【犀牛鸟实战issue】 macOS support with custom_rasterizer for CPU

[liangxg787]

【问题描述】目前代码环境配置都是在CUDA环境中，如何在macos中运行Hunyuan3D-2.1
【期望改进】在macos中运行Hunyuan3D-2.1
【完成方式】给出环境配置文档和代码

Summary of Changes

1. Added README_for_apple_silicon.md with setup instructions for Apple Silicon.

2. Added demo_for_apple_silicon.py as a usage example on Apple M-series chips.

3. Added device parameter to gradio_app.py for hardware flexibility.

4. Added compile_mesh_painter_for_apple_silicon.sh script for compiling on Apple Silicon.

5. Introduced new CPU-specific scripts for building custom_rasterizer:

   • grid_neighbor_cpu.cpp
   • rasterizer_cpu.cpp
   • rasterizer_cpu.h
   • setup.py

6. Added device parameter to textureGenPipeline.py.

7. Added new parameter face_count in simplify_mesh_utils.py, compatible with the latest trimesh version. Because the trimesh==4.4.7 could not work on MacOS.

8. Added requirements_for_apple_silicon.txt for environment setup on Apple Silicon.

Testing

• Platform: Apple Silicon (Apple M-series chip)

Notes

1. If you encounter a segmentation fault, it might be due to insufficient RAM on macOS.

2. Running the rasterizer in Hunyuan3DPaintPipeline on CPU is significantly slower. It is recommended to either:

   • Use an alternative rasterization method (to-do), or
   • Run Hunyuan3DPaintPipeline on a GPU for better performance.