Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on Sep 14, 2024

• Switch back to supported CI runners (#887)

  ZainRizvi

  authored
  8aa6533

  Copy full SHA for 8aa6533

• Move `autoround` from `generate.py` to `eval.py` (#868)

  Show description for d2bce6a
  yiliu30

  authored
  d2bce6a

  Copy full SHA for d2bce6a

Commits on Sep 13, 2024

• Fix inference_mode (#885)

  Show description for 90c8cbd
  jerryzh168

  authored
  90c8cbd

  Copy full SHA for 90c8cbd

• [Float8] Add static constructor that will be used in Observer workflow (#869)

  drisspg

  authored
  3fa38aa

  Copy full SHA for 3fa38aa

• readme-updates (#874)

  jcaip

  authored
  16b40fd

  Copy full SHA for 16b40fd

Commits on Sep 12, 2024

• Renaming fpx to floatx (#877)

  Show description for 8236a87
  jerryzh168

  authored
  8236a87

  Copy full SHA for 8236a87

Commits on Sep 11, 2024

• Fix tutorial example for my dtype tensor subclass (#865)

  Show description for f82071d
  jerryzh168

  authored
  f82071d

  Copy full SHA for f82071d

• [FSDP2] cast scale to float32 in precompute (#835)

  Show description for 85d03de
  weifengpy

  authored
  85d03de

  Copy full SHA for 85d03de

• README and benchmark improvements (#867)

  Show description for b4d0768
  HDCharles

  authored
  b4d0768

  Copy full SHA for b4d0768

Commits on Sep 10, 2024

• Revert "Pin PT version: Fix FPX Inductor error" (#843)

  Show description for e283743
  

  Mark Saroufim

  authored
  e283743

  Copy full SHA for e283743

• Fix build error - error: function "torchao::marlin_24_gemm" has already been defined (previous definition at line 83) (#863)

  Show description for d2226a4
  kshitij12345

  authored
  d2226a4

  Copy full SHA for d2226a4

• Fix Quantization subclass Int8DynamicallyQuantizedLinearWeight typo (#860)

  Show description for 3ac2ab8
  YihengBrianWu

  authored
  3ac2ab8

  Copy full SHA for 3ac2ab8

• add CI to disallow syntax errors and undefined vars in all Python files (#861)

  Show description for 93b5869
  vkuzo

  authored
  93b5869

  Copy full SHA for 93b5869

Commits on Sep 9, 2024

• Add ATEN parallel backend

  Show description for 3e9746c
  metascroy

  authored
  3e9746c

  Copy full SHA for 3e9746c

• Add test util for basic tensor subclass functionalities (#839)

  Show description for 00d18ec
  jerryzh168

  authored
  00d18ec

  Copy full SHA for 00d18ec

• Add symmetric quantization with no clipping error in the tensor subclass based API (#845)

  Show description for c842d50
  

  iseeyuan

  and

  Martin Yuan

  authored
  c842d50

  Copy full SHA for c842d50

• Add 2, 3, 4, 5 bit custom ops

  Show description for 9d169a8
  metascroy

  authored
  9d169a8

  Copy full SHA for 9d169a8

• Update README.md (#849)

  jerryzh168

  authored
  684f7cd

  Copy full SHA for 684f7cd

• fix syntax error in eval_combo.py (#852)

  Show description for d313cfa
  vkuzo

  authored
  d313cfa

  Copy full SHA for d313cfa

• Don't run mac builds per commit (#842)

  Show description for e1039ab
  

  Mark Saroufim

  authored
  e1039ab

  Copy full SHA for e1039ab

• Add INT8 mixed-precision training (#748)

  Show description for 9acc9a4
  gau-nernst

  authored
  9acc9a4

  Copy full SHA for 9acc9a4

Commits on Sep 8, 2024

• Update README (#823)

  Show description for 1b317f9
  

  Mark Saroufim

  authored
  1b317f9

  Copy full SHA for 1b317f9

Commits on Sep 7, 2024

• Update the way scale is calculated for affine Symmetric (#805)

  Show description for c6abf2b
  iseeyuan

  authored
  c6abf2b

  Copy full SHA for c6abf2b

• [Float8Quant] Add rowwise scaling option to float8 dyanmic quant (#819)

  drisspg

  authored
  e3ce0b9

  Copy full SHA for e3ce0b9

Commits on Sep 6, 2024

• Revert "Lint fixes #1 torchao/dtypes" (#836)

  Show description for 1ce7da9
  

  jerryzh168

  and

  msaroufim

  authored
  1ce7da9

  Copy full SHA for 1ce7da9

• Create build-wheels-m1.yml (#822)

  Show description for a60e312
  

  Mark Saroufim

  and

  seemethere

  authored
  a60e312

  Copy full SHA for a60e312

• Lint fixes #1 torchao/dtypes (#827)

  Show description for 144445a
  jainapurva

  authored
  144445a

  Copy full SHA for 144445a

• Unbreak build after #621 (#826)

  andrewor14

  authored
  88440e8

  Copy full SHA for 88440e8

• Expose hqq through `uintx_weight_only` API (#786)

  Show description for 0601b5c
  jerryzh168

  authored
  0601b5c

  Copy full SHA for 0601b5c

• Update docs + add deprecation warning (#825)

  Show description for 92dcc62
  jcaip

  authored
  92dcc62

  Copy full SHA for 92dcc62

• fix readme (#824)

  Hanxian97

  authored
  28dc4fb

  Copy full SHA for 28dc4fb

• Add sparse marlin AQT layout (#621)

  Show description for 65d86c6
  

  Diogo-V

  and

  jcaip

  authored
  65d86c6

  Copy full SHA for 65d86c6

• [StaticQuant] add a linear observer class and test (#807)

  drisspg

  authored
  422301b

  Copy full SHA for 422301b

• Lowbit custom torch op

  Show description for 038a0a2
  metascroy

  authored
  038a0a2

  Copy full SHA for 038a0a2

Commits on Sep 5, 2024

• Add static quant to float8 (#787)

  Show description for a246d87
  jainapurva

  authored
  a246d87

  Copy full SHA for a246d87