whisper-medium_mucs_68256_15_100_1e-4_None_hinglish_mixed_scripts_FT
This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5060
- Global Wer: 36.7156
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: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 500
- num_epochs: 100
Training results
| Training Loss | Epoch | Step | Validation Loss | Global Wer |
|---|---|---|---|---|
| 0.8348 | 1.0 | 432 | 0.4487 | 43.1898 |
| 0.5112 | 2.0 | 864 | 0.4393 | 46.1451 |
| 0.3586 | 3.0 | 1296 | 0.4408 | 54.0143 |
| 0.2652 | 4.0 | 1728 | 0.4417 | 37.4872 |
| 0.2037 | 5.0 | 2160 | 0.4607 | 39.6111 |
| 0.1637 | 6.0 | 2592 | 0.5060 | 36.7156 |
| 0.1314 | 7.0 | 3024 | 0.5392 | 37.4559 |
| 0.1076 | 8.0 | 3456 | 0.5432 | 37.8943 |
| 0.0949 | 9.0 | 3888 | 0.5591 | 40.0951 |
Framework versions
- Transformers 5.0.0.dev0
- Pytorch 2.9.0+cu126
- Datasets 3.6.0
- Tokenizers 0.22.2