metadata
language:
- bn
license: apache-2.0
tags:
- whisper-event
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_11_0
metrics:
- wer
model-index:
- name: Whisper Tiny Bengali
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: mozilla-foundation/common_voice_11_0 bn
type: mozilla-foundation/common_voice_11_0
config: bn
split: test
args: bn
metrics:
- name: Wer
type: wer
value: 32.89771261927907
Whisper Tiny Bengali
This model is a fine-tuned version of openai/whisper-tiny on the mozilla-foundation/common_voice_11_0 bn dataset. It achieves the following results on the evaluation set:
- Loss: 0.2314
- Wer: 32.8977
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: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.3362 | 0.96 | 1000 | 0.3536 | 45.0860 |
| 0.2395 | 1.91 | 2000 | 0.2745 | 37.1714 |
| 0.205 | 2.87 | 3000 | 0.2485 | 34.7353 |
| 0.1795 | 3.83 | 4000 | 0.2352 | 33.2469 |
| 0.1578 | 4.78 | 5000 | 0.2314 | 32.8977 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.8.1.dev0
- Tokenizers 0.13.2