Commits

Branch selector

 ecd0c1c

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 27, 2024

• [PERFORMANCE] Fix Packing thread regression in code (AutoGPTQ#642)

  Qubitium

  authored
  ecd0c1c

  Copy full SHA for ecd0c1c

Commits on Jun 24, 2024

• Supporting uint4 inference of pre-quantized models in HPU (AutoGPTQ#689)

  Show description for b57bea0
  

  HolyFalafel

  and

  Yantom1

  authored
  b57bea0

  Copy full SHA for b57bea0

Commits on Apr 16, 2024

• stream pytest output

  Qubitium

  committed
  991508a

  Copy full SHA for 991508a

• fix pack() thread regression via code

  Qubitium

  committed
  35cfceb

  Copy full SHA for 35cfceb

Commits on Apr 12, 2024

• Force warn users to reduce threads or packing may regress ~100x (AutoGPTQ#628)

  Show description for ea829c7
  Qubitium

  authored
  ea829c7

  Copy full SHA for ea829c7

• Cohere Command-R Added (AutoGPTQ#631)

  Show description for ce39d61
  

  egortolmachev

  and

  t3ga

  authored
  ce39d61

  Copy full SHA for ce39d61

Commits on Apr 5, 2024

• Fix padding not used correctly in exllama v2 layer (AutoGPTQ#626)

  Show description for b4b801c
  Qubitium

  authored
  b4b801c

  Copy full SHA for b4b801c

Commits on Mar 28, 2024

• Add checkpoint_format to unify all current/future saved weight formats (AutoGPTQ#603)

  Show description for 866b4c8
  

  Qubitium

  and

  fxmarty

  authored
  866b4c8

  Copy full SHA for 866b4c8

• Show meaningful error on Maxwell build (AutoGPTQ#622)

  Show description for ff3dcc4
  fxmarty

  authored
  ff3dcc4

  Copy full SHA for ff3dcc4

• Add --no-build-isolation to install instructions (AutoGPTQ#620)

  Show description for a78daec
  fxmarty

  authored
  a78daec

  Copy full SHA for a78daec

• Add support for MPT (#73)

  Show description for a9f3d46
  LaaZa

  authored
  a9f3d46

  Copy full SHA for a9f3d46

• fix tritonv2 in make_sure_no_tensor_in_meta_device (AutoGPTQ#617)

  LaaZa

  authored
  ca187fb

  Copy full SHA for ca187fb

Commits on Mar 26, 2024

• Patch load_checkpoint_in_model till accelerate 0.29 (AutoGPTQ#613)

  Show description for 9fd7a6c
  fxmarty

  authored
  9fd7a6c

  Copy full SHA for 9fd7a6c

• fix quantization samples

  fxmarty

  committed
  24efa31

  Copy full SHA for 24efa31

Commits on Mar 25, 2024

• Fix: bias option for qlinear should depend on origin layer (AutoGPTQ#602)

  Show description for 82070a1
  Coco58323

  authored
  82070a1

  Copy full SHA for 82070a1

• Fix quantization compatibility with transformers>=4.39 (AutoGPTQ#607)

  Show description for 65f3cf3
  fxmarty

  authored
  65f3cf3

  Copy full SHA for 65f3cf3

• Pin transformers to <4.39.0 Fix Issue AutoGPTQ#604 (AutoGPTQ#605)

  Show description for 1ba1e76
  Qubitium

  authored
  1ba1e76

  Copy full SHA for 1ba1e76

Commits on Mar 19, 2024

• Add MANIFEST.in (AutoGPTQ#600)

  Show description for ea4a997
  fxmarty

  authored
  ea4a997

  Copy full SHA for ea4a997

• Tritonv2 lint fix (AutoGPTQ#596)

  Show description for 56e4d88
  

  jeromeku

  and

  fxmarty

  authored
  56e4d88

  Copy full SHA for 56e4d88

• Add support for Phi models (AutoGPTQ#576)

  Show description for 1c8850d
  mgoin

  authored
  1c8850d

  Copy full SHA for 1c8850d

• Removing Redundant 'scales' Assignment (AutoGPTQ#585)

  Show description for f482ce6
  Danielohayon

  authored
  f482ce6

  Copy full SHA for f482ce6

• Fix marlin quantization using BaseQuantizeConfig from _pretrained AutoGPTQ#581 (AutoGPTQ#586)

  Show description for 04bca0d
  

  Qubitium

  and

  fxmarty

  authored
  04bca0d

  Copy full SHA for 04bca0d

Commits on Mar 18, 2024

• Add starcoder2 support (AutoGPTQ#578)

  Show description for 09289d8
  TechxGenus

  authored
  09289d8

  Copy full SHA for 09289d8

• Fix spelling mistakes. (AutoGPTQ#577)

  Show description for 80b0571
  duchengyao

  authored
  80b0571

  Copy full SHA for 80b0571

Commits on Mar 1, 2024

• Update install instructions on 0.7.1 release (AutoGPTQ#574)

  Show description for 896d820
  fxmarty

  authored
  896d820

  Copy full SHA for 896d820

• Compatibility with latest transformers (AutoGPTQ#573)

  Show description for d30f660
  fxmarty

  authored
  d30f660

  Copy full SHA for d30f660

Commits on Feb 26, 2024

• Explicitely check compute capability in marlin's QLinear (AutoGPTQ#567)

  Show description for a664836
  fxmarty

  authored
  a664836

  Copy full SHA for a664836

• Add support for Gemma models. (AutoGPTQ#561)

  LaaZa

  authored
  6b55300

  Copy full SHA for 6b55300

Commits on Feb 19, 2024

• Support loading sharded quantized checkpoints. (AutoGPTQ#425)

  Show description for 6906ce8
  

  LaaZa

  and

  fxmarty

  authored
  6906ce8

  Copy full SHA for 6906ce8

• Fix shape check marlin (AutoGPTQ#557)

  Show description for 33e8bd9
  fxmarty

  authored
  33e8bd9

  Copy full SHA for 33e8bd9

• Fix bias materialization for Marlin (AutoGPTQ#554)

  Show description for 05a7a2c
  fxmarty

  authored
  05a7a2c

  Copy full SHA for 05a7a2c

• Add back missing import (AutoGPTQ#553)

  Show description for 10cd170
  fxmarty

  authored
  10cd170

  Copy full SHA for 10cd170

Commits on Feb 16, 2024

• update version

  fxmarty

  committed
  74212f5

  Copy full SHA for 74212f5

• Add more guards on compute capability in Marlin kernel (AutoGPTQ#550)

  Show description for b05e059
  fxmarty

  authored
  b05e059

  Copy full SHA for b05e059

• fix typos in workflows

  fxmarty

  committed
  c1a1ef3

  Copy full SHA for c1a1ef3

Pagination

Previous

Next