Fix GLM headers via apt + symlink instead of sdist package_data

Trigger rebuild: gsplat package_data now includes GLM headers

Trigger rebuild: gsplat sdist now includes GLM headers

Fix torch_scatter: switch from cu126 to cu130 wheel

Add xformers back to requirements (missing module at startup)

Trigger rebuild: gsplat sdist updated with CUDA_HOME auto-detect patch

Install gsplat from direct sdist URL instead of git clone

Trigger rebuild: gsplat-patched repo recreated without git-lfs

Use patched gsplat fork that fixes pip build isolation issue

Revert to Gradio SDK; align deps with HF torch 2.11.0+cu126

Switch to Docker SDK with CUDA devel image for Blackwell support

Fix Blackwell GPU (sm_120) compatibility for ZeroGPU

tried to make safetensors loaded on GPU

set share to True

modified the torch scatter link

changed torch version to 2.2.2 for Zero GPUs

changed the torch version to 2.4.0 to be compatible with Zero GPUs

changed the torch version to 2.4.0 to be compatible with Zero GPUs

zero gpu still not working

added torch version to readme

added torch version to readme

added torch version to readme

added @spaces.GPU to app.py

tried to fix the pytorch version error in configurations

no args in app.py

updated app.py file

runtime txt file is not needed any more

added links to the requirenment.txt

added requirements.txt

added extra index url

added the index url to the requirements.txt

added python version to readme

added python version to readme

added python version to readme

updated runtime.txt with pythonpython-3.10.18

tried to fix the packagelist error

import space and snapdownload

change the demo codes to the initial one, but with support for huggingface weights

updated app py file

updated app

Track all files under examples/ with Git LFS

initial commit