Repository navigation

llama : attention split on main, for the KV cache in VRAM - #15

Merged

Piggidragon merged 5 commits into

mainfrom

mgpu/attn-split-v2

Oct 16, 2026

Merged

llama : attention split on main, for the KV cache in VRAM#15
Piggidragon merged 5 commits into
mainfrom
mgpu/attn-split-v2

Commits on Oct 16, 2026
- llama : add an attention split separate from the tensor split
  Piggidragon
  committed
  fc9df90
- llama : validate the attention split shares
  Piggidragon
  committed
  51f956c
- llama : pick the attention split by what the tensor is measured against
  Piggidragon
  committed
  f980340
- llama : rebase the attention split on main, describe its rounding and log the result
  
  Piggidragon
  and
  claude
  committed
  495df37
- llama : fix the review of the attention split
  
  Piggidragon
  and
  claude
  committed
  ff1a44e