Spaces:
No application file
Apply for community grant: Academic project (gpu and storage)
Unlock Pose Diversity: Accurate and Efficient Implicit Keypoint-based Spatiotemporal Diffusion for Audio-driven Talking Portrait. Code and demo are available at https://github.com/chaolongy/KDTalker.
Hi @ChaolongYang , we've assigned ZeroGPU to this Space. Please check the compatibility and usage sections of this page so your Space can run on ZeroGPU.
Hi, @hysts ; we plan to deploy the Demo in Spaces but found that the Storage limit is 1GB, which cannot accommodate our model weights. Can you give us more free space? Thank you for your reply.
@ChaolongYang Each Space has 50GB of ephemeral storage (docs), so I believe you don't need persistent storage. But if it didn't work, feel free to ping me again.
BTW, I'm not 100% sure, but the 1GB storage limit just means that you can't add large files to your Space repo. You can still upload your model to a model repo and use the huggingface-hub library to download it to your Space at runtime (docs). The reason we restrict the size of Space repo is because large files make the docker image of the Space and it significantly slows down the startup time of the Space.
@hysts
Thanks for the notice. We tried to deploy it on another space for testing but encountered the following error:TypeError: cannot pickle onnxruntime.capi.onnxruntime_pybind11_state.InferenceSession' object.
The project is located at the address YinuoGuo27/test_kdtalker. It works perfectly fine locally but raises this error when running it on ZeroGPU. Do you have any suggestions for fixing this problem?