Hello there! The diligent team has been hard at work, and I've compiled a summary of their efforts by analyzing 16 events from the past week.

This week, the team expanded BentoML's deployment options, added support for Pydantic v2, and squashed some pesky bugs. Keep up the great work!

Adaptive Advancements 🤖

• Added support for adaptive batching of custom parameter types in BentoML, allowing users to achieve adaptive batching by wrapping the inference function with a custom Runner (Frost)

Helpful Hints 📝

• Added note to inform users about file rotation issue when initializing custom logger in BentoML service.py (PenHsuanWang)

Bug Busters 🐛

• Fixed a bug in the transformers module that caused incorrect GPU device to be set (Aaron)
• Fixed an issue where subprocess build did not respect user environment, causing import_model logics to fail and create a default bentoml_home instead of finding the model from the custom BENTOML_HOME (Aaron)

Pydantic Power-ups 🌟

• Added support for Pydantic v2 in the pydantic_components_schema function and removed UnprocessableEntity exception and related test code (Aaron)

Model Magic 🧙

• Updated the version of a machine learning model and its dependencies in an example code (Bojiang)

Deployment Delights 🚀

• Added information on different deployment options for BentoML, including cloud deployment solutions, feature comparison, and how to containerize Bentos as Docker images (Chaoyu)
• Added details on deploying with Yatai on Kubernetes, deploying with BentoControl, and deploying to BentoCloud (currently in private beta) (Chaoyu)