Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Jul 14, 2026

• Gate int8 channelwise neondot qmatmul behind runtime FEAT_DotProd check

  Show description for eec7681
  metascroy

  authored
  eec7681View commit details

  Copy full SHA for eec7681

  Browse repository at this point

Commits on Jul 10, 2026

• optim: cache qmap values as tuples (#4552)

  Show description for bfbc842
  

  nascheme

  and

  rgommers

  authored
  bfbc842View commit details

  Copy full SHA for bfbc842

  Browse repository at this point

• cpu: lock ukernel registration tables (#4561)

  Show description for e98d7a0
  

  nascheme

  and

  rgommers

  authored
  e98d7a0View commit details

  Copy full SHA for e98d7a0

  Browse repository at this point

• Re-enable building under a free-threaded Python interpreter (#4537)

  Show description for d31dad6
  

  nascheme

  and

  rgommers

  authored
  d31dad6View commit details

  Copy full SHA for d31dad6

  Browse repository at this point

Commits on Jul 9, 2026

• [xpu][float8] device agnostic test base,compile, numerics_integration (#3823)

  Show description for edaf375
  arlesniak

  authored
  edaf375View commit details

  Copy full SHA for edaf375

  Browse repository at this point

Commits on Jul 8, 2026

• Fix typo in e2e model level benchmarks - h100->b200 (#4547)

  Show description for f229086
  Pqlet

  authored
  f229086View commit details

  Copy full SHA for f229086

  Browse repository at this point

Commits on Jul 7, 2026

• fix: use `Int4WeightOnlyConfig.group_size` in fake quant configs (#4518)

  Show description for 768fa14
  oriollinan

  authored
  768fa14View commit details

  Copy full SHA for 768fa14

  Browse repository at this point

• skip failing CI tests (#4549)

  Show description for 04084f2
  vkuzo

  authored
  04084f2View commit details

  Copy full SHA for 04084f2

  Browse repository at this point

Commits on Jul 6, 2026

• float8: snapshot FSDP precomputed scale (#4555)

  Show description for dd0efc7
  

  nascheme

  and

  rgommers

  authored
  dd0efc7View commit details

  Copy full SHA for dd0efc7

  Browse repository at this point

• [ROCm] Add HIP device synchronize event to profiler overhead filter (#4450)

  Show description for 3c466f5
  jagadish-amd

  authored
  3c466f5View commit details

  Copy full SHA for 3c466f5

  Browse repository at this point

Commits on Jun 30, 2026

• make nvfp4_linear handle a pre-quantized activation (#4504)

  Show description for d1e0e2d
  larryliu0820

  authored
  d1e0e2dView commit details

  Copy full SHA for d1e0e2d

  Browse repository at this point

Commits on Jun 26, 2026

• Fix spelling typo in 'deprecated_tenosr_subclass' (#4506)

  Show description for cb76f29
  GiGiKoneti

  authored
  cb76f29View commit details

  Copy full SHA for cb76f29

  Browse repository at this point

• Use smallest_normal instead of eps to clamp quant scales (#4534)

  GregoryComer

  authored
  9b3684aView commit details

  Copy full SHA for 9b3684a

  Browse repository at this point

Commits on Jun 24, 2026

• Integrate DeepGEMM FP8 blockwise MoE backend (#4473)

  iamzainhuda

  authored
  260ac85View commit details

  Copy full SHA for 260ac85

  Browse repository at this point

• Revert "pt2e: preserve mutable buffer inputs during prepare" (#4531)

  Show description for 214c37c
  liangel-02

  authored
  214c37cView commit details

  Copy full SHA for 214c37c

  Browse repository at this point

Commits on Jun 23, 2026

• [x86] Fix missing fake metadata in quantized concat linear CPU fusion (#4508)

  Show description for c850249
  Yuxingwang-intel

  authored
  c850249View commit details

  Copy full SHA for c850249

  Browse repository at this point

Commits on Jun 18, 2026

• fix(Int8DynActInt4WeightQuantizer): use scales_precision for runtime modules (#4459)

  Show description for 213d25c
  Anai-Guo

  authored
  213d25cView commit details

  Copy full SHA for 213d25c

  Browse repository at this point

• docs(finetuning): replace TorchTune QAT section with Unsloth (#4312)

