🎉Flyte v1.5.0 Release
Introducing experimental features. Use them and give us your feedback!
- Data subsystem revamped (now using fsspec exclusively) leading to:
- Higher performance for most use cases
- Improved streaming capabilities for flytekit
- Limited support for partial tasks
- Fix ("freeze") certain parameters of a Task/MapTask
- Support for Python 3.11
- Changes to the database migrations process, as a necessary preparation for supporting databases other than Postgres in future releases.
For full release notes, visit the page on GitHub.
⚙ New RFC process
➡️ In line with the recent updates to the Governance model, there's a new process in place to ensure enough community discussion for ideas that could change Flyte's user experience. How it works and how you can participate are described here
📅 Multiple in-person events coming soon
🎵On the road again…
April brings a lot of opportunities to catch up with Flyte maintainers and community members.
- Atlanta ML Meetup: “The fundamentals of type-safe, reproducible, scalable data pipelines”
- April 19, 5p.m. ET
- ➡️ RSVP
- PyCon Seattle
- 2 talks happening at April 26 and 28
- Union.ai is a sponsor — bring your Flyte questions to the booth!
- PyCon US
🔔 New examples covering DuckDB, Banana and Hugging Face
New blog posts
How to serve models with Banana
Using Hugging Face transformers and a Flyte workflow, Samhita Alla explores in this post how to serve models using Banana's ML inference system.
Analyzing COVID-19 Impact on NYC Taxis with DuckDB
Leveraging the existing integration with DuckDB, Samhita also explores in this piece how to run complex queries in a dataset with minimum effort.
Keep on the streak, Samhita!✨