🥳Check out: Token-Explorer! 🤖.Interact with and explore LLM token generation!.Features:.- Step through token selection.- Remove tokens to explore alt paths.- Fork prompt and quickly switch between them.- Visualize all token probabilities and entropy!.- OSS (github in replies)

RT @BEBischof: Hell yeah, blog came out about my failure funnel!. Come to @DataCouncilAI to see me explain it in depth!. .

Generate videos in just a few seconds. Try Grok Imagine, free for a limited time.

There are many things not ideal for with our current timeline… but we do have juice box sake!

Support for Structured Outputs in Token Explorer should be coming soon!

RT @venturetwins: A summary of consumer AI

Had an absolute blast chatting joining @cameron_pfiffer to chat with @CShorten30 about all the cool stuff we're working on @dottxtai!.

RT @cameron_pfiffer: Does anyone use exllamav2 with Outlines? If so, what has your experience been?.

RT @cameron_pfiffer: If you liked this course, you should come tell us in our Discord:

So excited to share what @cameron_pfiffer and I have been up to @dottxtai!. It was absolutely fantastic to work with Andrew and the incredible team @DeepLearningAI . We hope this course makes it easier to get started getting reliable outputs from your LLM using structured gen!.

RT @cameron_pfiffer: Who are the best public speakers/presenters in AI? . Looking for people to give interesting talks for AI by the Bay.….

RT @colin_fraser: tremendous alpha right now in sending your wife photos of yall converted to ren and stimpy characters .

It’s always odd to me when people gush about OpenAI releasing image models that can only just now do things open models have been doing about a year now!.

RT @RohanInference: I was interested in learning about Bayesian statistics and came across this fantastic book written by @willkurt ! Highl….

Choosing structure goes hand in hand with prompt engineering! A quick demo, here are the steps:. - Clone the initial prompt 5 times.- See what the unstructured output is.- Then generate through 4 different date formats and see how they compare!. We can see the model tends to

TokenExplorer now has basic support for structured generation implemented!. At any point in the prompt you can toggle on predefined structure (note with the 'Struct' label turns green) and it will constrain the output using @dottxtai's outlines-core!. Not quite ready for a PR but

I must say we do some cool things internally!.

It's funny how often you hear "you can't tell how uncertain an LLM is", but here is a case where you can investigate exactly what the model was thinking, and how close it was to getting the correct answer!.

RT @alonsosilva: TokenExplorer: 9.9 or 9.11, which one is bigger?

I think manually sampling high-entropy prompts is my new favorite way to write! . I had a short prompt talking about Burroughs cut-up technique and took the model from there! Most fun I've had with an LLM in a while!

It's fun to try to keep the prompt in a state of high entropy!

Here's a link to the repo: And some images of the "probability" and "entropy" visualization modes!