  Show description for 4ca0bad
  Anai-Guo

  authored
  4ca0badView commit details

  Copy full SHA for 4ca0bad

  Browse repository at this point

Commits on Jun 16, 2026

• Fix test_x86inductor_fusion CI failure (part 2) (#4503)

  Show description for ede5632
  andrewor14

  authored
  ede5632View commit details

  Copy full SHA for ede5632

  Browse repository at this point

Commits on Jun 15, 2026

• [PT2E][X86] Fall back to plain linear annotation for reused nn.Linear with fusable post op (#4480)

  Show description for 5647df9
  

  waynehacking8

  and

  claude

  authored
  5647df9View commit details

  Copy full SHA for 5647df9

  Browse repository at this point

Commits on Jun 12, 2026

• Fix H100 CI spmd-types install failure (#4481)

  Show description for c92676e
  andrewor14

  authored
  c92676eView commit details

  Copy full SHA for c92676e

  Browse repository at this point

• docs(quant_api): fix RST warnings in FqnToConfig and Int8DynamicActivationIntxWeightConfig docstrings (#3863) (#4475)

  Anai-Guo

  authored
  c42673dView commit details

  Copy full SHA for c42673d

  Browse repository at this point

• pt2e: preserve mutable buffer inputs during prepare (#4466)

  Show description for 7987abe
  

  mansnils

  and

  tom-arm

  authored
  7987abeView commit details

  Copy full SHA for 7987abe

  Browse repository at this point

• docs(qat): update supported configs list, fix unclickable link, add blog ref (#4305)

  Show description for 656e512
  Anai-Guo

  authored
  656e512View commit details

  Copy full SHA for 656e512

  Browse repository at this point

• Fix test_x86inductor_fusion CI failure (test_qconv2d_maxpool2d_linear) (#4479)

  Show description for 2f7431a
  andrewor14

  authored
  2f7431aView commit details

  Copy full SHA for 2f7431a

  Browse repository at this point

• Fix internal test_quantized_scaled_dot_product_op (#4472)

  Show description for 783ac5a
  andrewor14

  authored
  783ac5aView commit details

  Copy full SHA for 783ac5a

  Browse repository at this point

Commits on Jun 11, 2026

• [mxfp8] Fix mx_mm backward crash on non-contiguous grad_output (#4470)

  Show description for 9467cd8
  

  ultism

  and

  claude

  authored
  9467cd8View commit details

  Copy full SHA for 9467cd8

  Browse repository at this point

Commits on Jun 9, 2026

• add quantized qkv-fusion pass for x86 (#4434)

  Show description for abea9e0
  yuchengliu1

  authored
  abea9e0View commit details

  Copy full SHA for abea9e0

  Browse repository at this point

Commits on Jun 5, 2026

• Add blockwise FP8 MoE functionality with emulated GEMM (#4460)

  Show description for 5165bfb
  iamzainhuda

  authored
  5165bfbView commit details

  Copy full SHA for 5165bfb

  Browse repository at this point

• Implement aten.cat for Int8QuantizedTrainingLinearWeight (#2619) (#4465)

  Show description for 27fc8ff
  

  lollinng

  and

  claude

  authored
  27fc8ffView commit details

  Copy full SHA for 27fc8ff

  Browse repository at this point

• [xpu][feat] Add XPU support for blockwise FP8 (#4096)

  Show description for 24d796f
  Stonepia

  authored
  24d796fView commit details

  Copy full SHA for 24d796f

  Browse repository at this point

Commits on Jun 3, 2026

• Remove named tensor Dimname overload reference (#4462)

  jbschlosser

  authored
  4aa8101View commit details

  Copy full SHA for 4aa8101

  Browse repository at this point

• Remove MSLK FP8 tensorwise support (#4461)

  Show description for b2b58d8
  

  cthi

  and

  Chris Thi

  authored
  b2b58d8View commit details

  Copy full SHA for b2b58d8

  Browse repository at this point

Commits on Jun 2, 2026

• [float8] fix `==` typo in `float8_transpose` axiswise dim remapping (#4038)

  Show description for cc49e86
  GotFusion

  authored
  cc49e86View commit details

  Copy full SHA for cc49e86

  Browse repository at this point

Commits on Jun 1, 2026

• [CI] Switch release_model H100 job to linux_job_v3 (OSDC/ARC) (#4457)

  Show description for 8c244cb
  huydhn

  authored
  8c244cbView commit details

  Copy full SHA for 8c244cb

  Browse repository at this point