Commits

Branch selector

main

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

  Browse repository at this point

• Rename Layout -> TensorImpl (#1028)

  jainapurva

  authored
  cc8bf85View commit details

  Copy full SHA for cc8bf85

  Browse repository at this point

• fix failing fbcode tests (#1033)

  Show description for 35ea27b
  

  Jesse Cai

  authored
  35ea27bView commit details

  Copy full SHA for 35ea27b

  Browse repository at this point

• Optimize 3-bit packing

  Show description for 93ff876
  metascroy

  authored
  93ff876View commit details

  Copy full SHA for 93ff876

  Browse 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

  Browse repository at this point

• add axiswise scaling to Float8Linear (#920)

  Show description for e76db70
  vkuzo

  authored
  e76db70View commit details

  Copy full SHA for e76db70

  Browse repository at this point

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

  Show description for f81fe11
  vayuda

  authored
  f81fe11View commit details

  Copy full SHA for f81fe11

  Browse repository at this point

• Dynamic Float8 benchmarking llama (#1017)

  jainapurva

  authored
  92dd5f5View commit details

  Copy full SHA for 92dd5f5

  Browse repository at this point

• add axiswise granularity to Float8Tensor (#919)

  Show description for 52d27a1
  vkuzo

  authored
  52d27a1View commit details

  Copy full SHA for 52d27a1

  Browse repository at this point

• Refactor `tiktoken` import bare except (#1024)

  Show description for e7331ab
  

  Matthew Hoffman

  authored
  e7331abView commit details

  Copy full SHA for e7331ab

  Browse repository at this point

• Use `importlib.util.find_spec` to check if `lm_eval` is installed instead of trying to import it (#1023)

  Show description for e2301e9
  

  Matthew Hoffman

  authored
  e2301e9View commit details

  Copy full SHA for e2301e9

  Browse 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

  Browse 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

  Browse repository at this point

• training ir torchao migration

  Show description for d2982d5
  yushangdi

  authored
  d2982d5View commit details

  Copy full SHA for d2982d5

  Browse 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

  Browse repository at this point

• Add more videos (#1011)

  

  Mark Saroufim

  authored
  0cb91eaView commit details

  Copy full SHA for 0cb91ea

  Browse repository at this point

• Pin torch nightly - Oct 1 (#1008)

  Show description for 0ffbf85
  jainapurva

  authored
  0ffbf85View commit details

  Copy full SHA for 0ffbf85

  Browse repository at this point

Commits on Oct 3, 2024

• Introduce 6-bit quantization for Llama in torchchat

  Show description for 9ce7ebb
  ramreddymounica

  authored
  9ce7ebbView commit details

  Copy full SHA for 9ce7ebb

  Browse repository at this point

• sync with fbcode (#1005)

  metascroy

  authored
  8945fb3View commit details

  Copy full SHA for 8945fb3

  Browse repository at this point

Commits on Oct 2, 2024

• Add float8_static_activation_float8_weight to user-facing API (#994)

  jainapurva

  authored
  09b8b3cView commit details

  Copy full SHA for 09b8b3c

  Browse repository at this point

• Use checkpoint to enforece the recomputation of fp8 weight

  Show description for e919558
  y-sq

  authored
  e919558View commit details

  Copy full SHA for e919558

  Browse repository at this point

• Add example for awq like flow (#990)

  Show description for 378e6a8
  jerryzh168

  authored
  378e6a8View commit details

  Copy full SHA for 378e6a8

  Browse repository at this point

• Float8 dynamic autoquant (#946)

  jainapurva

  authored
  9229df9View commit details

  Copy full SHA for 9229df9

  Browse 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

  Browse 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

  Browse 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

  Browse 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

  Browse 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

  Browse repository at this point

• Don't call dequantize in `__repr__` (#965)

  Show description for b4055da
  jerryzh168

  authored
  b4055daView commit details

  Copy full SHA for b4055da

  Browse repository at this point

• BitNet b1.58 training (#930)

  Show description for a382752
  gau-nernst

  authored
  a382752View commit details

  Copy full SHA for a382752

  Browse repository at this point

Commits on Sep 30, 2024

• Change arg order in lowbit linear ops to match aten (#982)

  Show description for ae3e7c6
  metascroy

  authored
  ae3e7c6View commit details

  Copy full SHA for ae3e7c6

  Browse repository at this point

• 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

  Browse repository at this point

• no lintrunner (#973)

  

  Mark Saroufim

  authored
  8af72fcView commit details

  Copy full SHA for 8af72fc

  Browse repository at this point

• Add arm neon guards

  Show description for 60ffb86
  metascroy

  authored
  60ffb86View commit details

  Copy full SHA for 60ffb86

  Browse repository at this point

• Change op names to linear_8bit_act_xbit_weight

  Show description for b983f7d
  metascroy

  authored
  b983f7dView commit details

  Copy full SHA for b983f7d

  Browse repository at this point