Installing Guild AI

  1. Requirements
  2. Install Guild AI
    1. Install Guild AI with Virtualenv
    2. Install Guild AI with Conda
    3. Install Guild AI from source code
  3. Install TensorFlow
  4. Install optional libraries
    1. CUDA and cuDNN
    2. NVIDIA System Management Interface
  5. Verify your installation
  6. Next steps

Requirements

Guild AI has the following requirements:

  • Max OS, Linux
  • Python 2.7, Python 3
  • pip

Guild is installed from PyPI using pip. Refer to Installing pip to ensure you have it installed.

Install Guild AI

To install Guild AI, run the following on the command line:

pip install guildai

If you need to run install with administrative privileges, run:

sudo pip install guildai

You may alternatively install Guild AI in a Python virtual environment. Refer to the next sections for details.

Install Guild AI with Virtualenv

If you would like to install Guild within a Python virtual environment, run the following:

virtualenv guild
source guild/bin/activate
pip install guildai

This has the advantage of isolating Guild AI within a single directory. For more information on the advantages and disadvantages of this approach, see Working with Virtualenv.

Install Guild AI with Conda

Guild AI can be installed in Conda environments using pip.

conda create -n guild python=3.6
source activate guild
pip install guildai

If you’d like to install Guild with a different Python version, replace 3.6 above with the version you want.

Install Guild AI from source code

Note

This step is an alternative to installing Guild AI with pip described above. Install Guild AI from source code if you want a specific version from GitHub (e.g. an early release or development branch) or if you want to contribute to the project.

Additional required tools for installing from source code:

  • git
  • npm v5.8.0 or later
  • Python development library and headers for your system

To install Guild from source, clone the repository by running:

git clone https://github.com/guildai/guild.git

Change to the guild directory and install the required pip packages:

cd guild
pip install -r requirements.txt

Build Guild by running:

python setup.py build

Verify Guild by running:

guild/scripts/guild check

If see the message NOT INSTALLED (No module named 'tensorflow') that’s okay - you’ll install TensorFlow in the steps below. If you see other errors, please open an issue on GitHub and we’ll help!

You can run the GUILD_SOURCE_DIR/guild/scripts/guild executable directly (where GUILD_SOURCE_DIR is the location of your cloned Guild AI source repository) or modify your environment to make guild available on your PATH using one of these methods:

  • Add GUILD_SOURCE_DIR/guild/scripts directory to your PATH environment variable, OR
  • Create a symlink to GUILD_SOURCE_DIR/guild/scripts/guild that is available on your PATH

Install TensorFlow

Guild requires TensorFlow but does not install it for you. 1 You can use pip to install TensorFlow by running:

pip install tensorflow

If your system has a GPU, install the GPU enabled package by running:

pip install tensorflow-gpu

For alternative installation methods, refer to Installing TensorFlow.

Install optional libraries

If you system has a GPU or other accelerator supported by TensorFlow, you will need to install and configure support for your hardware.

CUDA and cuDNN

If you have an NVIDIA GPU and and want to use the GPU enabled TensorFlow package, you must install the NVIDIA CUDA and cuDNN libraries for your system. Refer to the links below for help installing the libraries.

NVIDIA System Management Interface

Guild uses NVIDIA System Management Interface (nvidia-smi) on GPU accelerated systems to collect GPU metrics. This tool is optional and Guild will run without it. However, to collect GPU stats on systems with one or more GPUs, ensure that nvidia-smi is installed.

Note

NVIDIA System Management Interface is typically installed with NVIDIA GPU drivers. Refer to NVIDIA System Management Interface for more information.

Verify your installation

Verify that Guild is installed properly by running the check command:

guild check

If there are problems with your installation, Guild will display the details and exit with an error. Refer to Troubleshooting for assistance.

Next steps

Congratulations, you’ve installed Guild AI! We’ve outlined some next steps for you below.

Train your first model

Dive in and train your first model using Guild AI. This introductory tutorial will walk you through the basics of Guild and cover most of its features.

Train your first model

Discover Guild models

Guild AI provides a catalog of state-of-the-art TensorFlow models that can be used to build deep learning applications. Start here to see what developers are building.

Discover Guild models

Browse the docs

If you're interested in a complete picture of Guild AI, start by browsing its comprehensives documentation.

Browse the docs

  1. TensorFlow is a rapidly evolving software library and is provided as both CPU and GPU supported packages. Guild leaves the specific package and version of TensorFlow up to the user.