.4058	17.2447	5.2623
2.3634	0.2837	40	1.7221	15.2216	4.7547
1.6372	0.3546	50	1.1584	12.9094	4.0948
1.2803	0.4255	60	1.0066	11.7534	3.5702
1.1453	0.4965	70	0.9125	9.1522	2.7580
1.0048	0.5674	80	0.8253	7.1291	2.1320
0.9227	0.6383	90	0.7304	5.7803	1.7428
0.8517	0.7092	100	0.6404	4.6243	1.4382
0.7304	0.7801	110	0.5414	3.1792	0.9306
0.6129	0.8511	120	0.4563	2.8902	0.7445
0.5272	0.9220	130	0.3733	2.1195	0.5922
0.4034	0.9929	140	0.3006	1.7341	0.4399
0.2945	1.0638	150	0.2279	1.1561	0.3384
0.2304	1.1348	160	0.1621	1.1561	0.3384
0.1543	1.2057	170	0.1051	1.2524	0.3215
0.1112	1.2766	180	0.0686	0.9634	0.2707
0.081	1.3475	190	0.0439	0.8671	0.2369
0.0593	1.4184	200	0.0333	0.7707	0.1523
0.0555	1.4894	210	0.0256	0.9634	0.1861
0.0438	1.5603	220	0.0203	0.7707	0.1523
0.0444	1.6312	230	0.0150	0.6744	0.1354
0.0435	1.7021	240	0.0111	0.6744	0.1354
0.0319	1.7730	250	0.0081	0.5780	0.1184
0.0256	1.8440	260	0.0065	0.4817	0.1015
0.0288	1.9149	270	0.0063	0.6744	0.1354
0.0161	1.9858	280	0.0055	0.6744	0.1354
0.0073	2.0567	290	0.0076	0.6744	0.1354
0.0064	2.1277	300	0.0035	0.4817	0.1015
0.0074	2.1986	310	0.0033	0.4817	0.1015
0.0064	2.2695	320	0.0029	0.4817	0.1015
0.0054	2.3404	330	0.0033	0.5780	0.1184
0.0107	2.4113	340	0.0026	0.4817	0.1015
0.0036	2.4823	350	0.0025	0.4817	0.1015
0.008	2.5532	360	0.0023	0.4817	0.1015
0.0043	2.6241	370	0.0021	0.4817	0.1015
0.004	2.6950	380	0.0020	0.4817	0.1015
0.0047	2.7660	390	0.0019	0.4817	0.1015
0.0034	2.8369	400	0.0019	0.4817	0.1015
0.0034	2.9078	410	0.0018	0.4817	0.1015
0.0042	2.9787	420	0.0018	0.4817	0.1015
0.0028	3.0496	430	0.0017	0.4817	0.1015
0.0021	3.1206	440	0.0017	0.4817	0.1015
0.0018	3.1915	450	0.0016	0.4817	0.1015
0.0031	3.2624	460	0.0016	0.4817	0.1015
0.0019	3.3333	470	0.0016	0.4817	0.1015
0.0019	3.4043	480	0.0016	0.4817	0.1015
0.002	3.4752	490	0.0016	0.4817	0.1015
0.002	3.5461	500	0.0016	0.4817	0.1015

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1