[Abandoned] Get the good type to pass to a transcriber model

[slain]

Hello,
I'm quite new to Python, and trying to get a working transcriber with Streamlit and HuggingFace models.

I have an hardware constraint, my script must work on a 4GB Nvidia GPU...

Here is the code at this moment:

## Imports ##
import torch
import io
import streamlit as st
from pathlib import Path
from tempfile import NamedTemporaryFile
from transformers import AutoModelForCTC, Wav2Vec2ProcessorWithLM
import nemo.collections.asr as nemo_asr
import torchaudio

## Initialization ##
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model_name = "bofenghuang/stt_fr_fastconformer_hybrid_large"
asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name=model_name)

## Display ##
st.title("Transcribe audio to text")
col1, col2 = st.columns(2)
audio_source=st.sidebar.file_uploader(label="Choose file", type=["wav","m4a","mp3","wma"])

## Variables ##
suffix = ""
predicted_text = ""

## Processing ##
if audio_source is not None:
    suffix = Path(audio_source.name).suffix
    col1.write("Starting process")
    with NamedTemporaryFile(suffix=suffix) as temp_file:
        temp_file.write(audio_source.getvalue())
        temp_file.seek(0)
        predicted_text = asr_model.transcribe(temp_file)
    col2.write("Transcribed text :")
    col2.write(predicted_text)
    st.sidebar.download_button(label="Download text", data=predicted_text, file_name="transcript.txt",mime="text/plain")

Getting this by a temporary file gives me this error :

TypeError: object of type '_TemporaryFileWrapper' has no len()

I guess I should convert my audio content to another type, but I don't know to which type I should convert, how I should convert it, and how I'll get the contents with the new type.

Does anyone has an idea what I miss?

Thank you