asr.py

import torchaudio
import torch
from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq

# Load processor and model
processor = AutoProcessor.from_pretrained("ixxan/whisper-small-ug-cv-15")
model = AutoModelForSpeechSeq2Seq.from_pretrained("ixxan/whisper-small-ug-cv-15")

def transcribe(audio_path: str) -> str:
    """
    Transcribes audio to text using the Whisper model for Uyghur.

    Args:
    - audio_path (str): Path to the audio file to transcribe.

    Returns:
    - str: The transcription of the audio.
    """

    # Load audio file
    audio_input, sampling_rate = torchaudio.load(audio_path)

    # Resample if needed
    if sampling_rate != processor.feature_extractor.sampling_rate:
        resampler = torchaudio.transforms.Resample(sampling_rate, processor.feature_extractor.sampling_rate)
        audio_input = resampler(audio_input)

    # Preprocess the audio input
    inputs = processor(audio_input.squeeze(), sampling_rate=16000, return_tensors="pt")

    # Move model to GPU if available
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model.to(device)
    inputs = {key: val.to(device) for key, val in inputs.items()}

    # Generate transcription
    with torch.no_grad():
        generated_ids = model.generate(inputs["input_features"], max_length=225)

    # Decode the output to get the transcription text
    transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
    
    return transcription