Hugging Face's logo

ymoslem
/
whisper-small-ar-v2

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
Arabic
whisper
audio
Eval Results
Inference Endpoints
Model card Files Metrics Community
whisper-small-ar-v2 / README.md
ymoslem's picture
ymoslem
Duplicate from ymoslem/whisper-small-ar-v1
3fbfbbb verified
preview code
|
raw
history blame
No virus
2.34 kB
metadata
license: apache-2.0
base_model: openai/whisper-small
tags:
  - audio
  - automatic-speech-recognition
datasets:
  - mozilla-foundation/common_voice_16_1
metrics:
  - wer
widget:
  - example_title: Sample 1
    src: sample_ar.mp3
model-index:
  - name: whisper-small-ar-v1
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: mozilla-foundation/common_voice_16_1
          type: mozilla-foundation/common_voice_16_1
          config: ar
          split: test
          args: ar
        metrics:
          - name: Wer
            type: wer
            value: 158.15321276282899
language:
  - ar
library_name: transformers
pipeline_tag: automatic-speech-recognition

whisper-small-ar-v1

This model is for Arabic automatic speech recognition (ASR). It is a fine-tuned version of openai/whisper-small on the Arabic portion of the mozilla-foundation/common_voice_16_1 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3354
  • Wer: 158.1532

Model description

Whisper model fine-tuned on Arabic data, following the official tutorial.

Intended uses & limitations

The model is not fully trained yet. Hence, it is not intended for professional use.

Training and evaluation data

Training Data: CommonVoice (v16.1) Arabic train + validation splits
Validation Data: CommonVoice (v16.1) Arabic test split

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 1e-05
  • train_batch_size: 32
  • 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: 500
  • training_steps: 4000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.2742 0.82 1000 0.3790 275.2463
0.1625 1.65 2000 0.3353 228.5252
0.1002 2.47 3000 0.3311 238.8858
0.0751 3.3 4000 0.3354 158.1532

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.2.0+cu121
  • Datasets 2.17.0
  • Tokenizers 0.15.2