README.md

metadata

language:
  - en
  - ar
tags:
  - automatic-speech-recognition
  - whisper
  - medical
  - asr
  - fp16
license: apache-2.0
datasets:
  - yashtiwari/PaulMooney-Medical-ASR-Data
model-index:
  - name: Whisper Large-v3 Medical
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: Medical ASR
          type: yashtiwari/PaulMooney-Medical-ASR-Data
        metrics:
          - type: wer
            value: 4.12
metrics:
  - wer
base_model:
  - openai/whisper-large-v3
pipeline_tag: automatic-speech-recognition

Whisper Large-v3 Medical

This model is a fine-tuned version of openai/whisper-large-v3 on medical speech data. It is trained for Automatic Speech Recognition (ASR) of doctor-patient dialogues and medical narratives, with support for English and Arabic.

🩺 Use Cases

• Transcribing clinical interviews.
• Building medical dictation tools.
• Precision: float16 (better for inference), float32 (better for fine-tuning)

📊 Performance

• **WER (Word Error Rate): 4.12% **
• Optimized for clean and domain-specific spoken medical data.

🔧 Model Details

• Base model: openai/whisper-large-v3
• Fine-tuned on: yashtiwari/PaulMooney-Medical-ASR-Data
• Languages: Multilingual
• Framework: 🤗 Transformers

🧪 How to Use

from transformers import WhisperProcessor, WhisperForConditionalGeneration

model_id = "yehiazak/whisper-largev3-medical"

# Load FP16 model
model = WhisperForConditionalGeneration.from_pretrained(model_id, revision="fp16")

# Load FP32 model
model = WhisperForConditionalGeneration.from_pretrained(model_id)

processor = WhisperProcessor.from_pretrained(model_id)