A Few Commentary-Free Resources & A Note About An Unscheduled Temporary Drop Hiatus

Greetings and felicitations on this somewhat fine Thursday!

Due to a very distressing and also very unexpected octogenarian critical family emergency on my spouse’s side of the family, I’ll be in transit most of Thursday, then navigating 🏥 corridors both Thursday and Friday, followed by a soul-crusing large amount of transit this weekend. The time I would have waxed poetic for Thursday’s resource Drop, Friday’s WPE, and this weekend’s Bonus Drop is being spent preparing for said adventure and communicating with extended fam. Oh, and likely collapsing for a while on Sunday.

For your AI-consideration, I present:

which is a very intriguing take on something we were/are all doing manually before ~February this year, and are now likely (eventually) going to relegate to AI and some clever meta-programming. However, they are not mutually exclusive beasties.

The ☂️ for two other AI-related resource drops is “vector databases” — a.k.a. “embedding databases” / “vector similarity search engines” — which are all the rage this year (for fairly obvious reasons).

Here’s what they are, and here’s a pretty hot-but-accurate take on them.

Finally, these fancy new tools aren’t of any use if we can’t use them in a practical sense. I’ll let this last resource speak for itself:

LLMs promise to fundamentally change how we use AI across all industries. However, actually serving these models is challenging and can be surprisingly slow even on expensive hardware. Today we are excited to introduce vLLM, an open-source library for fast LLM inference and serving. vLLM utilizes PagedAttention, our new attention algorithm that effectively manages attention keys and values. vLLM equipped with PagedAttention redefines the new state of the art in LLM serving: it delivers up to 24x higher throughput than HuggingFace Transformers, without requiring any model architecture changes.

FIN

Barring degradation in the fam health situation, Drops will resume Monday.