GPUs

In this guide, we will show you how to run your extractor on a AWS EC2 instance running on Ubuntu. We will skip the steps on how to spin up an EC2 instance as there are already many guides available online on how to do that.

​

Step 1: Install NVIDIA Drivers

First of all, try running the command below to see if you have NVIDIA drivers installed:

nvidia-smi

If you are able to see the NVIDIA driver version, then you are good to go and you can skip this step. If not, you need to install the NVIDIA drivers by running following this guide: Installing NVIDIA Drivers on Ubuntu.

You might want to choose the driver dedicated for a server instead of the desktop version. After installing the drivers, you might need to reboot your machine.

​

Step 2: Install NVIDIA Container Toolkit

Next, you need to install the NVIDIA Container Toolkit. This is required to run Docker containers on your GPU. You can install it by following this guide: Installing NVIDIA Container Toolkit.

Make sure to follow the guide to configure the toolkit with Docker.

​

Step 3: Run Executor with GPU

docker run \
    --gpus all \
    tensorlake/default-executor-image \
    indexify-cli\
    executor \
    --server-addr $ADDRESS:8900