metadata
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- google/fleurs
metrics:
- wer
model-index:
- name: whisper-small-bn-in
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: google/fleurs
type: google/fleurs
config: bn_in
split: train+validation
args: bn_in
metrics:
- name: Wer
type: wer
value: 0.45676500508647
whisper-small-bn-in
This model is a fine-tuned version of openai/whisper-small on the google/fleurs dataset. It achieves the following results on the evaluation set:
- Loss: 0.1842
- Wer: 0.4568
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: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 5
- training_steps: 2000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.4443 | 0.53 | 100 | 0.3399 | 0.7272 |
| 0.249 | 1.07 | 200 | 0.2222 | 0.6244 |
| 0.1662 | 1.6 | 300 | 0.1778 | 0.5807 |
| 0.1221 | 2.14 | 400 | 0.1602 | 0.5397 |
| 0.0965 | 2.67 | 500 | 0.1484 | 0.5168 |
| 0.0646 | 3.21 | 600 | 0.1475 | 0.4966 |
| 0.0566 | 3.74 | 700 | 0.1420 | 0.4812 |
| 0.028 | 4.28 | 800 | 0.1511 | 0.4910 |
| 0.0325 | 4.81 | 900 | 0.1476 | 0.4766 |
| 0.0177 | 5.35 | 1000 | 0.1593 | 0.4876 |
| 0.0176 | 5.88 | 1100 | 0.1589 | 0.4715 |
| 0.0127 | 6.42 | 1200 | 0.1622 | 0.4634 |
| 0.0126 | 6.95 | 1300 | 0.1706 | 0.4673 |
| 0.0089 | 7.49 | 1400 | 0.1777 | 0.4712 |
| 0.0087 | 8.02 | 1500 | 0.1776 | 0.4666 |
| 0.0076 | 8.56 | 1600 | 0.1788 | 0.4505 |
| 0.007 | 9.09 | 1700 | 0.1906 | 0.4685 |
| 0.0057 | 9.63 | 1800 | 0.1840 | 0.4573 |
| 0.0064 | 10.16 | 1900 | 0.1841 | 0.4569 |
| 0.0057 | 10.7 | 2000 | 0.1842 | 0.4568 |
Framework versions
- Transformers 4.32.0.dev0
- Pytorch 1.12.1+cu116
- Datasets 2.4.0
- Tokenizers 0.12.1