THUDM / CogVideo

Text-to-video generation: CogVideoX (2024) and CogVideo (ICLR 2023)
Apache License 2.0
7.4k stars 678 forks source link

I2V model PT seems corrupt #291

Open phr00t opened 3 days ago

phr00t commented 3 days ago

System Info / 系統信息

transformers-4.43.3 Windows 11 x64

Information / 问题信息

Reproduction / 复现过程

Download the PT seen here, URL provided in GitHub comments:

https://cloud.tsinghua.edu.cn/d/5cc62a2d6e7d45c0a2f6/?p=%2F1&mode=list

... try to load it in PyTorch:

model = torch.load(model_path) An error occurred: PytorchStreamReader failed reading zip archive: failed finding central directory

Expected behavior / 期待表现

Correctly loading the PyTorch file.

zRzRzRzRzRzRzR commented 2 days ago

We have not yet released the I2V model, it is expected to be released this month, and the final preparations for the model are underway

xvjiarui commented 16 hours ago

Hi @zRzRzRzRzRzRzR

Great job on releasing I2V model! Is there any plan on updating the I2V model download link for SAT?

zRzRzRzRzRzRzR commented 13 hours ago

https://cloud.tsinghua.edu.cn/d/5cc62a2d6e7d45c0a2f6/?p=%2F1&mode=list here for how to use it, follow sat/README.md

xvjiarui commented 12 hours ago

Hi @zRzRzRzRzRzRzR Thanks for your prompt reply. I just tried this file, it still raise "RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory"

runzeer commented 12 hours ago

@zRzRzRzRzRzRzR It seemed the link https://cloud.tsinghua.edu.cn/d/5cc62a2d6e7d45c0a2f6/?p=%2F1&mode=list
has not been refreshed. It seemed that the model weights is 5 days ago. And the weights can not be loaded by pytorch.

zRzRzRzRzRzRzR commented 9 hours ago

I discovered this issue, and my colleague is handling it. It should be resolved within 12 hours. In the worst case, we will re-upload the model and update the link. Please keep an eye on this issue, and we will update as soon as possible

The diffusers version can be used normally

zRzRzRzRzRzRzR commented 4 hours ago

will it work now? upload again

hanshumin001 commented 3 hours ago

hi, When I use the latest version model,there still are some errors.

[rank0]: RuntimeError: Given groups=1, weight of size [3072, 32, 2, 2], expected input[26, 16, 60, 90] to have 32 channels, but got 16 channels instead.