| import torch | |
| import librosa | |
| from transformers import WhisperProcessor, pipeline | |
| def predict_from_file(audio_path, model_name="ahmad1703/whis_ee"): | |
| Make a prediction using the WhisperClassifier model | |
| Args: | |
| audio_path: Path to the audio file | |
| model_name: The Hugging Face model name | |
| Returns: | |
| Prediction (0 or 1) | |
| # Load audio | |
| audio, sr = librosa.load(audio_path, sr=16000) | |
| # Create pipeline | |
| classifier = pipeline("audio-classification", model=model_name) | |
| # Get prediction | |
| result = classifier(audio) | |
| # Convert probability to binary class | |
| prediction = 1 if result["score"] > 0.5 else 0 | |
| return prediction | |
| if __name__ == "__main__": | |
| # Example usage | |
| audio_path = "example.wav" # Replace with your audio file | |
| prediction = predict_from_file(audio_path) | |
| print(f"Prediction: {prediction}") | |