Code

ggml-cuda : support stablelm rope#4156

Code

Merged

slaren merged 4 commits intomasterggerganov/llama.cpp:masterfrom cuda-stablelm-ropeggerganov/llama.cpp:cuda-stablelm-ropeCopy head branch name to clipboardNov 24, 2023

+24-17

Commits

Commits on Nov 21, 2023

• ggml-cuda : support stablelm rope

  slaren

  committed
  5844493

• remove unused freq_base kernel parameter

  slaren

  committed
  4a3469f

• add n_dims parameter to llm_build_k_shift, default to n_rot via overload

  slaren

  committed
  84adb54

Commits on Nov 24, 2023

• llama : fix llm_build_k_shift args

  ggerganov

  committed
  5ed3e1a