Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Oct 8, 2024

• Revert "Rename Layout -> TensorImpl" (#1040)

  Show description for 745085f
  

  Mark Saroufim

  authored
  745085fView commit details

  Copy full SHA for 745085f

  View code at this pointBrowse repository at this point

• Rename Layout -> TensorImpl (#1028)

  jainapurva

  authored
  cc8bf85View commit details

  Copy full SHA for cc8bf85

  View code at this pointBrowse repository at this point

• fix failing fbcode tests (#1033)

  Show description for 35ea27b
  

  Jesse Cai

  authored
  35ea27bView commit details

  Copy full SHA for 35ea27b

  View code at this pointBrowse repository at this point

Commits on Oct 7, 2024

• float8 training axiswise scaling support with per-gemm-argument configuration (#940)

  Show description for dec0313
  vkuzo

  authored
  dec0313View commit details

  Copy full SHA for dec0313

  View code at this pointBrowse repository at this point

• add axiswise scaling to Float8Linear (#920)

  Show description for e76db70
  vkuzo

  authored
  e76db70View commit details

  Copy full SHA for e76db70

  View code at this pointBrowse repository at this point

• Activation Aware Weight Quantization (AWQ) (#743)

  Show description for f81fe11
  vayuda

  authored
  f81fe11View commit details

  Copy full SHA for f81fe11

  View code at this pointBrowse repository at this point

• add axiswise granularity to Float8Tensor (#919)

  Show description for 52d27a1
  vkuzo

  authored
  52d27a1View commit details

  Copy full SHA for 52d27a1

  View code at this pointBrowse repository at this point

Commits on Oct 6, 2024

• [low-bit optim] Fix load state dict when device is different (#1021)

  Show description for c187f87
  gau-nernst

  authored
  c187f87View commit details

  Copy full SHA for c187f87

  View code at this pointBrowse repository at this point

Commits on Oct 4, 2024

• Add Float8 support for AQT tensor parallel (#1003)

  jainapurva

  authored
  9e2a253View commit details

  Copy full SHA for 9e2a253

  View code at this pointBrowse repository at this point

• training ir torchao migration

  Show description for d2982d5
  yushangdi

  authored
  d2982d5View commit details

  Copy full SHA for d2982d5

  View code at this pointBrowse repository at this point

• Unskip `test_choose_qparams_token_asym` in 2.6 (#1004)

  Show description for 5dd0132
  jerryzh168

  authored
  5dd0132View commit details

  Copy full SHA for 5dd0132

  View code at this pointBrowse repository at this point

• Pin torch nightly - Oct 1 (#1008)

  Show description for 0ffbf85
  jainapurva

  authored
  0ffbf85View commit details

  Copy full SHA for 0ffbf85

  View code at this pointBrowse repository at this point

Commits on Oct 2, 2024

• Float8 dynamic autoquant (#946)

  jainapurva

  authored
  9229df9View commit details

  Copy full SHA for 9229df9

  View code at this pointBrowse repository at this point

Commits on Oct 1, 2024

• Gptq mt refactor (#914)

  Show description for 83d5b63
  

  danielpatrickhug

  and

  HDCharles

  authored
  83d5b63View commit details

  Copy full SHA for 83d5b63

  View code at this pointBrowse repository at this point

• Add int4 weight-only embedding QAT (#947)

  Show description for 5a4857e
  andrewor14

  authored
  5a4857eView commit details

  Copy full SHA for 5a4857e

  View code at this pointBrowse repository at this point

• [float8] improve eager numerics for dynamic scales and gets on par with torch.compile (#904)

  Show description for 71be315
  

  14 peopleauthored
  71be315View commit details

  Copy full SHA for 71be315

  View code at this pointBrowse repository at this point

• add ruff lint rule to remove unused imports via ruff (#969)

  Show description for 1137f39
  jimexist

  authored
  1137f39View commit details

  Copy full SHA for 1137f39

  View code at this pointBrowse repository at this point

• Print args/kwargs types for utils and examples (#963)

