Whisper Small Eg - Tariq Hasaballah 100330
This model is a fine-tuned version of openai/whisper-small on the ASR-EGARBCSC: AN EGYPTIAN ARABIC CONVERSATIONAL SPEECH CORPUS dataset. It achieves the following results on the evaluation set:
- Loss: 0.5626
- Wer: 47.4960
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: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- training_steps: 500
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.7309 | 0.7267 | 125 | 0.5984 | 52.4512 |
| 0.3608 | 1.4535 | 250 | 0.5488 | 48.6031 |
| 0.1789 | 2.1802 | 375 | 0.5537 | 46.5999 |
| 0.1844 | 2.9070 | 500 | 0.5626 | 47.4960 |
Framework versions
- Transformers 4.40.0
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.19.1