Nice article from the Harvard Crimson about our 'CBS-NTT Program in Physics of Intelligence'! 🧠

“This is new for all of us. How do you explain intelligent behavior in equations or in physics terms?”

I'm co-organizing the inaugural research workshop on Compound AI Systems on June 13th: sites.google.com/view/compound-… . Send in your work on designing & optimizing such systems!

Thrilled to have Richard Socher, Monica Lam and Noam Brown as speakers, and host this at #DataAISummit.

🦙🦙🦙🔥

We're thankful to Databricks for the great training experience we had with them for OLMo 1.7! Cheers to supporters of open science 🥳

The next set of OLMo models is here, with HUGE improvements in performance!

Announcing our latest addition to the OLMo family, OLMo 1.7!🎉Our team's efforts to improve data quality, training procedures and model architecture have led to a leap in performance. See how OLMo 1.7 stacks up against its peers and peek into the technical details on the blog:

Our team is incredibly proud to partner with Allen Institute for AI and thrilled to see them cook! Achieving such a massive improvement in MMLU, while reducing the compute budget, is a fantastic win. And doing it fully open? Everyone wins. Congrats! Can't wait to see what's next 👀

We released OLMo 1.7 7B + Dolma 1.7 today 🔥
With the juiciness of Dolma 1.7 + staged training we have improved OLMo’s MMLU score by 24 pts, clearly better than Llama2 7B!
Blog post: blog.allenai.org/olmo-1-7-7b-a-…
Model: huggingface.co/allenai/OLMo-1…
Dataset: huggingface.co/datasets/allen…

c'mon! u'd better take this! 🤣

Fixed the fix.

Fixed it for you, Cody Blakeney

Friday is LLM battle day.

I added DBRX to the financial metrics challenge.

Overall, very impressed with DBRX.

Main takeaways:
• correctly calculated metrics
• ranked top 4 fastest models
• competitive pricing

DBRX was +50% cheaper and +100% faster than models in its tier.

So crazy the things that emerge at scale. It’s so satisfying releasing models I got the wild and watching people do such cool stuff with them. Makes it all worth it ❤️

Databricks DBRX Instruct is twice as fast as new Mixtral 8x22B. When asked to write a reverse proxy in python both produce comparable quality results.

Latest AI showdown 🤜🏼🤛🏼

1. Mistral AI 8x22B vs Databricks DBRX Instruct
2. Google AI Gemini 1.5 Pro vs new OpenAI GPT-4-Turbo vision.

Details 👇

HumanEval benchmark results for the new Mixtral 8x22B, DBRX-base, Qwen1.5-72B, and Mixtral 8x7B:

Grateful to Ameet Talwalkar for the chance to present my recent work at CMU today! There are very exciting things happening in industry these days.

🚨Open Source Drop🚨

Databricks is adopting MegaBlocks, and we're releasing the MegaBlocks integration into LLMFoundry. This is a critical component in our Dbrx training stack, and we're super excited to bring MoE training to the community (1/N)

Please welcome MegaBlocks to the Databricks family! databricks.com/blog/bringing-…