Installation Instructions

NVIDIA Drivers

Install the latest proprietary NVIDIA drivers. The minimum supported CUDA version on neosr will always be the same as Pytorch's latest stable release.

Windows

Download and install from official nvidia website.

Linux-based systems

It is recommended that you install the proprietary drivers by adding the official repository to your package manager. Follow the instructions on NVIDIA website. The package cuda-toolkit is recommended as well. The open-dkms versions of the drivers have not been tested, but they should work as well. If you have issues with open-dkms, use the older dkms instead. Verify the CUDA version by using nvidia-smi on command line.

Note

Proprietary drivers are required, since we need CUDA.

---

Install scripts

To make installation easier, neosr has installation scripts that setup everything you need in a single command.

Windows:

Open a powershell terminal and paste:

powershell -ExecutionPolicy ByPass -c "irm https://raw.githubusercontent.com/neosr-project/neosr/refs/heads/master/install_windows.ps1 | iex"

Linux:

Open a terminal and paste:

curl -LsSf https://raw.githubusercontent.com/neosr-project/neosr/refs/heads/master/install_linux.sh | sh

Manual Installation

If you prefer not to use the installation scripts, manually installing is easy.

• Use git to clone the neosr repository:

git clone https://github.com/neosr-project/neosr

• Install uv by using the recommended command in their website.

• Go to the neosr folder and run:

cd neosr
uv cache clean
uv sync

• Make aliases for each neosr command (optional). Example (linux):

alias neosr-train='uv run --isolated train.py -opt'
alias neosr-test='uv run --isolated test.py -opt'
alias neosr-convert='uv run --isolated convert.py'
alias neosr-update='git pull --autostash && uv sync'

Uninstalling

If you only use uv for neosr, you can run the following commands to uninstall neosr completely:

uv cache clean
rm -r "$(uv python dir)"
rm -r "$(uv tool dir)"
rm -rf neosr/