Prerequisites#

docker run -it --rm --runtime=nvidia --gpus all ubuntu nvidia-smi

Example output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.78.01    Driver Version: 525.78.01    CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| 41%   30C    P8     1W / 260W |   2244MiB / 11264MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

Note

For more information on enumerating multi-GPU systems, please see the NVIDIA Container Toolkit’s GPU Enumeration Docs

Supported Hardware#

The AlphaFold2-Multimer NIM is configured to run on a single GPU. The minimum GPU memory requirement for the AlphaFold2-Multimer NIM is 32GB. The AlphaFold2-Multimer NIM should run on any NVIDIA GPU that meets this minimum hardware requirement and has compute capability ≥ 8.0. The AlphaFold2-Multimer NIM also requires at least 512GB of free hard drive space to store the various MSA databases required by endpoints exposed from the NIM.

In summary, users looking to successfully run the AlphaFold2-Multimer NIM for short sequences / multimers should have as system with:

  • One NVIDIA GPU with ≥ 32GB of VRAM and Compute Capability ≥ 8.0.

  • At least 64 GB of RAM.

  • A CPU with at least 24 available cores.

  • At least 512GB of free SSD drive space.

For optimal performance on long sequences / multimers and multiple MSA databases, we recommend a system with:

  • At least one NVIDIA GPU with 80GB of RAM (e.g. A100 80GB).

  • At least 128GB of RAM.

  • A CPU with at least 36 available cores.

  • At least 512GB of free fast NVMe SSD drive space.

NGC (NVIDIA GPU Cloud) Account#

  1. Create an account on NGC.

  2. Generate an API Key.

  3. Log in to Docker with your NGC API key: docker login nvcr.io --username='$oauthtoken' --password=${NGC_CLI_API_KEY}

NGC CLI Tool#

  1. Download the NGC CLI Tool <https://org.ngc.nvidia.com/setup/installers/cli> for your OS.

Important

Use NGC CLI version 3.41.1 or newer. Here is the command to install this on AMD64 Linux in your home directory:

wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/ngc-apps/ngc_cli/versions/3.41.3/files/ngccli_linux.zip -O ~/ngccli_linux.zip && \
unzip ~/ngccli_linux.zip -d ~/ngc && \
chmod u+x ~/ngc/ngc-cli/ngc && \
echo "export PATH=\"\$PATH:~/ngc/ngc-cli\"" >> ~/.bash_profile && source ~/.bash_profile

  1. Set up your NGC CLI Tool locally. (You’ll need your API key for this!)

ngc config set

Note

After you enter your API key, you may see multiple options for the org and team. Select as desired or hit enter to accept the default.

  1. Log in to NGC.

You’ll need to log in to NGC via Docker and set the NGC_API_KEY environment variable to pull images:

docker login nvcr.io
Username: $oauthtoken
Password: <Enter your NGC key here>

Then, set the relevant environment variables in your shell. You will need to set the NGC_CLI_API_KEY variable:

export NGC_CLI_API_KEY=<Enter your NGC key here>

  1. Set up your NIM (model) cache.

The NIM cache allows you to download models and store previously-downloaded models so that you don’t need to download them again later when you run the NIM again. The NIM cache must be readable and writable by the NIM, so in addition to creating the directory, the permissions on this directory need to be set to globally readable writable. The NIM cache directory can be set up as follows:

## Create the NIM cache directory.
mkdir -p /home/$USER/.cache/nim

## Set the NIM cache directory permissions to 777.
chmod -R 777 /home/$USER/.cache/nim

## If you hit permissions issues after running the NIM and downloading the model for AlphaFold2,
## set model & database permissions to 777 as well. Required for running the NIM!
sudo chmod -R 777 /home/$USER/.cache/nim/alphafold2-data_v1.1.0

Important

If you experience issues downloading the model via running the NIM, such as deploying in an “air-gapped” environment, you can manually download the AlphaFold2 model using the following command: ngc registry model download-version nim/deepmind/alphafold2-data:1.1.0 followed by sudo chmod -R 777 /home/$USER/.cache/nim/alphafold2-data_v1.1.0

Throughout this documentation, the $LOCAL_NIM_CACHE environment variable is assumed to use the local ~/.cache/nim directory. You can set the model cache path to any directory. NVIDIA recommends using a high-speed SSD for fast read and write access to the AF2 model and databases. Run the sudo chmod -R 777 $LOCAL_NIM_CACHE command to avoid file permission issues with the NIM container, regardless of the specified directory.

Now, you should be able to pull the container and download the model using the environment variables. To get started, review the Quickstart Guide.

On this page