Sage Elliott

The Flyte 2.0 SDK is officially here: This release brings some really exciting local development features that you can right out of the box even with connecting to Kubernetes or Docker! The features I've been loving to my local AI workflows: - Terminal User Interface (live workflows and history) - HTML Reports (tracks to workflow version - Caching (skip rerunning a task by reading from cache) - Retries (auto retry a task if it fails - super useful for API calls) - And more All this together gives you: - great lightweight experiment tracking - a lift in reliability - a great interface for debugging large pipelines or AI agent runs I can't wait to show you whats next in both Flyte and @union_ai

Scale AI Agents & Workflows Across Containers Using asyncio Flyte 2 lets you use asyncio.gather(), the same tool you'd use in any async Python app to fan out tasks in parallel. Each awaited task call becomes a distributed task on your cluster, running in its own container with its own resources.

Open-Source Video Generation: Diffusion & Transformer Models - AI Build & Learn x.com/i/broadcasts/1…

working from balcony with rain and thunder is glorious

I brought this up in my talk tonight showing open source video generation models. Some still produce weird things but I'm going to miss the weirdness when it's gone 😭

You can read it here: flytelog.substack.com/p/build-report…

The latest Flyte log is out: Build Reports Into Your Flyte Tasks and See Model Results Live! I might be the biggest "power user" when it comes to reports in flyte. I've embedded things most orchestrators probably couldn't handle, like robotic simulation rollout playbacks and generated videos. Of course I also use it for all my training metrics and evals too (it can up update live). Being able to embed basically any HTML lets you to use interactive plotting like Plotly, etc, and its all embedded on that specific run so you maintain complete lineage. I LOVE reports. If you're building on Flyte and not using them, you should check them out!

If you run a group on meetup you should turn off discussions. I've seen nothing but spam in them for years and surprised that it's still a feature.

All of those blacksmith's hands are gonna hurt 😅

Moving on from images we're looking at generating videos this week with open source models! These are all prompt to video, but I'll be looking to try and squeeze in some image to video and video to video by Fridays stream Some of these are older ones but really fun to compare with the quality! Any open source AI video generation models I should checkout this week?

I want another dgx spark to experiment on while the other one is doing a longer runs 😅

Open-Source Image Generation: Diffusion Models & LoRA - AI Build & Learn x.com/i/broadcasts/1…

I've been playing around with open source image generation models for this Friday's stream. I created a list of prompts to evaluate how they handle different types of situations. I'm using Flyte Devbox on a NVIDIA AI DGX Spark to run this all locally, The results get embedded directly in a Flyte report for quick scanning and every run is versioned which is great for coming back to explore prompt and generations. In this run I tried the models: - Flux1 schnell - SDXL - NVIDIA's Sana sprint - Qwen Image 2512 - zimage turbo - chroma-hd - Chroma Flash I want to try getting Flux2, SD35 large, and Krea2 going as well (but may not have time before friday). Any other open source models that have come up you think are worth trying out?

I finally built a cool image generation comparison pipeline with Flyte debox on my DGX spark What are the coolest open source image gen models, or ones I should check out?

See some examplere here or check out the docs! flytelog.substack.com/p/build-contai…

This week on Flyte log: Build container images in pure Python Writing Dockerfiles for ML projects can be tedious. You need to get the base image right, install system packages, install your dependencies, copy source files, and inevitably debug layer caching issues. What if you could do all of that in Python, right next to your task code? Flyte 2’s flyte.Image lets you define container images with a fluent builder API. No Dockerfile. No context switching. Just chain methods and Flyte handles the rest. See the code snippets in the latest flyte log or check out the docs

Computer break this weekend (Little si)

@rasbt @ManningBooks Would love to have you if you're available, and no pressure to stay the whole time.

@sagecodes @ManningBooks Wow awesome! Happy to stop by (virtually) as a guest 😊

@rasbt @ManningBooks RSVP on Luma: luma.com/j0l3hy2z

August's book for AI book club is "Build a Reasoning Model (From Scratch)" by @rasbt You can order both digital or physical copies now from @ManningBooks Building an LLM from Scratch was a group favorite, and I'm excited to for the "sequel"!