Code

imatrix : offload to GPU support#4957

Code

Merged

ggerganov merged 10 commits intomasterggerganov/llama.cpp:masterfrom gg/imatrix-gpu-4931ggerganov/llama.cpp:gg/imatrix-gpu-4931Copy head branch name to clipboardJan 17, 2024

+128-53

Commits

Commits on Jan 15, 2024

• backend : add eval callback

  Show description for 65648b3
  ggerganov

  committed
  65648b3

• backend : group nodes in a single compute when user don't need them

  ggerganov

  committed
  01b6f68

• backend : clean-up the implementation

  Show description for 83f3d7a
  ggerganov

  committed
  83f3d7a

• simple : do not perform tensor data copy if not needed

  ggerganov

  committed
  e1b1db9

• simple : fix

  ggerganov

  committed
  e049380

• imatrix : offload to GPU support

  ggerganov

  committed
  0b2fca9

Commits on Jan 17, 2024

• imatrix : fix ggml_mul_mat_id hanlding

  Show description for a722d05
  ggerganov

  committed
  a722d05

• ci : add imatrix test

  Show description for 10b25e0
  ggerganov

  committed
  10b25e0

• ci : rearrange output

  Show description for 4fb5284
  ggerganov

  committed
  4fb5284

• Merge branch 'master' into gg/imatrix-gpu-4931

  Show description for 2917e6b
  ggerganov

  committed
  2917e6b