Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 16, 2024

• Add generic fake quantized embedding for QAT (pytorch#1085)

  Show description for 0b71b8d
  andrewor14

  authored
  0b71b8d

  Copy full SHA for 0b71b8d

• Add AQT tensor parallel for float8_dynamic_quant (pytorch#1078)

  jainapurva

  authored
  7a35695

  Copy full SHA for 7a35695

Commits on Oct 15, 2024

• Add generic fake quantized linear for QAT (pytorch#1020)

  Show description for 48bc81c
  andrewor14

  authored
  48bc81c

  Copy full SHA for 48bc81c

Commits on Oct 11, 2024

• Skip test_fpx_weight_only in fbcode (pytorch#1056)

  jainapurva

  authored
  d4b2f33

  Copy full SHA for d4b2f33

Commits on Oct 10, 2024

• Rename AQT#2 LayoutType -> Layout (pytorch#1049)

  jainapurva

  authored
  7038f8b

  Copy full SHA for 7038f8b

• SpinQuant (pytorch#983)

  Show description for 590f8fb
  tobiasvanderwerff

  authored
  590f8fb

  Copy full SHA for 590f8fb

• Move and rename GranularityType -> Granularity (pytorch#1038)

  Show description for 0f6bae5
  andrewor14

  authored
  0f6bae5

  Copy full SHA for 0f6bae5

• Aqt rename#1 Layout -> TensorImpl (pytorch#1046)

  jainapurva

  authored
  fa6d156

  Copy full SHA for fa6d156

Commits on Oct 9, 2024

• Make module swap the main QAT flow again (pytorch#1037)

  Show description for 9e0a59f
  andrewor14

  authored
  9e0a59f

  Copy full SHA for 9e0a59f

Commits on Oct 8, 2024

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

  Show description for 745085f
  msaroufim

  authored
  745085f

  Copy full SHA for 745085f

• Rename Layout -> TensorImpl (pytorch#1028)

  jainapurva

  authored
  cc8bf85

  Copy full SHA for cc8bf85

• fix failing fbcode tests (pytorch#1033)

  Show description for 35ea27b
  jcaip

  authored
  35ea27b

  Copy full SHA for 35ea27b

Commits on Oct 7, 2024

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

  Show description for dec0313
  vkuzo

  authored
  dec0313

  Copy full SHA for dec0313

• add axiswise scaling to Float8Linear (pytorch#920)

  Show description for e76db70
  vkuzo

  authored
  e76db70

  Copy full SHA for e76db70

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

  Show description for f81fe11
  vayuda

  authored
  f81fe11

  Copy full SHA for f81fe11

• add axiswise granularity to Float8Tensor (pytorch#919)

  Show description for 52d27a1
  vkuzo

  authored
  52d27a1

  Copy full SHA for 52d27a1

Commits on Oct 6, 2024

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

  Show description for c187f87
  gau-nernst

  authored
  c187f87

  Copy full SHA for c187f87

Commits on Oct 4, 2024

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

  jainapurva

  authored
  9e2a253

  Copy full SHA for 9e2a253

• training ir torchao migration

  Show description for d2982d5
  yushangdi

  authored
  d2982d5

  Copy full SHA for d2982d5

• Unskip test_choose_qparams_token_asym in 2.6 (pytorch#1004)

  Show description for 5dd0132
  jerryzh168

  authored
  5dd0132

  Copy full SHA for 5dd0132

• Pin torch nightly - Oct 1 (pytorch#1008)

  Show description for 0ffbf85
  jainapurva

  authored
  0ffbf85

  Copy full SHA for 0ffbf85

Commits on Oct 2, 2024

• Float8 dynamic autoquant (pytorch#946)

  jainapurva

  authored
  9229df9

  Copy full SHA for 9229df9

Commits on Oct 1, 2024

• Gptq mt refactor (pytorch#914)

  Show description for 83d5b63
  

  danielpatrickhug

  and

  HDCharles

  authored
  83d5b63

  Copy full SHA for 83d5b63

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

  Show description for 5a4857e
  andrewor14

  authored
  5a4857e

  Copy full SHA for 5a4857e

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

  Show description for 71be315
  

  14 peopleauthored
  71be315

  Copy full SHA for 71be315

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

  Show description for 1137f39
  jimexist

  authored
  1137f39

  Copy full SHA for 1137f39

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

  Show description for 20f3796
  jerryzh168

  authored
  20f3796

  Copy full SHA for 20f3796

• BitNet b1.58 training (pytorch#930)

  Show description for a382752
  gau-nernst

  authored
  a382752

  Copy full SHA for a382752

Commits on Sep 30, 2024

• Skip test_choose_qparams_token_asym on pt 2.6 (pytorch#981)

  Show description for 68e1886
  msaroufim

  authored
  68e1886

  Copy full SHA for 68e1886

Commits on Sep 28, 2024

• Fix WOQ int8 failures (pytorch#884)

  Show description for 39ce823
  leslie-fang-intel

  authored
  39ce823

  Copy full SHA for 39ce823

Commits on Sep 27, 2024

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

  Show description for 72d2518
  

  jerryzh168

  and

  kwen2501

  authored
  72d2518

  Copy full SHA for 72d2518

Commits on Sep 26, 2024

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

  gau-nernst

  authored
  ceec750

  Copy full SHA for ceec750

• int8 dynamic quant + bsr support (pytorch#821)

  Show description for 4b5b5ee
  jcaip

  authored
  4b5b5ee

  Copy full SHA for 4b5b5ee

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

  Show description for da0bbe3
  weifengpy

  authored
  da0bbe3

  Copy full SHA for da0bbe3

Commits on Sep 25, 2024

• Fix fp8-all-gather buck errors

  Show description for b521c9b
  y-sq

  authored
  b521c9b

  Copy full SHA for b521c9b

Pagination

Previous

Next