metadata
language:
- hi
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- shields/whisper-small-hindi
metrics:
- wer
model-index:
- name: Whisper Small Hi
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: shields/whisper-small-hindi
type: shields/whisper-small-hindi
args: 'config: hi, split: test'
metrics:
- name: Wer
type: wer
value: 68.6046511627907
Whisper Small Hi
This model is a fine-tuned version of openai/whisper-small on the shields/whisper-small-hindi dataset. It achieves the following results on the evaluation set:
- Loss: 0.9716
- Wer: 68.6047
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-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.039 | 4.0 | 200 | 0.7402 | 70.9302 |
| 0.002 | 8.0 | 400 | 0.8662 | 62.7907 |
| 0.0003 | 12.0 | 600 | 0.9329 | 66.2791 |
| 0.0002 | 16.0 | 800 | 0.9613 | 68.6047 |
| 0.0002 | 20.0 | 1000 | 0.9716 | 68.6047 |
Framework versions
- Transformers 4.39.2
- Pytorch 2.2.0+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2