  Show description for 20f3796
  jerryzh168

  authored
  20f3796View commit details

  Copy full SHA for 20f3796

  View code at this pointBrowse repository at this point

• BitNet b1.58 training (#930)

  Show description for a382752
  gau-nernst

  authored
  a382752View commit details

  Copy full SHA for a382752

  View code at this pointBrowse repository at this point

Commits on Sep 30, 2024

• Skip test_choose_qparams_token_asym on pt 2.6 (#981)

  Show description for 68e1886
  

  Mark Saroufim

  authored
  68e1886View commit details

  Copy full SHA for 68e1886

  View code at this pointBrowse repository at this point

Commits on Sep 28, 2024

• Fix WOQ int8 failures (#884)

  Show description for 39ce823
  leslie-fang-intel

  authored
  39ce823View commit details

  Copy full SHA for 39ce823

  View code at this pointBrowse repository at this point

Commits on Sep 27, 2024

• Supporting tensor parallelism for int8 weight only quant (#939)

  Show description for 72d2518
  

  jerryzh168

  and

  kwen2501

  authored
  72d2518View commit details

  Copy full SHA for 72d2518

  View code at this pointBrowse repository at this point

Commits on Sep 26, 2024

• Update INT8 mixed-precision training test to be less flaky (#950)

  gau-nernst

  authored
  ceec750View commit details

  Copy full SHA for ceec750

  View code at this pointBrowse repository at this point

• int8 dynamic quant + bsr support (#821)

  Show description for 4b5b5ee
  

  Jesse Cai

  authored
  4b5b5eeView commit details

  Copy full SHA for 4b5b5ee

  View code at this pointBrowse repository at this point

• [float8] all-reduce amax on dp mesh instead of global pg (#933)

  Show description for da0bbe3
  weifengpy

  authored
  da0bbe3View commit details

  Copy full SHA for da0bbe3

  View code at this pointBrowse repository at this point

Commits on Sep 25, 2024

• Fix fp8-all-gather buck errors

  Show description for b521c9b
  y-sq

  authored
  b521c9bView commit details

  Copy full SHA for b521c9b

  View code at this pointBrowse repository at this point

• Add composable QAT quantizer (#938)

  Show description for 6a4d064
  andrewor14

  authored
  6a4d064View commit details

  Copy full SHA for 6a4d064

  View code at this pointBrowse repository at this point

Commits on Sep 24, 2024

• Float8 autoquant weight only (#866)

  jainapurva

  authored
  7dff17aView commit details

  Copy full SHA for 7dff17a

  View code at this pointBrowse repository at this point

• [low-bit optim] Change 8-bit and FP8 optim block size from 2048 to 256 to match new bnb v0.44 (#927)

  gau-nernst

  authored
  728d629View commit details

  Copy full SHA for 728d629

  View code at this pointBrowse repository at this point

• clean up device checks in float8 unit test files (#923)

  Show description for 26e790d
  vkuzo

  authored
  26e790dView commit details

  Copy full SHA for 26e790d

  View code at this pointBrowse repository at this point

Commits on Sep 23, 2024

• [float8] fix typo in bitwise_identical unit test (#918)

  Show description for 1d6f8e2
  weifengpy

  authored
  1d6f8e2View commit details

  Copy full SHA for 1d6f8e2

  View code at this pointBrowse repository at this point

Commits on Sep 16, 2024

• Add empty_like for NF4Tensor to support offloading (#881)

  Show description for a584e24
  janeyx99

  authored
  a584e24View commit details

  Copy full SHA for a584e24

  View code at this pointBrowse repository at this point

Commits on Sep 13, 2024

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

  drisspg

  authored
  3fa38aaView commit details

  Copy full SHA for 3fa38aa

  View code at this pointBrowse repository at this point

Commits on Sep 12, 2024

• Renaming fpx to floatx (#877)

  Show description for 8236a87
  jerryzh168

  authored
  8236a87View commit details

  Copy full SHA for 8236a87

  View code at this pointBrowse repository at this point

Commits on Sep 10, 2024

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

  Show description for e283743
  

  Mark Saroufim

  authored
  e283743View commit details

  Copy full SHA for e283743

  View code at this pointBrowse repository at this point