whisper-medium-pt-cv16-fleurs-coraa
This model is a fine-tuned version of openai/whisper-medium on the fsicoli/cv16-fleurs-coraa-mls-ted-alcaim-cf-cdc-lapsbm-lapsmail default dataset. It achieves the following results on the evaluation set:
- Loss: 0.4546
- Wer: 0.1652
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-06
- train_batch_size: 8
- 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: 6000
- training_steps: 20000
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.748 | 0.02 | 2000 | 0.3752 | 0.2156 |
0.6051 | 0.05 | 4000 | 0.4198 | 0.1885 |
0.5578 | 0.07 | 6000 | 0.4245 | 0.1757 |
0.4632 | 0.1 | 8000 | 0.4393 | 0.1736 |
0.5151 | 0.12 | 10000 | 0.4457 | 0.1701 |
0.5158 | 0.15 | 12000 | 0.4314 | 0.1671 |
0.55 | 0.17 | 14000 | 0.4566 | 0.1675 |
0.4524 | 0.19 | 16000 | 0.4648 | 0.1663 |
0.4667 | 0.22 | 18000 | 0.4606 | 0.1656 |
0.5133 | 0.24 | 20000 | 0.4546 | 0.1652 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.2.1+cu121
- Datasets 2.18.1.dev0
- Tokenizers 0.15.0
- Downloads last month
- 2
Finetuned from
Evaluation results
- Wer on fsicoli/cv16-fleurs-coraa-mls-ted-alcaim-cf-cdc-lapsbm-lapsmail defaultself-reported0.165