Installation

Linux

We distribute WarpTools as part of a conda package for Warp on x86-64 (linux-64) and ARM64 (linux-aarch64, including NVIDIA GH200). Both packages use CUDA 12.9; Conda selects the native architecture automatically. ARM64 packages are available starting with v2.0.0dev41.

Installing Conda

If you're new to the conda package manager we recommend installing mambaforge.

Creating a conda environment and installing Warp into it

The following command will create a new environment called warp and install warp and all dependencies into it.

conda create -n warp warp -c warpem -c nvidia/label/cuda-12.9.0 -c conda-forge --channel-priority flexible

The environment can then be activated whenever you want to use WarpTools

conda activate warp

Updating

To update your installation, run the following command

conda update warp -c warpem -c nvidia/label/cuda-12.9.0 -c conda-forge --channel-priority flexible

Checking what version you are running

To check which version you are running use

conda list warp

Windows

We don't currently provide pre-built binaries for WarpTools on Windows.