Repository navigation

GPU groundwork: standard-ops forecast graph and exact-order device linear kernels - #1

Draft

mudler-agent wants to merge 8 commits into

localai-org:mainfrom

mudler-agent:gpu/standard-ops

Draft

GPU groundwork: standard-ops forecast graph and exact-order device linear kernels#1
mudler-agent wants to merge 8 commits into
localai-org:mainfrom
mudler-agent:gpu/standard-ops

Commits on Oct 24, 2026
- Add a standard-ggml-op forecast graph beside the CPU custom ops
  
  mudler
  and
  claude
  committed
  7bd33c8
- Add NORM gates for the standard-op forecast graph
  
  mudler
  and
  claude
  committed
  25b0f8c
- Document the standard-op forecast graph
  
  mudler
  and
  claude
  committed
  3843c1f
- Split standard-op linear layers into reference-sized K panels
  
  mudler
  and
  claude
  committed
  0a4ca8d
- Build the standard-op RMS norm from the reference F32 cascade
  
  mudler
  and
  claude
  committed
  2c96f4e
Commits on Oct 25, 2026
- Add exact-order GPU linear kernel and bitwise test
  
  mudler
  and
  claude
  committed
  b5f9309
- Add an exact-order F32 linear test for Metal
  
  mudler
  and
  claude
  committed
  9efae51
- Share one input generator between the CUDA and Metal tests
  
  mudler
  and
  claude
  committed
  e4fe365