Spaces:

ali-vilab

/

modelscope-text-to-video-synthesis

like 1.41k

Running on A10G

@Zaesar
I'm not sure, but did you assign a GPU to your Space? Maybe the error occurs if a GPU is not available. FYI, this Space currently uses "A10G large".

I use a script provided by Poli to when somebody wants to use it they have to duplicate it to not use GPU from the ORganisation. In all the other spaces it works fine.

@Zaesar I tried to run this Space on a CPU environment and I got an error too. It seems the model weight files are saved as CUDA tensor and the ModelScope library tries to load them on GPU, so the error occurs on CPU-only environment.

Hope by putting the "A10G large" and adding the snippet Poli gives me when a user wants to use it it doesn't use the GPU from the Organization.

https://huggingface.co/spaces/AIFILMS/modelscope-text-to-video-synthesis

Any suggestions?

@Zaesar I realized that you'd like to provide a shared UI Space, so I opened a PR. https://huggingface.co/spaces/AIFILMS/modelscope-text-to-video-synthesis/discussions/1
I think the error won't occur on CPU and it will work as a shared UI Space that users need to duplicate and assign a GPU to use it.

Ok, done!

Appreciate your help

@Zaesar

Have seen you opened a pull on Community. Tried to pull it into main but now it isn't working.

Sorry. Found a bug and opened another PR to fix it. https://huggingface.co/spaces/AIFILMS/modelscope-text-to-video-synthesis/discussions/2

Have you done it to work on CPU but if user wants to use it has to use their GPU?

The shared UI Space can be up on a CPU, but people need a GPU to run inference.