INSTALL.md

Installation

The code has been tested with python 3.10, torch==2.4.0, torchvision==0.19.0 and torchsparse==2.1.0, with CUDA 12.4. We provide different options to set up the environment and repreduce the results.

Python Environment

Quick Setup

Create a virtual environment and install packages from the requirements.txt file:

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Step-by-step installation

Install PyTorch 2.4.0 with CUDA 12.4:

pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu124

Clone and build TorchSparse 2.1.0:

git clone https://github.com/mit-han-lab/torchsparse.git && cd torchsparse
pip install -r requirements.txt
python setup.py install

Install remaining dependencies:

pip install pyyaml opencv-python-headless matplotlib tensorboard fvcore timm scipy strictyaml scikit-learn tqdm easydict

Singularity

Pre-built Image

We provide a pre-built Singularity image for a quick setup and usage of the code. You can find the image in this Drive folder.

singularity exec --nv ubuntu_cuda124_torchsparse.sif <script>

Build Image

We provide a .def file to build from scratch a singularity image with the required packages to reproduce and use the code.

### build singularity image
singularity build --fakeroot ubuntu_cuda124_torchsparse.sif ubuntu_cuda124_torchsparse.def

### run the script with the singularity image
singularity exec --nv ubuntu_cuda124_torchsparse.sif <script>