Industrial Uses

Google Cloud Vertex AI

PyGX is the driving force behind Vertex AI NAS, which has brought advanced AutoML technologies to industries with significant impact. Companies such as Qualcomm, Varian, and Oppo have all benefited from this groundbreaking technology.

Vertex AI has published a few search spaces for computer vision as examples:

• EfficientNet — paper · code
• SpineNet — paper · code
• MNASNet — paper · code
• NASFpn — paper · code
• AutoAugment — paper · code

Pax

Pax is a powerful machine learning framework developed by Google, based on Jax, that is designed for training large-scale models. It utilizes PyGX to enable its hyperparameter tuning and AutoML capabilities. Pax serves as a good example of how PyGX can be seamlessly integrated into a large-scale ML codebase based on dynamic evaluation:

• Inspecting the search space
• Implementing the tuning loop

Vizier

Vizier is the distributed tuning solution at Alphabet. PyGX uses it as a backend for serving distributed AutoML scenarios at Google. The open-source Vizier has shown how PyGX can be used together with Vizier; it also serves as an example of how a PyGX backend could be developed.

• Implementing the Backend interface
• Implementing the Feedback interface