J S
J S
That is the name of the organization. It could be just GenAI for simplicity if it looks better? But the org is called JuliaGenAI.
I agree - I really like the right-hand side one! Would it be possible to actually change the letters to be the endpoints of those lines (with the same colours)?...
This is great! Thank you for bearing with us! I really like the first red one, because it’s the closest to the colors in the picture already. I wonder if...
It’s awesome! I love it ❤️
Not from my end and haven’t seen any on Slack either! The biggest question on my mind is how to make it into a sticker 🤣 What usually happens now?...
Thank you so much for your help!!!
Btw there is a typo in a kwarg name: https://github.com/thomasahle/dspy/blob/719a8d1805b4a59780ed9cda6840e782be39c013/dspy/teleprompt/bootstrap.py#L99 “bootsraps” vs bootstraps But I intentionally didn’t fix it, because that could break code and would need to be versioned...
> Ah thanks for the note. Actually please do feel free to fix `max_bootsraps` too. It's not used anywhere based on quick search (and I'm pretty sure it's not used...
For reference, I did some benchmarking few weeks ago and Bool embeddings performed really well (especially assuming that we would use a reranker/cross-encoder downstream With `top_k=20`, 1024dims in Bool still...
I'm actually slightly hesitant to enforce threading under the hood, because: - if we have composite indices (eg, 5 indices with different types of data), it might be more natural...