ST/inference/inference1.py

import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "6"  # SET the GPUs you want to use
import csv
from transformers import WhisperProcessor, WhisperForConditionalGeneration

# device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

model_id = "openai/whisper-large-v3"

model = WhisperForConditionalGeneration.from_pretrained(model_id)
processor = WhisperProcessor.from_pretrained(model_id)

mp3_folder = "./extra_epi_wav/"

# Get a list of all the mp3 files in the folder
mp3_files = [file for file in os.listdir(mp3_folder) if file.endswith(".wav")]

# Create a CSV file to store the transcripts
csv_filename = "transcripts_with_lang.csv"

with open(csv_filename, mode='w', newline='', encoding='utf-8') as csv_file:
    fieldnames = ['File Name', 'Transcript', 'lang']
    writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
    
    # Write the header to the CSV file
    writer.writeheader()

    # Process each mp3 file and write the results to the CSV file
    processed_files_counter = 0
    for mp3_file in mp3_files:
        mp3_path = os.path.join(mp3_folder, mp3_file)
        result = pipe(mp3_path)
        print(result)

        transcript = result["text"]
        # p

        processed_files_counter += 1

        # Check progress after every 10 files
        if processed_files_counter % 10 == 0:
            print(f"{processed_files_counter} files processed.")
        
        # Write the file name and transcript to the CSV file
        writer.writerow({'File Name': mp3_file, 'Transcript': transcript})

print(f"Transcripts saved to {csv_filename}")