metadata
library_name: transformers
language:
- ru
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- custom
metrics:
- wer
model-index:
- name: Whisper Small RU (Phone) v6 - Stepler
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: ru_audiobooks_sova
type: custom
args: 'config: ru, split: test'
metrics:
- name: Wer
type: wer
value: 14.931947459319
Whisper Small RU (Phone) v6 - Stepler
This model is a fine-tuned version of openai/whisper-small on the ru_audiobooks_sova dataset. It achieves the following results on the evaluation set:
- Loss: 0.1732
- Wer: 14.9319
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: 1000
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2325 | 0.3501 | 4000 | 0.2328 | 20.7527 |
| 0.1867 | 0.7002 | 8000 | 0.2006 | 18.3470 |
| 0.0968 | 1.0503 | 12000 | 0.1889 | 16.9944 |
| 0.1109 | 1.4004 | 16000 | 0.1805 | 16.6653 |
| 0.0947 | 1.7505 | 20000 | 0.1746 | 15.7622 |
| 0.0435 | 2.1007 | 24000 | 0.1756 | 15.4416 |
| 0.0566 | 2.4508 | 28000 | 0.1756 | 15.1253 |
| 0.0462 | 2.8009 | 32000 | 0.1732 | 14.9319 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.1+cu124
- Datasets 2.21.0
- Tokenizers 0.19.1