ad1ttya

I pretty much got that idea when reading the post, on how it didnt have improvements on a Gemma 270M model which didn't have any looping issue. However, it still made me look at some bottlenecks in models I had trained (like MLP network for stylus input automation and a 17M reranker with NLI stance and noul heads) and allowed me to finetune without training, but just by adjusting some weights. But what's pretty cool is the zero shot sandbox development method they showed for a Go model. I tried to do that for a 17M re-ranker from zero weights and Astra built it using its reasoning without me having to use a GPU and find training datasets that are not CC by NC, and add some hours of tweaking and checkpoints. I don't know if its the best way to build something, but we might be building small models (maybe not LLMs) like with just programming the weights instead of finetuning.

i guess it's not needing to finetune for each specific use case

discourse is down for me.

From Ublock Origin's github(https://github.com/gorhill/uBlock): uBlock Origin is (and has always been) COMPLETELY UNRELATED to the web site ublock.org.

edit: and here is a commit blocking the site ublock.org - https://github.com/uBlockOrigin/uAssets/commit/82067d1ef3370...

This looks amazing! Just started studying intervallic improvisation and this will help alot! do you have any plans to add support for alternate tunings or let the user to modify the tuning (like this tool: http://fretviz.com/) ?

F#

He was complaining about meteor using node 0.10. This uses node 5.x

Maybe its because OpenTK has type safe bindings(OpenGL, OpenCL and OpenAL) for F# and C#.. http://www.opentk.com/

You install lodash globally, and can then use the above line to use the module inside repl..