Remove remaining @spaces.GPU decorators from bake_albedo and bake_ao

Fix ZeroGPU proxy token expiry: single GPU allocation for full pipeline

Fix nvdiffrast SM mismatch: add +PTX arch + GL context fallback

Preload libcudart.so.13 via ctypes.CDLL before nvdiffrast import

Add CUDA 13.0 lib64 to LD_LIBRARY_PATH for nvdiffrast runtime linking

Fix nvdiffrast build: set TORCH_CUDA_ARCH_LIST=8.0;9.0 for headless compile

Add full build output to nvdiffrast install for diagnostics

Bypass CUDA version mismatch for nvdiffrast build via nvcc shim

Switch to torch 2.9.1 + runtime nvdiffrast install

Revert to torch 2.6.0 + Jeff Xiang prebuilt nvdiffrast wheel

Install nvdiffrast at runtime instead of build time

Fix packages.txt: remove comments (apt-get chokes on them)

Fix nvdiffrast ABI mismatch + pymeshfix TrackedArray crash

Fix pymeshlab pipeline crash + add post-process progress streaming

Fix all post-processing errors on ZeroGPU/headless environment

Fix state persistence across ZeroGPU subprocess boundary

Fix viewer not updating: direct GLB output + allowed_paths

Fix 3D viewer not updating after generation

Add streaming progress to generate_hunyuan + fix omegaconf dep

Add onnxruntime for rembg background removal

Implement Hunyuan3D-2.1 generation with background removal

Fix: preprocess image on remote TRELLIS.2 Space before calling image_to_3d

Polish: preset loading, quota recording, coacd fallback

Fix: replace PyPI nvdiffrast (doesn't exist) with Jeff Xiang's prebuilt wheel

Milestones 9-12: UniRig, UE5 Export, Presets, Polish

Milestones 6+8: albedo/material/AO baking + ORM pack, LODs, collision, pivot, scale

Milestone 4+5: UV unwrap (xatlas), symmetry, and normal map baking (nvdiffrast)

Milestone 3: Stage 2A-2C Mesh Cleanup (repair, cleanup, decimation)

Switch Stage 1 to gradio_client: all TRELLIS.2 custom wheels ABI-broken on torch 2.8+

Patch trellis2/representations/mesh/base.py: guard cumesh and flex_gemm imports

Fix: un-ignore TRELLIS.2/trellis2/models/ — was blocked by 'models/' in .gitignore

Fix: evict pip-installed trellis2 stub from sys.modules before loading vendored copy

Fix: make trellis2/__init__.py fully lazy to survive partial-module namespace

Fix circular import: bootstrap trellis2 at app.py top-level before src.stages import

Fix: bump torch to 2.8.0 for ZeroGPU compatibility (2.6.0 no longer supported)

Milestone 2: Wire TRELLIS.2 generation + vendor both model packages

Upload app.py

Delete app.py

Update requirements.txt

Update README.md

Upload README.md

Delete README.md

Upload 2 files

Delete app.py

Delete requirements.txt

Upload app.py

Delete app.py

Upload app.py

Delete app.py

Create presets/.gitkeep