Installation

NeCT has been tested on Windows and Linux with the following dependencies:

Package	Version	Notes
python	3.11 \| 3.12
pytorch	2.4 - 2.7
CUDA	12.X
CMake (Linux)	3.24	For tiny-cuda-nn
C++17 (Windows)		For tiny-cuda-nn

Recommended: Use conda or uv to manage your Python environment.

For video export, the avc1 codec for ffmpeg is only available with conda. With uv, video export uses the mp4v codec. If video export using avc1 is vital, use conda.

Tested with python=3.11, 3.12 and pytorch>=2.4,<2.8.

To install for multiple compute capabilities, see below.

Note: Ensure PATH and LD_LIBRARY_PATH include the CUDA binaries as described in tiny-cuda-nn. If you encounter installation errors related to tiny-cuda-nn, check their issues page. Building binaries for both tiny-cuda-nn and NeCT may take several minutes.

Uv Installation¶

If you don't have uv installed, follow the uv installation guide. This will install pytorch with CUDA 12.4.

uv venv --python=3.12
source venv/bin/activate
uv pip install git+https://github.com/haakonnese/nect[torch]
uv pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch --no-build-isolation

Custom PyTorch Version¶

To use a specific PyTorch version (2.4-2.7) visit the PyTorch Installation Page and install the desired PyTorch version into your uv environment. Then install NeCT with:

uv pip install git+https://github.com/haakonnese/nect --no-build-isolation-package torch
uv pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch --no-build-isolation

Conda Installation¶

conda create -n nect python=3.12 -y
conda activate nect
conda install pytorch==2.5.1 torchvision==0.20.1 pytorch-cuda=12.4 lightning==2.1 conda-forge::opencv -c pytorch -c nvidia -c conda-forge -y
pip install -v git+https://github.com/haakonnese/nect
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

Install Multiple Compute Capabilities¶

To build for multiple compute capabilities (e.g., 60=P100, 70=V100, 80=A100, 90=H100), set these environment variables before installing NeCT:

export CUDA_ARCHITECTURES="60;70;80;90"
export CMAKE_CUDA_ARCHITECTURES=${CUDA_ARCHITECTURES}
export TCNN_CUDA_ARCHITECTURES=${CUDA_ARCHITECTURES}
export TORCH_CUDA_ARCH_LIST="6.0 7.0 8.0 9.0"
export FORCE_CUDA="1"