Feature Request: Support for Poolside's new Laguna line up

[dandandelion]

Prerequisites

• I am running the latest code. Mention the version if possible as well.
• I carefully followed the README.md.
• I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• I reviewed the Discussions, and have a new and useful enhancement to share.

Feature Description

Support for the newer Laguna M 1/S 2.1/XS 2.1

PoolSide - HuggingFace

Motivation

Being one of the only few to support kvarn and other extreme kv cache compression, it might serve well or make the model usable for even more devices that otherwise might struggle to fit the model on their devices.

Possible Implementation

I have no clue

[crusaderky]

I have a franken-fork right here: https://github.com/crusaderky/llama.cpp/tree/beellama-laguna
Note: I didn't test DFlash.

Note: the laguna fork contains a commit that conflicts with ggml-org#25165
I did not investigate if they are identical or not.
For the sake of long term maintainability, I'd suggest to merge from llaamcpp main first, and then cherry-pick the commits from the laguna fork.

[Anbeeld]

Merged in from upstream in v0.4.2 branch.

[crusaderky]

Merged in from upstream in v0.4.2 branch.

Not quite - upstream supports laguna xs and m. You still need the laguna patches for S. There are also two dflash commits.

Broad question: what is the scope of this fork? Should it cover models that are not yet in mainline, or should it stay focused on core functionality?