Installing Guild AI on your system and use it to build and monitor TensorFlow™ models
This is a short overview of Guild and its features. If you’re ready to install Guild, go to Setup.
Guild AI is a set of tools to enhance TensorFlow™
development. Start by adding a simple
Guild file to the root of your
Guild uses this information to execute the script
train.py when you
run this command:
Just before executing your script, Guild creates a unique run
directory that contains training logs, checkpoints, exported models,
and additional data that Guild collects. This directory can be
referenced using the
RUNDIR environment variable, or provided as a
command line option to your script.
Here’s what Guild provides:
You can view your training operation in progress by running the
command in a separate terminal:
Because each model training is unique, Guild lets you customize your view for each model. Here’s an example that shows the MNIST training in progress — it reads the validation and training accuracy from the TensorFlow event logs:
When it’s time to run your model as a part of your application, use
This runs your model in an light weight HTTP server, creating a web
service for making predictions. You can even use
curl (or any HTTP
client) to test your model:
That’s Guild AI in a nutshell! Use it to augment your TensorFlow development to train, evaluate, and finally run your models. Read more about Guild features or click below to get started using Guild on your system.