# Install CUDA

In order to run the inference service on Atoma, Nvidia cuda toolkit is required to be installed locally. Please follow the official NVIDIA installation guide for you OS:<br>

<https://developer.nvidia.com/cuda-downloads?>