<!-- 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. -->
kkkh_whisper_small_distillation_att_loss_libri360_epochs_50_batch_4_concat_dataset_try4
This model is a fine-tuned version of rohitp1/kkkh_whisper_small_distillation_att_loss_libri360_epochs_100_batch_4_concat_dataset_try2_ckpt200 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 10.8587
- Wer: 5.3390
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: 0.0005
- train_batch_size: 8
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 256
- 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 |
|---|---|---|---|---|
| 6.2271 | 0.98 | 100 | 5.0642 | 4.7958 |
| 6.0311 | 1.97 | 200 | 4.6785 | 4.9318 |
| 6.7957 | 2.95 | 300 | 6.8347 | 5.0522 |
| 7.9809 | 3.94 | 400 | 7.9823 | 5.2277 |
| 9.2538 | 4.92 | 500 | 7.9344 | 5.1974 |
| 19.6094 | 5.91 | 600 | 10.8587 | 5.3390 |
Framework versions
- Transformers 4.27.3
- Pytorch 1.12.1
- Datasets 2.6.1
- Tokenizers 0.13.1
