Fix Genesis from_torch patch: avoid torch ops on genesis tensors inside ZeroGPU

Patch Genesis from_torch to fix PyTorch 2.5 tensor subclass error

Fix Genesis RuntimeError: don't move simulator tensors to GPU

CPU-first startup: load all models/scenes to CPU at module level, GPU transfer at generation time

Fix TypeError: gr.Group does not accept label argument in gr.update()

Fix ZeroGPU illegal duration: defer all GPU work to do_generate

Fix ZeroGPU illegal duration: reduce to 240s, download models outside GPU slot

Fix ZeroGPU Error: make _on_load itself @spaces.GPU instead of calling it

Fix empty scene dropdown on ZeroGPU: move startup into demo.load()

Fix 3 bugs and remove tracked sim_tmp runtime files

Patch gradio_client boolean schema bug at import time

Fix EGL/OpenGL unavailable for Genesis headless rendering

Fix checkpoint discovery: use rglob instead of glob

Migrate to Gradio SDK with ZeroGPU + H200 support

load all cases at first

init huggingface deployment