OSError: PixArt-alpha/pixart_sigma_sdxlvae_T5_diffusers does not appear to have a file named config.json.

[Lexarymade]

now I face following error： OSError: PixArt-alpha/pixart_sigma_sdxlvae_T5_diffusers does not appear to have a file named config.json. how to fix this ?

[Lexarymade]

I've download the opensora model, the vae, and the t5. I just want to load the pre-downloaded model

[syc11-25]

same problem，by export HF_ENDPOINT=https://hf-mirror.com，sometimes works，sometimes also OSError

[zhengzangw]

I've download the opensora model, the vae, and the t5. I just want to load the pre-downloaded model

If you downloaded in your local machine, you should modify the config/opensora-v1-2/inference/sample.py, including

https://github.com/hpcaitech/Open-Sora/blob/65c61d0d5dd12a1cb7649d00caa65c2ce6600cc9/configs/opensora-v1-2/inference/sample.py#L18

https://github.com/hpcaitech/Open-Sora/blob/65c61d0d5dd12a1cb7649d00caa65c2ce6600cc9/configs/opensora-v1-2/inference/sample.py#L26

https://github.com/hpcaitech/Open-Sora/blob/65c61d0d5dd12a1cb7649d00caa65c2ce6600cc9/configs/opensora-v1-2/inference/sample.py#L33

[zhengzangw]

Otherwise, if you want the automatic downloading, just follow

same problem，by export HF_ENDPOINT=https://hf-mirror.com，sometimes works，sometimes also OSError

use

export HF_ENDPOINT=https://hf-mirror.com

[zhengzangw]

We've tested our inference script both in China and outside China, and both work. If after downloading it is still not working, please share your config and error, I will try to fix.

[liuxi0099]

Though, I have defined the path of pre-downloaded 3 models in Open-Sora/configs/opensora-v1-2/inference/sample.py. Using export HF_ENDPOINT=https://hf-mirror.com will download 2 models:

vae/config.json: 631B [00:00, 1.05MB/s]
diffusion_pytorch_model.safetensors: 100%|███████████████████████████████| 335M/335M [00:17<00:00, 18.9MB/s]
Loading checkpoint shards: 100%|██████████████████████████████████████████████| 2/2 [00:31<00:00, 15.61s/it]
config.json: 688B [00:00, 1.01MB/s]
model.safetensors: 100%|███████████████████████████████████████████████| 4.84G/4.84G [07:35<00:00, 10.6MB/s]

And a new bug occurs:

python3 gradio/app.py 
...
TypeError: STDiT3.__init__() got an unexpected keyword argument 'force_huggingface'

Maybe there are some bugs when migrating from v1.1 to v1.2?

[zhengzangw]

The downloading process seems correct.

For the next problem, we have fixed it as mentioned in #507. Please try pull the lastest code and run again.

[github-actions[bot]]

This issue is stale because it has been open for 7 days with no activity.