<!-- 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. -->
dgx1_whisper_tiny_finetune_teacher_babble_noise_mozilla_100_epochs_batch_32_resume_training
This model is a fine-tuned version of rohitp1/dgx1_whisper_tiny_finetune_teacher_babble_noise_mozilla_40_epochs_batch_32 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.5241
- Wer: 41.3829
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: 32
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 32
- total_train_batch_size: 1024
- 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: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.0063 | 7.35 | 500 | 1.2717 | 41.0104 |
0.04 | 14.7 | 1000 | 1.2952 | 42.2359 |
0.0585 | 22.06 | 1500 | 1.2859 | 42.7616 |
0.0338 | 29.41 | 2000 | 1.3966 | 42.5144 |
0.0225 | 36.76 | 2500 | 1.4849 | 42.1141 |
0.014 | 44.12 | 3000 | 1.5241 | 41.3829 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1
- Datasets 2.8.0
- Tokenizers 0.13.2