metadata
language:
- mn
license: apache-2.0
tags:
- generated_from_trainer
base_model: openai/whisper-small
datasets:
- audiofolder
metrics:
- wer
model-index:
- name: Whisper Small MN with custom data + Common voice + Fluers combined - Zagi
results:
- task:
type: automatic-speech-recognition
name: Automatic Speech Recognition
dataset:
name: audiofolder
type: audiofolder
config: default
split: None
args: 'config: mn, split: test'
metrics:
- type: wer
value: 23.64547043436441
name: Wer
Whisper Small MN with custom data + Common voice + Fluers combined - Zagi
This model is a fine-tuned version of openai/whisper-small on the audiofolder dataset. It achieves the following results on the evaluation set:
- Loss: 0.2272
- Wer: 23.6455
- Cer: 7.3582
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.4602 | 0.39 | 500 | 0.4537 | 44.8910 | 13.9069 |
| 0.3075 | 0.79 | 1000 | 0.3099 | 33.6078 | 10.3059 |
| 0.1829 | 1.18 | 1500 | 0.2715 | 28.9205 | 8.8684 |
| 0.1838 | 1.58 | 2000 | 0.2486 | 26.7223 | 8.2105 |
| 0.1619 | 1.97 | 2500 | 0.2328 | 25.3029 | 7.8554 |
| 0.1038 | 2.37 | 3000 | 0.2339 | 24.4936 | 7.6527 |
| 0.0915 | 2.76 | 3500 | 0.2287 | 23.7944 | 7.3275 |
| 0.0815 | 3.16 | 4000 | 0.2272 | 23.6455 | 7.3582 |
Framework versions
- Transformers 4.39.1
- Pytorch 2.0.1+cu117
- Datasets 2.18.0
- Tokenizers 0.15.2