anzorq/openai-whisper-large-v2-LORA-colab

import torch
import gradio as gr
from transformers import (
    AutomaticSpeechRecognitionPipeline,
    WhisperForConditionalGeneration,
    WhisperTokenizer,
    WhisperProcessor,
)
from peft import PeftModel, PeftConfig
from pytube import YouTube

peft_model_id = "anzorq/openai-whisper-large-v2-LORA-colab"
# peft_model_id = "/content/whisper_large_kbd_lora/checkpoint-64"
language = "Georgian"
task = "transcribe"
peft_config = PeftConfig.from_pretrained(peft_model_id)
model = WhisperForConditionalGeneration.from_pretrained(
    peft_config.base_model_name_or_path, load_in_8bit=True, device_map="auto"
)

model = PeftModel.from_pretrained(model, peft_model_id)
tokenizer = WhisperTokenizer.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
processor = WhisperProcessor.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
feature_extractor = processor.feature_extractor
forced_decoder_ids = processor.get_decoder_prompt_ids(language=language, task=task)
pipe = AutomaticSpeechRecognitionPipeline(model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)

def transcribe(path_to_audio):
    with torch.cuda.amp.autocast():
        text = pipe(audio_path, generate_kwargs={"forced_decoder_ids": forced_decoder_ids}, max_new_tokens=255)["text"]
    return text

transcribe(path_to_audio)