Code

fix lora issues#9

Code

Merged

ngxson merged 2 commits intongxson:xsn/fix_lorangxson/llama.cpp:xsn/fix_lorafrom ggerganov:sl/fix_fix_loraggerganov/llama.cpp:sl/fix_fix_loraCopy head branch name to clipboardJul 10, 2024

+37-49

Commits

Commits on Jul 10, 2024

• cuda : do not use dmmv if the tensor does not have enough cols

  slaren

  committed
  f15167a

  Copy full SHA for f15167a

• llama : lora fixes

  slaren

  committed
  9841fbd

  Copy full SHA for 9841fbd