Code

0cc4m/vulkan iq4 nl#253

Code

Merged

Nexesenex merged 10 commits intoNexesenex:lcpp_pr_vulk_iq4nl_supportNexesenex/croco.cpp:lcpp_pr_vulk_iq4nl_supportfrom ggerganov:0cc4m/vulkan-iq4_nlggerganov/llama.cpp:0cc4m/vulkan-iq4_nlCopy head branch name to clipboardJul 21, 2024

+597-702

Commits

Commits on Jul 19, 2024

• convert-*.py: add general.name kv override (#8571)

  mofosyne

  authored
  3d0e436

  Copy full SHA for 3d0e436

• fix: typo of chatglm4 chat tmpl (#8586)

  Show description for f299aa9
  thxCode

  authored
  f299aa9

  Copy full SHA for f299aa9

• ggml : add friendlier error message to fopen errors (#8575)

  Show description for b57eb9c
  HanClinto

  authored
  b57eb9c

  Copy full SHA for b57eb9c

• readme : fix server badge

  ggerganov

  authored
  be0cfb4

  Copy full SHA for be0cfb4

• llama : bump max layers from 256 to 512 (#8530)

  Show description for d197545
  ggerganov

  authored
  d197545

  Copy full SHA for d197545

• convert-*.py: remove add_name from ChatGLMModel class (#8590)

  mofosyne

  authored
  57b1d4f

  Copy full SHA for 57b1d4f

• ggml : fix quant dot product with odd number of blocks (#8549)

  Show description for 87e397d
  

  slaren

  and

  ggerganov

  authored
  87e397d

  Copy full SHA for 87e397d

Commits on Jul 20, 2024

• Fix Vulkan matmul tests compile errors

  0cc4m

  committed
  c8ee1bc

  Copy full SHA for c8ee1bc

Commits on Jul 21, 2024

• Add Vulkan IQ4_NL support

  0cc4m

  committed
  6274b3f

  Copy full SHA for 6274b3f

• Fix Vulkan DeepSeek-Coder-V2-Lite MoE support

  0cc4m

  committed
  3252afb

  Copy full SHA for 3252afb