<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
subh_whisper_small_distil_att_loss_mozilla_epochs_50_batch_4_try2
This model is a fine-tuned version of rohitp1/kkkh_whisper_small_distillation_att_loss_mozilla_epochs_100_batch_4_concat_dataset on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.4047
- Wer: 26.9184
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: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 512
- total_train_batch_size: 2048
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine_with_restarts
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.1019 | 1.47 | 100 | 1.6488 | 24.8451 |
| 1.0977 | 2.94 | 200 | 1.6543 | 24.8816 |
| 1.0992 | 4.41 | 300 | 1.6592 | 24.8625 |
| 1.093 | 5.88 | 400 | 1.6705 | 24.8903 |
| 1.1001 | 7.35 | 500 | 1.6851 | 24.9043 |
| 1.0575 | 8.82 | 600 | 1.4047 | 26.9184 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1
- Datasets 2.8.0
- Tokenizers 0.13.2
