News for hackers. Stuff that matters.

foldl2022

Gemma-4 E2B/E4B models reuses K-V cache from other layers, which do things in a "transposed" way: not reuse Q/K/V matrices within a single layer, but reuse across different layers.
Re:Meta Superintelligence's surprising first paper on Saturday October 11, 2025 @09:26PM (#45554299)
So, show me the model weights, please.
Re:Titania Programming Language on Sunday September 14, 2025 @11:18PM (#45245750)
A nice side-effect of "variables at the top": you keep your functions short.
Re:Lazarus Release 4.0 on Monday May 12, 2025 @02:27AM (#43960185)
Lazarus is still my favorite when developing desktop apps.

Language is not a problem. Pascal is just C/C++ in another favor. LCL/VCL is a wonderful library, everything just works like what I am expecting.

Re:Intel appoints Lip-Bu Tan as its CEO on Wednesday March 12, 2025 @07:18PM (#43348741)
Tan looks like Pat. Amazing.
Re:Performance of LLMs on Advent of Code 2024 on Monday December 30, 2024 @08:25PM (#42555461)
Just open-wights it. We need this. :)
Re:I Run LLMs Locally on Sunday December 29, 2024 @08:32PM (#42545478)
Off the topic: I am also in favor of minimalist.

In the case of AlphaGeometry, I made AlphaGeometryRE to get rid of tensorflow/jax/flax, and 100+ Python packages.

Re:Show HN: AlphaGeometryRE AlphaGeometry Re-Engineered on Friday December 27, 2024 @09:39PM (#42528109)
Author here. It took quite a lot of time and efforts to implement this old model. The situation is worsened by `meliad` framework. Glad to see that `meliad` is not active now.
Re:Ping Rate Clock on Thursday November 07, 2024 @05:35AM (#42075459)
Is one ESP32 enough?