metadata
language:
- ig
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- google/fleurs
metrics:
- wer
model-index:
- name: Whisper Small Igbo
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: google/fleurs-jboat
type: google/fleurs
config: ig_ng
split: test
args: ig_ng
metrics:
- name: Wer
type: wer
value: 44.01272438082254
Whisper Small Igbo
This model is a fine-tuned version of openai/whisper-small on the google/fleurs-jboat dataset. It achieves the following results on the evaluation set:
- Loss: 1.0619
- Wer Ortho: 47.8937
- Wer: 44.0127
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: 50
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.3161 | 2.6455 | 500 | 0.7413 | 50.2340 | 46.1448 |
| 0.0421 | 5.2910 | 1000 | 0.8582 | 49.0269 | 44.8004 |
| 0.0168 | 7.9365 | 1500 | 0.9246 | 47.6351 | 43.5204 |
| 0.0075 | 10.5820 | 2000 | 0.9912 | 47.7541 | 43.3121 |
| 0.0051 | 13.2275 | 2500 | 1.0277 | 47.7377 | 43.3954 |
| 0.0067 | 15.8730 | 3000 | 1.0354 | 47.6638 | 43.1644 |
| 0.0041 | 18.5185 | 3500 | 1.0722 | 48.3864 | 44.1112 |
| 0.0028 | 21.1640 | 4000 | 1.0619 | 47.8937 | 44.0127 |
Framework versions
- Transformers 4.42.4
- Pytorch 2.3.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1