Code

IBM Granite MoE Architecture#9438

Code

Merged

ggerganov merged 10 commits intoggerganov:masterggerganov/llama.cpp:masterfrom gabe-l-hart:GraniteMoEgabe-l-hart/llama.cpp:GraniteMoECopy head branch name to clipboardSep 25, 2024

+88-13

Commits

Commits on Sep 24, 2024

• feat(gguf-py): Add granitemoe architecture

  Show description for 8a4ca23
  gabe-l-hart

  committed
  8a4ca23

• feat(convert_hf_to_gguf): Add GraniteMoeModel

  Show description for e0b7229
  gabe-l-hart

  committed
  e0b7229

• fix(granitemoe convert): Split the double-sized input layer into gate and up

  Show description for 014e59d
  gabe-l-hart

  committed
  014e59d

• feat(granitemoe): Implement granitemoe

  Show description for eca37cd
  gabe-l-hart

  committed
  eca37cd

• Typo fix in docstring

  Show description for 71bc4c1
  

  gabe-l-hart

  and

  ggerganov

  committed
  71bc4c1

• fix(conversion): Simplify tensor name mapping in conversion

  Show description for 5eb28c4
  gabe-l-hart

  committed
  5eb28c4

• fix(convert): Remove unused tensor name mappings

  Show description for f236099
  gabe-l-hart

  committed
  f236099

• fix(convert): Sanity check on merged FFN tensor sizes

  Show description for 317b15b
  gabe-l-hart

  committed
  317b15b

• fix: Allow "output" layer in granite moe architecture (convert and cpp)

  Show description for 1c8b3e4
  gabe-l-hart

  committed
  1c8b3e4

• fix(granite): Add missing 'output' tensor for Granite

  Show description for a843f1f
  gabe-l-hart

  committed
  a843f1f