app.py

from transformers import AutoProcessor, MusicgenForConditionalGeneration
from IPython.display import Audio
import scipy
import torch
import streamlit as st


def mu_gen(prompt):
  processor = AutoProcessor.from_pretrained("facebook/musicgen-small")
  model = MusicgenForConditionalGeneration.from_pretrained("facebook/musicgen-small")

  device = torch.device("cpu")
  model.to(device)

  inputs = processor(
      text = [str(prompt)],     # This line is correct
      padding=True,
      return_tensors="pt",
  )
  
  inputs = {key: value.to(device) for key, value in inputs.items()}

    # Generate audio on CPU
  audio_values = model.generate(**inputs, max_new_tokens=256)
  sampling_rate = model.config.audio_encoder.sampling_rate

    # Create an Audio object from the generated audio
  result = Audio(audio_values[0].numpy(), rate=sampling_rate)

  return result


def main():
    st.title("Text to music")

    # Input text prompt
    title = st.text_input('Write a prompt (จะใช้เวลาค่อนข้างมากในการสร้างเนื่องจากใช้ CPU ในการรันโมเดล)', "")

    if st.button('Generate Image'):
        # Call the pic_mo function to generate an image
        generated_music = mu_gen(prompt)

        # Display the generated image
        st.image(generated_music, caption='Generated Music', use_column_width=True)

if __name__ == '__main__':
    main()