WhisperForSpokenNER-end2end
This model is a fine-tuned version of openai/whisper-small on the facebook/voxpopuli de+es+fr+nl dataset. It achieves the following results on the evaluation set:
- Loss: 0.3933
- Wer: 0.1464
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: 0.0001
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.3562 | 0.36 | 200 | 0.3265 | 0.1920 |
0.3149 | 0.71 | 400 | 0.3136 | 0.1842 |
0.2778 | 1.07 | 600 | 0.3204 | 0.1786 |
0.2288 | 1.43 | 800 | 0.3156 | 0.1717 |
0.2307 | 1.79 | 1000 | 0.3056 | 0.1708 |
0.1482 | 2.14 | 1200 | 0.3138 | 0.1682 |
0.1368 | 2.5 | 1400 | 0.3136 | 0.1656 |
0.1405 | 2.86 | 1600 | 0.3082 | 0.1617 |
0.0639 | 3.22 | 1800 | 0.3201 | 0.1612 |
0.0673 | 3.57 | 2000 | 0.3242 | 0.1612 |
0.0688 | 3.93 | 2200 | 0.3235 | 0.1584 |
0.0227 | 4.29 | 2400 | 0.3420 | 0.1558 |
0.0232 | 4.65 | 2600 | 0.3430 | 0.1525 |
0.0229 | 5.0 | 2800 | 0.3450 | 0.1528 |
0.0064 | 5.36 | 3000 | 0.3631 | 0.1498 |
0.0059 | 5.72 | 3200 | 0.3652 | 0.1482 |
0.0043 | 6.08 | 3400 | 0.3756 | 0.1482 |
0.0021 | 6.43 | 3600 | 0.3798 | 0.1477 |
0.002 | 6.79 | 3800 | 0.3824 | 0.1484 |
0.0014 | 7.15 | 4000 | 0.3876 | 0.1471 |
0.0013 | 7.51 | 4200 | 0.3900 | 0.1473 |
0.0013 | 7.86 | 4400 | 0.3917 | 0.1461 |
0.0012 | 8.22 | 4600 | 0.3929 | 0.1462 |
0.0012 | 8.58 | 4800 | 0.3932 | 0.1465 |
0.0012 | 8.94 | 5000 | 0.3933 | 0.1464 |
Framework versions
- Transformers 4.37.0.dev0
- Pytorch 2.1.0
- Datasets 2.14.6
- Tokenizers 0.14.1
- Downloads last month
- 3
Inference API (serverless) is not available, repository is disabled.
Model tree for qmeeus/whisper-small-multilingual-spoken-ner-end2end
Base model
openai/whisper-small
Finetuned
this model
Dataset used to train qmeeus/whisper-small-multilingual-spoken-ner-end2end
Evaluation results
- Wer on facebook/voxpopuli de+es+fr+nlself-reported0.146