Code

ggerganov master to upstreamchanges#2

Code

Closed

YellowRoseCx wants to merge 7 commits intoYellowRoseCx:upstreamchangesYellowRoseCx/koboldcpp-rocm:upstreamchangesfrom ggml-org:masterggml-org/llama.cpp:masterCopy head branch name to clipboard

+558-387

Commits

Commits on Apr 15, 2023

• cmake : add finding the OpenBLAS header file (#992)

  katsu560

  authored
  106faaf

  Copy full SHA for 106faaf

• benchmark : fix result validation in benchmark-q4_0-matmult (#987)

  dfyz

  authored
  c12b14b

  Copy full SHA for c12b14b

• ggml : use posix_memalign on non-Windows env

  ggerganov

  committed
  aa485ce

  Copy full SHA for aa485ce

• ggml : add Q8_0 quantization for intermediate results (#951)

  Show description for e95b655
  

  ggerganov

  and

  sw

  authored
  e95b655

  Copy full SHA for e95b655

• Refactor ggml.c for future tensor types (#1001)

  sw

  authored
  0ad9646

  Copy full SHA for 0ad9646

• Fix potential int8 overflow in non-SIMD vec_dot (#986)

  sw

  authored
  2f7c8e0

  Copy full SHA for 2f7c8e0

• convert.py: Fix loading safetensors and ggml format on Windows (#991)

  Show description for 74f5899
  comex

  authored
  74f5899

  Copy full SHA for 74f5899