Meatgrinder/transcribe.py

import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
def process(audio_path, model_name):
    model = AutoModelForSpeechSeq2Seq.from_pretrained(
        model_name,
        torch_dtype=torch.float32,
        low_cpu_mem_usage=True,
        use_safetensors=True
    )
    model.to("mps")
    processor = AutoProcessor.from_pretrained(model_name)
    pipe = pipeline(
        "automatic-speech-recognition",
        model=model,
        tokenizer=processor.tokenizer,
        feature_extractor=processor.feature_extractor,
        torch_dtype=torch.float32,
        return_timestamps=True,
        device="mps"
    )
    output = pipe(audio_path)["text"]
    return output