lubitelpospat

Impressive work! I wonder if there's any plan for refining the classification based on user feedback? I.e. Martin Klepman's the wild boar book is currently being shown in "Business & Management" category, and "Pattern Recognition and Machine Learning" is in "Psychology and Behavior". By the way, is there some book naming normalization based on learned context embeddings (i.e., will this mention of the wild boar book be counted)?

Q8_0 on my laptop, hope LiquidAI team can give some feedback on what model they offer in the playground!

Surprisingly (or maybe not?), LFM3-VL-2B gave me better results than LFM3-1.7B on text-only prompts (general knowledge, chatting) through liquid playground. As an example, when asked “what does a cow say?”, LFM3-1.7B said that a cow can say “Oink”, and when I asked the same question to LFM3-VL-3B, it (correctly) said “moo”. I assume this is lower quant problem, though: Q8 versions I’ve played with through llama both give the correct “moo” answer.

Inverse Cramer never fails

Recently tried to initialize one Mac from another - kids, never ever attempt that! Microsoft stuff doesn't get copied over properly, and refuses to behave even if you delete an account and create a new one - the only solution is re-installing the OS from scratch. And if you want to do so, and you erase the disk - surprise surprise, in recovery mode macOS doesn't work with the WPA2/WPA3 Enterprise network; you need to go and find a WPA2 network somewhere to re-install the OS. Does the recovery menu explain that somewhere? Hell no. Damn, installing a linux distro from a usb stick is a better experience than re-installing macOS these days!

Great work! Will be curious to see how it competes with the existing workflow languages such as Nextflow or Snakemake - do you by any chance have a "Why migrate from Nextflow/Snakemake to Rex" pitch I could read somewhere?

I couldn’t find an option to connect to a generic OpenAI-compatible endpoint - did I miss it?

Nice thanks

What do I read to quickly get a grasp of how economics works, from a high level perspective?

Nice project! Would be nice to have an option without a Rapsberry Pi 5 though - I think a decent fraction of your audience already has one and maybe doesn’t want to pay for another

If you're using litert-lm on a Mac with Apple Silicon - DO NOT forget to use "--backend gpu"! On my M1 Pro laptop this single setting resulted in 10x prefill performance and 2x decode performance. To anyone who knows how the internals of litert-lm work - what quantization does it use? How come the model is just 3.4 GB in size?

EDIT: typo fix.

Wow, that's quite beefy.

All right guys, this is your time - what consumer device do you use for local LLM inference? GPU poor answers only