a small trailer of my latest DSPy exploration...
full video on lycee.ai

2012: A University of Toronto team created AlexNet. It showed the power of DNNs, outperforming ImageNet's runner up by 11 pts.

2024: Bo Wang Lab at University of Toronto built & optimized several DSPy programs, beating the next best MEDIQA system by *20 pts*!

The AlexNet moment for optimized LM programs?🧵

What if I write a script to visualize DSPy modules as graphs ?? Should I ?

And here is the preview of what I am building to visualize DSPy modules

Prompt formatting in #DSPy

here are three different ways to create prompts (ok signatures) in DSPy and control exactly what comes in and out.

First is the shorthand.
Second is class based.
Third is fully customized.

Use the third to work with multi-line inputs and outputs.

Omar Khattab dropping gems on natural language modules and DSPy!

Users from Lycee AI are asking me to create more advanced DSPy tutorials. I'm starting a new series today with an exploration of DSPy and knowledge graphs.

I wrote a script to build a knowledge graph using DSPy. I'll be dropping the video on Lycee AI later today. It's actually…

Amazing talk by Omar Khattab on his “pytorch for Language Programming” library called DSPy.

Have been using DSPy a lot lately and it’s absolutely fantastic to code language models with it.

Check it out if you work with language models.

Thanks for hosting Arize AI

We've aggregated all of the DSPy resources from the Weaviate team on one page! It is broken into two categories: 1. Hands on Learning and 2. Read and Listen 📚

DSPy round-up: weaviate.io/developers/wea…

We're also very excited to host our in-person meetup with DSPy, Arize AI, and…

It’s happening, over 500+ signups? A packed venue at Edge, amazing partnerships with Arize AI cohere and of course Weaviate • vector database. Excited to learn more about DSPy from Omar Khattab!

I have finally completed the tutorial on using Spacy-llm and DSPy for Named Entity Recognition. It concludes with an unexpected twist. Link in the following tweet. Preview:

1/13 A good prompt nowadays can be 6000+ characters. That's like a 3 page essay to get a world class prompt.

DSPY can help you write prompts and run them better.

Here are my notes from our recent paper club meeting

arxiv.org/abs/2310.03714

Added two new chapters to the advanced DSPy tutorial 🔥

In the second chapter, I explain how to improve the knowledge graph extractor we built in the first chapter, aiming to reduce duplicate entities.

In the third chapter, I explain what coreference resolution is and why it is…

I cannot emphasize how much of a game changer DSPy is. It’s prompt engineering for the first time.

The DSPy meetup is so on!
Edge

I know that -instruct is not a chat model, but is doing dspy something that the answer is that negative? 😂

Awesome event tonight, and an extremely engaging presentation of DSPY by Omar Khattab

— may I say, LFGomar? 😂

arize-phoenix Weaviate • vector database

Most people don't know that I studied HISTORY in undergrad.
A long way from reading cutting edge AI systems.
If you're just getting started, stick with it.
'It' being whatever interests you.
Your life won't be linear.

Like all the code I'm tracing to understand DSPy Agents 🤣

From unstructured to structured data using DSPy

Full tutorial on Lycee AI (link in second tweet)

How DSPy optimizers work — a Q&A part of the talk not recorded by the venue!