Code

[WIP] Quark Quantizer Support#1207

Code

Open

shobrienDMA wants to merge 18 commits intomicrosoft:mainmicrosoft/onnxruntime-genai:mainfrom shobrienDMA:shobrien/per_layer_supportshobrienDMA/onnxruntime-genai:shobrien/per_layer_supportCopy head branch name to clipboard

+257-72

Commits

Commits on Dec 8, 2024

• get a quark quantized model through OGA without config modifications

  shobrienDMA

  committed
  16e23b7

  Copy full SHA for 16e23b7

Commits on Dec 10, 2024

• remove potentially errant lm_head.g_idx = 0

  shobrienDMA

  committed
  4a458cf

  Copy full SHA for 4a458cf

• fix lm_head layers not getting the right group size

  shobrienDMA

  committed
  9cd1b85

  Copy full SHA for 9cd1b85

Commits on Dec 17, 2024

• refactor quantized_model.py

  shobrienDMA

  committed
  c1d6341

  Copy full SHA for c1d6341

• Merge branch 'amd/main' into amd_shobrien/per_layer_support

  Show description for ae05635
  shobrienDMA

  committed
  ae05635

  Copy full SHA for ae05635

• remove GPTQ extra argument

  shobrienDMA

  committed
  c661471

  Copy full SHA for c661471

• use the quark model class

  shobrienDMA

  committed
  2e8e13e

  Copy full SHA for 2e8e13e

Commits on Dec 18, 2024

• tidy formatting, remove unecessary print and comments

  shobrienDMA

  committed
  67b5c09

  Copy full SHA for 67b5c09

Commits on Dec 19, 2024

• tidy up comments and unused variables

  shobrienDMA

  committed
  a5891e7

  Copy full SHA for a5891e7

• Update the quark model class

  shobrienDMA

  committed
  38ddbf2

  Copy full SHA for 38ddbf2

• remove the make quantized lm_head method since the existing method works for this

  shobrienDMA

  committed
  229b30f

  Copy full SHA for 229b30f

Commits on Jan 13, 2025

• Merge branch 'main' of https://github.com/microsoft/onnxruntime-genai into amd_shobrien/per_layer_support

  shobrienDMA

  committed
  06d1f27

  Copy full SHA for 06d1f27

• update the ONNX Runtime Extensions dependency version

  shobrienDMA

  committed
  3c39016

  Copy full SHA for 3c39016

Commits on Jan 29, 2025

• Merge branch 'oga/main' into amd/shobrien/per_layer_support

  shobrienDMA

  committed
  0e4b0d0

  Copy full SHA for 0e4b0d0

Commits on Feb 10, 2025

• Merge branch 'microsoft:main' into shobrien/per_layer_support

  shobrienDMA

  authored
  e33446a

  Copy full SHA for e33446a

Commits on Feb 25, 2025

• Extra layernorms

  BowenBao

  committed
  4c9c545

  Copy full SHA for 4c9c545

• Address comments; Update local_bits, local_group_size passing

  BowenBao

  committed
  09dafca

  Copy full SHA for 09dafca

• drop cmake/deps.txt changes

  BowenBao

  committed
  d1b212f

  Copy full SHA for d1b212f