THUDM / CogVideo

text and image to video generation: CogVideoX (2024) and CogVideo (ICLR 2023)
Apache License 2.0
9.31k stars 875 forks source link

Frame Interpolation is causing an error #345

Open Enchante503 opened 2 months ago

Enchante503 commented 2 months ago

System Info / 系統信息

Windows11Pro WSL2 Ubuntu 22.04.4 LTS CUDA12 RTX4090

Information / 问题信息

Reproduction / 复现过程

I tried to make a video from the sample image at https://huggingface.co/spaces/THUDM/CogVideoX-5B-Space using I2V + Frame Interpolation (8fps -> 16fps), but I got an error.

When I tried it locally (gradio_composite_demo) I got the following error: RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 480 but got size 496 for tensor number 1 in the list.

In the local environment, add the following code:

pipe_image.enable_sequential_cpu_offload()
pipe_image.vae.enable_tiling()

Expected behavior / 期待表现

The video is created successfully.

zRzRzRzRzRzRzR commented 2 months ago

@glide-the