metadata
language:
- ar
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
datasets:
- tarteel-ai/everyayah
metrics:
- wer
model-index:
- name: quran-whisper-ar-tiny-1
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: tarteel-ai/everyayah
type: tarteel-ai/everyayah
metrics:
- name: Wer
type: wer
value: 164.83516483516482
quran-whisper-ar-tiny-1
This model is a fine-tuned version of openai/whisper-tiny on the tarteel-ai/everyayah dataset. It achieves the following results on the evaluation set:
- Loss: 0.5858
- Wer: 164.8352
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: 0.0001
- 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
- training_steps: 45
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.2643 | 0.3333 | 5 | 4.2666 | 209.8901 |
| 3.5798 | 0.6667 | 10 | 1.9318 | 173.6264 |
| 1.2993 | 1.0 | 15 | 1.0968 | 102.1978 |
| 0.8612 | 0.6667 | 20 | 0.9798 | 145.0549 |
| 0.7235 | 0.8333 | 25 | 0.7984 | 131.8681 |
| 0.5694 | 1.0 | 30 | 0.7010 | 127.4725 |
| 0.5781 | 0.7778 | 35 | 0.7017 | 378.0220 |
| 0.5704 | 0.8889 | 40 | 0.6511 | 341.7582 |
| 0.4247 | 1.0 | 45 | 0.5858 | 164.8352 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.1.2
- Datasets 2.19.2
- Tokenizers 0.19.1