File size: 795 Bytes
a6f4cc3 998bdae b954d78 998bdae a9e4715 92a9c4e 998bdae a9e4715 92a9c4e a9e4715 998bdae |
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 |
import streamlit as st
import torch
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
from diffusers.utils import export_to_video
# Explicitly set the device to CPU
device = torch.device("cpu")
# Load the model onto the CPU
pipe = DiffusionPipeline.from_pretrained(
"damo-vilab/text-to-video-ms-1.7b",
torch_dtype=torch.float32 # Use float32 for CPU
).to(device)
# Initialize the scheduler from the pipeline's configuration, no need to move it to the CPU
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
prompt = "Pop international experimental music"
# Generate the video frames on the CPU
video_frames = pipe(prompt, num_inference_steps=25).frames
# Export the frames to a video file
video_path = export_to_video(video_frames)
|