If you are having a hard time visualizing all the layers and matrix operations inside an LLM, then you'll love this website! Check it out: bbycroft.net/llm

Run 20+ models on your data with a few clicks in Oxen.ai 🪄 Here's the full list 👇 oxen.ai/explore/models

POV you want to make sure the next iteration of LLM training data is high quality when they crawl your open source repository.

We are diving into "Upcycling LLMs with Mixture of Experts" tomorrow with Ethan He from NVIDIA on Arxiv Dives. Feel free to join us live at 10am PST to learn from the some of best and the brightest in the field. lu.ma/arxivdive-31

For all my Rustaceans, this is from our open source project today 🤦‍♂️ 🦀 Who knew the println! macro could be one of the most dangerous functions when working with enums

If it’s not highly technical, I send it back. See you all there 🤓

Fridays are for increasing our rate of learning as well as increasing model learning rates

Last night was one year of oxen sponsoring AI Tinkerers in LA. We always buy the egg rolls and Sean McDonald made us the most sweet (and sour) gift of all time 🥹 3D printed Oxen.ai Egg Rolls. Thanks my dude, going front and center on my desk.

G'mooooonday ☕️

Then collaborate on them in Oxen.ai

Save your chats, upload the jsonl to Oxen.ai, view, query, edit, run models, the world is yours.

Talking Upcycling LLMs to MoE with Ethan He from Nvidia, and learned a lot. It's a pretty simple approach that costs 1/8 the compute to get performance 2x the model size (according to scaling laws). Highly recommend listening to the whole presentation.

Tomorrow at 10amPT we're back with AI Water Cooler. Kirk Greninger will be joining us to talk about his magazine wtllm_mag “What the LLM?” a physical AI magazine with a focus on LLMs. Greg Schoeninger will show experiments with Llama 3.2-1B to try to re-create o1-mini at home!

This

It's almost like we could use a better way of sharing and collaborating on datasets 🤔 🐂 oxen.ai/datasets/mnist…

Then version your evals in Oxen.ai so you can always go back to them and compare over time

Oxen.ai can now see! 🖼️👀 Simply select a dataset with images, type in your prompt, and let GPT-4o or GPT-4o-mini do the work for you. Let us know if there are other vision models you'd like to see us integrate.

Posting on Reddit is awesome because you get random experts telling you about models and techniques you’ve never heard of before

Oxen.ai now supports Groq Inc for Visual LLMs including Llama 3.2 Vision 90B, 11B and LLaVA 7B Can confirm, it's fast AF 🔥 First test processed 125 receipts in 24 seconds for $0.02