LHM++ Installation

March 16, 2026 · View on GitHub

Requirements

  • Linux (tested on Ubuntu)
  • Python 3.10
  • PyTorch 2.3.0
  • torchvision 0.18.0
  • CUDA 12.1 (recommended)

1. Clone the repository

git clone https://github.com/aigc3d/LHM-plusplus
cd LHM-plusplus

2. Install PyTorch and xformers

# CUDA 12.1 (recommended)
pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu121
pip install -U xformers==0.0.26.post1 --index-url https://download.pytorch.org/whl/cu121

3. Install base dependencies

pip install -r requirements.txt
pip install rembg[cpu]  # used during extracting sparse view inputs

4. Install pointops

cd ./lib/pointops/ && python setup.py install && cd ../../

5. Install spconv and torch_scatter

pip install spconv-cu121

# torch_scatter: see [wheel](https://data.pyg.org/whl/) for your CUDA version
# Example (PyTorch 2.3 + CUDA 12.1 + Python 3.10):
pip install torch_scatter-2.1.2+pt23cu121-cp310-cp310-linux_x86_64.whl

6. Install PyTorch3D

pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/py310_cu121_pyt230/download.html

7. Install diff-gaussian-rasterization

pip install git+https://github.com/ashawkey/diff-gaussian-rasterization/
# or
# git clone --recursive https://github.com/ashawkey/diff-gaussian-rasterization
# pip install ./diff-gaussian-rasterization

8. Install simple-knn

pip install git+https://github.com/camenduru/simple-knn/

9. Install gsplat

Download the pre-compiled wheel from gsplat whl.

# Example (PyTorch 2.3 + CUDA 12.1 + Python 3.10):
pip install gsplat-1.4.0+pt23cu121-cp310-cp310-linux_x86_64.whl

10. Download model weights

# Download prior models + pretrained weights (default)
python scripts/download_pretrained_models.py

# Prior models only (human_model_files, voxel_grid, BiRefNet, etc.)
python scripts/download_pretrained_models.py --prior

# LHM++ model weights only (LHMPP-700M, LHMPP-700MC, LHMPPS-700M)
python scripts/download_pretrained_models.py --models

Optional dependencies

SAM2 (for video segmentation)

We use a modified version of SAM2. Install only if needed for video processing:

pip install git+https://github.com/hitsz-zuoqi/sam2/
# or
# git clone --recursive https://github.com/hitsz-zuoqi/sam2
# pip install ./sam2

Windows installation

  1. Install Python 3.10 from python.org.
  2. Install CUDA 12.1 toolkit.
  3. Create a virtual environment and follow steps 2–10 above:
python -m venv lhmpp_env
lhmpp_env\Scripts\activate
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
# ... then run the remaining pip install commands from steps 3–9

Note: Adjust wheel filenames (torch_scatter, gsplat) for your Python version and CUDA. See PyG wheels and gsplat whl.

The installation has been tested with Python 3.10 and CUDA 12.1. For issues, refer to README.md or open an issue on GitHub.