Union at MLOps World Conference 2023
As the team behind Flyte (the Kubernetes-native AI orchestrator for production-grade ML and data pipelines), Union.ai is thrilled to attend and sponsor the MLOps World Conference for the second year in a row! The event will be held at the Renaissance Austin Hotel from October 25-26. Be sure to register here if you haven’t already!
MLOps World is an international community of practitioners established by the Toronto Machine Learning Society (TMLS) to support deployment of machine learning models in live production environments. We are proud to be a part of this event and support the ML community and ecosystem.
Learn Your Codebase: Fine-tuning Code Llama with Flyte … to Learn Flyte
Foundation LLMs can only be trained by a handful of organizations; few possess the compute resources required to pre-train models with more than a hundred billion parameters of data.
Instead of building their own foundation models from scratch, the larger ML community fine-tunes existing LLMs for specific applications. However, fine-tuning carries its own challenges — including provisioning the appropriate infrastructure.
During a virtual workshop Oct. 24, Union.ai Chief Machine Learning Engineer Niels Bantilan will demonstrate how Flyte addresses the infrastructure challenge by providing a declarative way to specify the resources required for machine-learning tasks, including fine-tuning LLMs.
Attendees will learn how to fine-tune Code Llama with Flyte, tailoring it to their specific codebase to create their own GitHub co-pilot or LLM-driven code assistant. They will also gain hands-on experience using Flyte to leverage state-of-the-art deep learning tools such as `torchrun` distributed training, LoRA, 4/8-bit quantization and FSDP, while benefiting from Flyte's reproducibility, versioning and cost-management capabilities.
At the end of this workshop, attendees will possess the knowledge and code to apply the same fine-tuning principals to their own codebases to answer user support questions, generate boilerplate starter code or tackle any other downstream tasks!
The workshop runs from 12:30 to 2:00 p.m. CST. Register for it here.
Meanwhile, feel free to check out additional resources below:
- Large Language Models in Production Cheat Sheet
- Fine Tune Llama 2 with Limited Resources
- Fine-Tuning vs Prompt Engineering Large Language Models
- Fine-Tuning Insights: Lessons from Experimenting with RedPajama Large Language Model on Flyte Slack Data
- Fine-Tuning Insights: Using LLMs as Preprocessors to Improve Data Quality
Meet the Team + Giveaway!
If you have any questions about Flyte, come stop by Booth 23 to chat with our team or feel free to reach out to us on Slack! We’ll also be hosting a raffle for the Sony WH-1000XM4 Wireless Premium Noise Canceling Headphones, so be sure to stop by to enter for a chance to win. We look forward to meeting and hearing all about your current ML projects and initiatives. Register to attend this event here.
How to enter our giveaways
1. Sony Headphones Giveaway
Choose one of the following ways to enter below:
- Fill out this raffle form
- Share a picture featuring our swag at the booth and tag us either on LinkedIn at Union.ai or X at @union_ai with the hashtags #BetterAIPipelines #MLOpsWorld2023 (this option will be 3 entries).
1 winner will be drawn from the entires each conference day, so visit us both days for chances to win!
We will announce a winner at our booth at the following times:
- Wed, Oct 25 at 4:30 PM
- Thu, Oct 26 at 3:00 PM
If you are chosen and are not present, we will email you. Please coordinate to meet to receive your prize.
2. Union Tumblers
Choose one of the following ways to enter below:
- Follow us on X at @union.ai or LinkedIn at Union.ai
- Star us on GitHub at flyteorg/flyte
- Join our Community on Slack
Learn more about Flyte
- Join the Flyte Community on Slack
- Check out and star the Flyte repo on GitHub
- Try the Flyte hosted sandbox
- Check out the Flyte docs
- Follow us on LinkedIn and Twitter
