Code

backend : add eval callback#4935

Code

Merged

ggerganov merged 9 commits intomasterggml-org/llama.cpp:masterfrom gg/sched-eval-callback-4931ggml-org/llama.cpp:gg/sched-eval-callback-4931Copy head branch name to clipboardJan 17, 2024

+64-2

Commits

Commits on Jan 15, 2024

• backend : add eval callback

  Show description for 65648b3
  ggerganov

  committed
  65648b3

  Copy full SHA for 65648b3

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

  ggerganov

  committed
  01b6f68

  Copy full SHA for 01b6f68

• backend : clean-up the implementation

  Show description for 83f3d7a
  ggerganov

  committed
  83f3d7a

  Copy full SHA for 83f3d7a

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

  ggerganov

  committed
  e1b1db9

  Copy full SHA for e1b1db9

• simple : fix

  ggerganov

  committed
  e049380

  Copy full SHA for e049380

Commits on Jan 16, 2024

• simple : no need for ggml_is_contiguous + fix bool parse

  ggerganov

  committed
  aa16b54

  Copy full SHA for aa16b54

• llama : fix callback placement in llama_context_params

  ggerganov

  committed
  0c96c72

  Copy full SHA for 0c96c72

• backend : avoid double-ask callback calls

  ggerganov

  committed
  012ecec

  Copy full SHA for 012ecec

Commits on Jan 17, 2024

• simple : restore examples, imatrix will serve as a demo

  ggerganov

  committed
  200dcaf

  Copy full SHA for 200dcaf