I2V 为什么很慢 10分钟生成一个视频

[THEBEST-cloud]

System Info / 系統信息

cuda_version: cuda_11.8.r11.8/compiler.31833905_0 diffusers_version: 0.30.3 python_version: 3.10.0 operating_system: Linux-5.15.0-92-generic-x86_64-with-glibc2.35

Information / 问题信息

• [X] The official example scripts / 官方的示例脚本
• [ ] My own modified scripts / 我自己修改的脚本和任务

Reproduction / 复现过程

用于图像生成视频

import torch from diffusers import CogVideoXImageToVideoPipeline from diffusers.utils import export_to_video, load_image from time import time import cv2 import imageio

start = time() prompt = "A woman picks up a pair of sunglasses and smiles." image = load_image(image="old1.jpg") original_size = image.size pipe = CogVideoXImageToVideoPipeline.from_pretrained( "CogVideoX-5b-I2V", torch_dtype=torch.bfloat16 )

pipe.enable_sequential_cpu_offload() pipe.vae.enable_tiling() pipe.vae.enable_slicing()

video = pipe( prompt=prompt, image=image, num_videos_per_prompt=1, num_inference_steps=50, num_frames=49, guidance_scale=6, generator=torch.Generator(device="cuda").manual_seed(42), ).frames[0]

export_to_video(video, "output.mp4", fps=8)

尝试过禁用其中的优化，速度会更慢或者相当 pipe.enable_sequential_cpu_offload() pipe.vae.enable_tiling() pipe.vae.enable_slicing()

Expected behavior / 期待表现

有方法可以加速吗？