w2v2-base_kabir
This model is a fine-tuned version of facebook/wav2vec2-base on the Medical Speech, Transcription, and Intent dataset. It achieves the following results on the evaluation set:
- Loss: 0.9705
- Wer: 0.3289
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 250
- training_steps: 2500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.8771 | 20.8333 | 500 | 2.8897 | 1.0 |
| 0.3115 | 41.6667 | 1000 | 0.9687 | 0.4125 |
| 0.1248 | 62.5 | 1500 | 0.9421 | 0.3502 |
| 0.0658 | 83.3333 | 2000 | 0.9894 | 0.3348 |
| 0.0703 | 104.1667 | 2500 | 0.9705 | 0.3289 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.0
- Datasets 3.0.1
- Tokenizers 0.20.0