metadata
license: apache-2.0
base_model: openai/whisper-medium.en
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: whisper-medium.en-cit-do015-wd0-lr1e-06-SF-500
results: []
whisper-medium.en-cit-do015-wd0-lr1e-06-SF-500
This model is a fine-tuned version of openai/whisper-medium.en on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.7070
- Wer Ortho: 28.9359
- Wer: 18.7657
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: 1e-06
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 200
- training_steps: 500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 1.6528 | 3.1746 | 100 | 1.1367 | 40.6706 | 29.8170 |
| 0.8589 | 6.3492 | 200 | 0.7969 | 30.5029 | 20.0215 |
| 0.6147 | 9.5238 | 300 | 0.7363 | 28.9359 | 18.7298 |
| 0.5156 | 12.6984 | 400 | 0.7134 | 28.7536 | 18.8375 |
| 0.4706 | 15.8730 | 500 | 0.7070 | 28.9359 | 18.7657 |
Framework versions
- Transformers 4.44.0
- Pytorch 1.13.1+cu117
- Datasets 2.20.0
- Tokenizers 0.19.1