Skip to content

installation_cpu.md

Latest commit

 

History

History
66 lines (55 loc) · 2.09 KB

installation_cpu.md

File metadata and controls

66 lines (55 loc) · 2.09 KB

Installation Instruction (CPU-only version)

We recommend setting up the environment using Miniconda or Anaconda. We have tested the code on Linux with Python 3.10, PyTorch 2.3.1 (CPU), but it should also work in other environments.

Step 1: create an environment

Clone this repo:

git clone https://github.com/your-repo/PinPoint3D.git
cd PinPoint3D

Create and activate conda environment:

conda create -n pinpoint3d python=3.10
conda activate pinpoint3d

Step 2: Install PyTorch (CPU version)

pip install torch==2.3.1 torchvision==0.18.1 --index-url https://download.pytorch.org/whl/cpu

Step 3: Install Minkowski Engine (CPU version)

3.1 Install dependencies:

conda install -c intel mkl mkl-include

3.2 Install MinkowskiEngine from source:

git clone https://github.com/NVIDIA/MinkowskiEngine.git
cd MinkowskiEngine
python setup.py install --blas=mkl --cpu_only
cd ..

Alternative method (if above fails):

pip install MinkowskiEngine --no-deps

If you encounter issues, refer to MinkowskiEngine's official CPU-only compilation guide.

Step 4: Install other required packages

# Essential packages (tested versions)
pip install numpy==2.2.5 matplotlib==3.10.1 scipy==1.15.2
pip install open3d==0.19.0
pip install wandb==0.19.8  # For experiment tracking
pip install plyfile==1.1  # For PLY file I/O

Step 5: Verify installation

Test if everything is installed correctly:

python -c "import torch; print('PyTorch:', torch.__version__)"
python -c "import MinkowskiEngine as ME; print('MinkowskiEngine installed successfully')"
python -c "import numpy, matplotlib, wandb, open3d; print('All packages installed')"

Performance Note

⚠️ Training on CPU is extremely slow! A single epoch that takes minutes on GPU may take hours on CPU. Consider:

  • Using a smaller dataset for testing
  • Reducing the number of points per scene
  • Using lower batch sizes
  • Cloud GPU services (Google Colab, AWS, etc.) for actual training