README.md

metadata

language:
  - en
license: apache-2.0
base_model: openai/whisper-tiny.en
tags:
  - generated_from_trainer
datasets:
  - mispeech/speechocean762
metrics:
  - wer
model-index:
  - name: Whisper Tiny En - speechocean762
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: speechocean762
          type: mispeech/speechocean762
        metrics:
          - name: Wer
            type: wer
            value: 38.84869455803712

Whisper Tiny En - speechocean762

This model is a fine-tuned version of openai/whisper-tiny.en on the speechocean762 dataset. It achieves the following results on the evaluation set (Best results):

• Loss: 0.6837
• Wer: 28.3422

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: 128
• 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: 250
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.0683	0.2778	10	4.0618	38.3139
3.9216	0.5556	20	3.8951	37.4961
3.7275	0.8333	30	3.6337	45.8635
3.3621	1.1111	40	3.2960	36.4580
2.9818	1.3889	50	2.8749	40.7046
2.5404	1.6667	60	2.3590	44.1648
1.9537	1.9444	70	1.7972	49.4495
1.4184	2.2222	80	1.3603	66.3731
1.1875	2.5	90	1.1660	55.5206
1.1203	2.7778	100	1.0743	44.8569
1.024	3.0556	110	1.0085	44.2277
0.905	3.3333	120	0.9581	42.4347
0.8787	3.6111	130	0.9169	40.9877
0.8677	3.8889	140	0.8844	37.2130
0.7563	4.1667	150	0.8573	36.4895
0.7497	4.4444	160	0.8324	35.9862
0.7283	4.7222	170	0.8097	35.2941
0.7055	5.0	180	0.7907	30.6071
0.6259	5.2778	190	0.7770	30.9531
0.6115	5.5556	200	0.7601	30.3555
0.5998	5.8333	210	0.7457	29.8207
0.5752	6.1111	220	0.7368	29.9465
0.5031	6.3889	230	0.7284	29.7892
0.5079	6.6667	240	0.7140	29.0028
0.4969	6.9444	250	0.7006	29.3174
0.4285	7.2222	260	0.6951	32.5889
0.466	7.5	270	0.6886	31.6766
0.4101	7.7778	280	0.6837	28.3422
0.4021	8.0556	290	0.6755	31.4250
0.359	8.3333	300	0.6763	32.5260
0.3281	8.6111	310	0.6727	32.2114
0.3604	8.8889	320	0.6695	36.1120
0.3085	9.1667	330	0.6698	32.1799
0.3007	9.4444	340	0.6698	32.3372
0.3313	9.7222	350	0.6659	35.7974
0.2862	10.0	360	0.6638	32.0226
0.278	10.2778	370	0.6639	31.9912
0.2645	10.5556	380	0.6639	32.0856
0.2708	10.8333	390	0.6649	32.0541
0.257	11.1111	400	0.6620	32.1799
0.2455	11.3889	410	0.6621	31.8025
0.2506	11.6667	420	0.6636	38.9745
0.2545	11.9444	430	0.6635	38.9116
0.2266	12.2222	440	0.6644	31.8339
0.2072	12.5	450	0.6652	32.1799
0.2382	12.7778	460	0.6661	31.9597
0.219	13.0556	470	0.6653	38.7858
0.2256	13.3333	480	0.6649	38.9431
0.2178	13.6111	490	0.6652	38.9431
0.2229	13.8889	500	0.6654	38.8487

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.2.dev0
• Tokenizers 0.19.1