🚀 Scale Smarter, Not Harder 🚀
Flyte unifies your data and ML pipelines with cloud-native scalability!
✅ Dynamic workflows adapt to changing needs—no more rigid DAGs.
✅ Cut costs by 80%+ with spot instances—just set interruptible=True.
👉 Learn more: https://t.co/EmMg7Xgde7
🚀 We're excited to share what's new in Flyte 1.15!
Major highlights:
📊 Streaming Decks
🔐 Better Secrets Handling
⚡️ Enhanced Task Caching
📦 Quality of Life Updates
Full release notes: https://t.co/f3W0cbFKHq
Thanks to 12 new contributors who made this possible! 🙌
- Deeply integrated with Kubernetes
- Using AWS-native constructs
- Reliable by design
This is what the Flyte reference implementation for EKS achieves in minutes using Terraform/OpenTofu.
Get started: https://t.co/KgrR9Q2IIQ
Read the blog: https://t.co/bqgGvL9WEz
@hyperprior@Wooltard Hey!
So the immutability aspect partially has to do with caching but is optional so not completely a system property.
By default, every execution gets a version ID that cannot be altered.
Also, it instantiates a CRD every time which stores state
🔥Go from hidden states and hard-to-share experiments to reproducible executions.
🚀Tap into elastic compute resources straight from your Notebooks!
Learn how Flyte makes it possible:
https://t.co/9UXHbkWQBO
🔥 Learn how to go from zero to production-grade ML orchestration on #Azure:
https://t.co/c4z14ZITOe
Thanks to @Azure team for their input and feedback; it enabled us to add integrations with Azure Container Insights, AKS node auto-upgrade, and more!
We just reproduced benchmarks of the Liger Kernels package using:
- Flyte
- One A100 GPU
- Phi3 Mini
Results?
🔥 Effectively 20% -30% higher token throughput with 6% - 13% less reserved memory compared to vanilla transformers.
Dive into the details!
https://t.co/zzeGDpUzhS
🏎️The new reference implementation for @Azure takes you to a production Flyte deployment in minutes, using Terraform/OpenTofu.
Learn more: https://t.co/Ug6w1rao1R
Code: https://t.co/9S7ISLiHsX
Welcoming Rafael Raposo to Flyte's Technical Steering Committee. We're grateful for @SpotifyEng's support to the project throughout the years, and Rafael's personal commitment to represent our community values.
https://t.co/2fA51efHcQ
Welcoming Rafael Raposo to Flyte's Technical Steering Committee. We're grateful for @SpotifyEng's support to the project throughout the years, and Rafael's personal commitment to represent our community values.
https://t.co/b8r6d70Hv4
Our next Community sync will include 2 🔥 sessions:
1️⃣9:00 - 9:30 am PT
How @go_brain_space uses Flyte to develop a new breed of models 🧠
2️⃣9:30 - 10:00 am PT
💥Live demo of @union_ai's Serverless platform +Q&A moderated by @kelseyhightower!
Register: https://t.co/yYIl4RBUEQ