kuan2/whisper-tiny-en-book

Whisper tiny Book

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

• Loss: 0.0251
• Wer: 1.9268
• Cer: 0.3553

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
3.9929	0.0709	10	3.3716	29.5761	10.8629
3.9386	0.1418	20	3.0177	28.4200	10.2030
3.375	0.2128	30	2.6123	27.7457	9.8477
3.0007	0.2837	40	2.2620	26.1079	9.0186
2.5946	0.3546	50	1.9080	23.6031	7.8849
2.1729	0.4255	60	1.5682	20.2312	6.7682
1.8715	0.4965	70	1.3811	18.5934	6.0914
1.6468	0.5674	80	1.2722	17.2447	5.7868
1.5396	0.6383	90	1.1741	15.7033	5.0254
1.4669	0.7092	100	1.0837	14.5472	4.6024
1.3851	0.7801	110	0.9828	11.4644	3.7394
1.2107	0.8511	120	0.9004	9.9229	3.0457
1.1375	0.9220	130	0.8206	9.7303	3.2318
0.9685	0.9929	140	0.7487	8.9595	2.7750
0.8194	1.0638	150	0.6772	8.2852	2.5719
0.7976	1.1348	160	0.6092	7.8998	2.5550
0.7104	1.2057	170	0.5373	7.5145	2.2335
0.6255	1.2766	180	0.4742	7.0328	2.0981
0.5621	1.3475	190	0.4145	6.5511	1.9966
0.5151	1.4184	200	0.3615	5.3950	1.5905
0.4697	1.4894	210	0.3107	5.3950	1.4890
0.4032	1.5603	220	0.2671	5.2023	1.4382
0.3962	1.6312	230	0.2254	4.6243	1.2183
0.3278	1.7021	240	0.1901	4.1426	1.1337
0.2661	1.7730	250	0.1614	3.9499	1.0660
0.2927	1.8440	260	0.1362	3.6609	0.9475
0.2197	1.9149	270	0.1165	3.5645	0.9306
0.1859	1.9858	280	0.0991	3.4682	0.9137
0.1446	2.0567	290	0.0885	3.6609	0.8799
0.13	2.1277	300	0.0770	3.8536	0.9475
0.1376	2.1986	310	0.0675	3.5645	0.8629
0.1107	2.2695	320	0.0610	3.2755	0.8291
0.1187	2.3404	330	0.0568	2.9865	0.7614
0.1173	2.4113	340	0.0521	2.8902	0.7445
0.0859	2.4823	350	0.0476	2.8902	0.7445
0.0933	2.5532	360	0.0440	2.4085	0.5922
0.0787	2.6241	370	0.0397	2.0231	0.4399
0.0901	2.6950	380	0.0363	1.7341	0.3215
0.0833	2.7660	390	0.0337	1.9268	0.4230
0.0751	2.8369	400	0.0325	2.0231	0.3723
0.071	2.9078	410	0.0311	2.0231	0.3723
0.068	2.9787	420	0.0298	1.9268	0.3553
0.0724	3.0496	430	0.0286	2.1195	0.4569
0.0484	3.1206	440	0.0277	2.1195	0.4569
0.0422	3.1915	450	0.0272	1.9268	0.3553
0.0536	3.2624	460	0.0266	2.0231	0.4399
0.0443	3.3333	470	0.0262	2.0231	0.4399
0.0465	3.4043	480	0.0257	1.9268	0.3553
0.0555	3.4752	490	0.0253	1.9268	0.3553
0.0467	3.5461	500	0.0251	1.9268	0.3553

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1