TensorFlow projects

Add guild.yml to your TensorFlow or Keras project to automate tasks and enable experiment tracking, testing, distribution and remote training. Learn more.

Model operations

Train your model by running the command guild run train to generate a unique experiment that preserves training files and metadata. Learn more.

Experiments

Capture results as unique experiments by running operations—compare performance, diff changes, visualize with TensorBoard and backup to the cloud. Learn more.

End-to-end workflow

Automate your model life cycle from data collection and pre-processing to training and evaluation to optimization and deployment. Learn more.

Automate testing

Verify all stages of your model workflow with tests that exercise code and check results such as expected loss and accuracy ranges. Learn more.

Packages

Save time and reduce errors by reusing tested, proven models and operations—developed by you and others in the TensorFlow community. Learn more.


More features

Train remotely on Amazon EC2

Guild supports remote training on EC2, including setup and tear down of EC2 infrastructure. If your local compute resources aren't enough for a training operation, start an EC2 remote and run your operation there.

Learn more

Backup runs to S3

Backup runs on S3 for safe keeping with Guild. All of Guild's run management commands are supported for S3, including list, delete, restore, and label.

Learn more

Visualize runs with TensorBoard

TensorBoard is a powerful visualization tool for TensorFlow. Use Guild to start TensorBoard to view runs with a single command. Guild automatically synchronizes TensorBoard with your current set of runs---you only need to start TensorBoard once.

Learn more

Compare run performance and diff changes

As you run operations, Guild automatically captures run logs and files so you can study and compare them. Guild seamlessly indexes TensorFlow event logs to capture training and evaluation metrics, letting you quickly view model performance across runs.

Learn more

Publish packaged models to PyPI

To share your models with other developers, Guild lets you quickly generate Python packages and upload them to PyPI. Users install your models using pip, Conda, or Guild.

Learn more


Get started

Guild AI introduction

If you're new to Guild AI, this introduction covers core features and functionality. Start here to learn about projects, models and operations, runs, end-to-end workflow, automated testing and packaging.

Guild AI introduction

Add Guild to your project

If you have scripts that train your TensorFlow or Keras models, add a Guild file (i.e. a file named guild.yml) to your project to enable Guild features.

Add Guild to your project

Create an image classifier

Guild supports a variety of feature-rich packages and project templates. This guide steps you through the process of creating an image classifier.

Create an image classifier

Browse Guild AI documentation

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

Browse documentation