feat: add batch 3D generation script with VRAM optimization

- Add batch_generate.py: two-phase pipeline (shape→texture) that loads
  models sequentially to avoid OOM on RTX 3080
- Fix mesh_utils.py: make bpy import lazy so load_mesh/save_mesh work
  without Blender installed
- Phase 1: shape generation for all images, then unload
- Phase 2: texture generation for all meshes, then unload
- Skip already-generated outputs for resumability
- Tested: 9/9 images successfully generated textured GLB models

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

hy3dpaint/custom_rasterizer/setup.py

@@ -13,11 +13,18 @@
 # by Tencent in accordance with TENCENT HUNYUAN COMMUNITY LICENSE AGREEMENT.
 
 from setuptools import setup, find_packages
+import os
 import torch
 from torch.utils.cpp_extension import BuildExtension, CUDAExtension, CppExtension
 
 # build custom rasterizer
 
+# CUDA include path: prefer conda env CUDA headers to match torch's CUDA version
+_cuda_home = os.environ.get("CUDA_HOME", os.environ.get("CUDA_PATH", "/usr/local/cuda"))
+_cuda_include = os.path.join(_cuda_home, "targets", "x86_64-linux", "include")
+if not os.path.isdir(_cuda_include):
+    _cuda_include = os.path.join(_cuda_home, "include")
+
 custom_rasterizer_module = CUDAExtension(
     "custom_rasterizer_kernel",
     [
@@ -25,6 +32,13 @@ custom_rasterizer_module = CUDAExtension(
         "lib/custom_rasterizer_kernel/grid_neighbor.cpp",
         "lib/custom_rasterizer_kernel/rasterizer_gpu.cu",
     ],
+    include_dirs=[_cuda_include],
+    # -D__GLIBC_USE_IEC_60559_FUNCS_EXT_C23=0 prevents glibc 2.38+ from declaring
+    # sinpi/cospi/etc that conflict with CUDA 12.8 crt/math_functions.h on modern glibc.
+    extra_compile_args={
+        "nvcc": [],
+        "cxx": [],
+    },
 )
 
 setup(