metadata
language:
- nyn
license: apache-2.0
base_model: openai/whisper-small
tags:
- whisper-event
- generated_from_trainer
datasets:
- tericlabs
metrics:
- wer
model-index:
- name: Whisper Small Runyankore
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Yogera data
type: tericlabs
metrics:
- name: Wer
type: wer
value: 47.48073503260225
Whisper Small Runyankore
This model is a fine-tuned version of openai/whisper-small on the Yogera data dataset. It achieves the following results on the evaluation set:
- Loss: 0.6975
- Wer: 47.4807
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
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.7277 | 0.43 | 100 | 2.6880 | 98.2810 |
| 2.265 | 0.85 | 200 | 1.8155 | 84.8251 |
| 1.6872 | 1.28 | 300 | 1.3734 | 75.1630 |
| 1.25 | 1.7 | 400 | 0.9704 | 65.2638 |
| 0.9047 | 2.13 | 500 | 0.8496 | 57.9727 |
| 0.7852 | 2.55 | 600 | 0.7705 | 57.0836 |
| 0.7065 | 2.98 | 700 | 0.7080 | 51.4523 |
| 0.4952 | 3.4 | 800 | 0.6939 | 50.3853 |
| 0.516 | 3.83 | 900 | 0.6786 | 49.0219 |
| 0.4005 | 4.26 | 1000 | 0.6975 | 47.4807 |
Framework versions
- Transformers 4.38.0.dev0
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0