File size: 1,283 Bytes
64e127f 7b8d71f 9847153 7b8d71f 64e127f 9847153 64e127f 4f5f255 08bd4c4 7b8d71f 4f5f255 64e127f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 |
import random
import subprocess
import torch
import numpy as np
import gradio as gr
from transformers import AutoProcessor, MusicgenForConditionalGeneration
COLORS = [
["#ff0000", "#00ff00"],
["#00ff00", "#0000ff"],
["#0000ff", "#ff0000"],
]
path = "facebook/musicgen-large"
processor = AutoProcessor.from_pretrained(path)
model = MusicgenForConditionalGeneration.from_pretrained(path, torch_dtype=torch.float16).to("cuda")
def predict(text):
inputs = processor(
text=[text],
padding=True,
return_tensors="pt",).to("cuda")
with torch.autocast("cuda"):
outputs = model.generate(**inputs, do_sample=True, guidance_scale=3, max_new_tokens=512)
video_path = gr.make_waveform((32000, outputs[0].cpu().numpy().astype(np.float16).ravel()), bars_color=random.choice(COLORS), bar_count=75)
command = f'ffmpeg -y -i {video_path} -vf "scale=250:150" result.mp4'
subprocess.run(command, shell=True)
return (32000, outputs[0][0].cpu().numpy().astype(np.float16)), "result.mp4"
title = "MusicGen"
gr.Interface(
fn=predict,
inputs=[
gr.Textbox(label="Text prompt"),
],
outputs=["audio", "video"],
title=title,
theme="gradio/monochrome",
).queue(max_size=10).launch()
|