Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Feb 25, 2026

• Add parameter_name support to _intx_weight_only_transform (#3905)

  

  Jesse Cai

  authored
  cd062f2View commit details

  Copy full SHA for cd062f2

  Browse repository at this point

• Add parameter_name support to _int8_dynamic_activation_intx_weight_transform (#3904)

  

  Jesse Cai

  authored
  1efc5d0View commit details

  Copy full SHA for 1efc5d0

  Browse repository at this point

• Fix _int8_dynamic_activation_int8_weight_transform to use parameter_name properly (#3903)

  

  Jesse Cai

  authored
  323c2ebView commit details

  Copy full SHA for 323c2eb

  Browse repository at this point

• Add parameter_name support to _float8_dynamic_activation_int4_weight_transform (#3902)

  

  Jesse Cai

  authored
  1f016a0View commit details

  Copy full SHA for 1f016a0

  Browse repository at this point

Commits on Feb 24, 2026

• Add parameter_name support to _int4_weight_only_transform (#3901)

  

  Jesse Cai

  authored
  d1fa9a2View commit details

  Copy full SHA for d1fa9a2

  Browse repository at this point

• fix up nvfp4 roofline sweep script (#3927)

  Show description for be10b2d
  vkuzo

  authored
  be10b2dView commit details

  Copy full SHA for be10b2d

  Browse repository at this point

• [optim] Add appearance dtype for optim subclasses to make things work better with DTensor (#3934)

  Show description for 1a9a884
  gau-nernst

  authored
  1a9a884View commit details

  Copy full SHA for 1a9a884

  Browse repository at this point

Commits on Feb 23, 2026

• Add pruning-aware training in torchao.prototype.pat (#3429)

  Show description for 2a37912
  lisjin

  authored
  2a37912View commit details

  Copy full SHA for 2a37912

  Browse repository at this point

Commits on Feb 21, 2026

• Add Linear+BN folding to _fuse_conv_bn_()

  Show description for 450699a
  SS-JIA

  authored
  450699aView commit details

  Copy full SHA for 450699a

  Browse repository at this point

• Lowbit Embedding on x86

  Show description for 3d3dd50
  JacobSzwejbka

  authored
  3d3dd50View commit details

  Copy full SHA for 3d3dd50

  Browse repository at this point

Commits on Feb 20, 2026

• [moe training] refactor configs, recipes; support converting linears + grouped gemms in a single quantize_() call (#3862)

  Show description for 4a42d32
  danielvegamyhre

  authored
  4a42d32View commit details

  Copy full SHA for 4a42d32

  Browse repository at this point

Commits on Feb 19, 2026

• Land 3894, 3887, 3884, 3883 into main from base branch (#3920)

  Show description for 920c502
  

  Jesse Cai

  authored
  920c502View commit details

  Copy full SHA for 920c502

  Browse repository at this point

• Make `_emulated_mxfp8_scaled_grouped_mm_2d_2d` torch.compile compatible (#3906)

  Show description for 01b37b2
  lyprince

  authored
  01b37b2View commit details

  Copy full SHA for 01b37b2

  Browse repository at this point

• move mx documentation from README.md to docs site (#3915)

  Show description for f8bbce5
  vkuzo

  authored
  f8bbce5View commit details

  Copy full SHA for f8bbce5

  Browse repository at this point

• LowBitPacking kernels on x86

  Show description for 74c6d3d
  JacobSzwejbka

  authored
  74c6d3dView commit details

  Copy full SHA for 74c6d3d

  Browse repository at this point

• Add support for `torch._higher_order_ops.while_loop` in pt2e quantization (#3916)

  Show description for 9f8ae93
  jerryzh168

  authored
  9f8ae93View commit details

  Copy full SHA for 9f8ae93

  Browse repository at this point

• Add flashinfer-python to CI (#3910)

  Show description for b5ca5dc
  jerryzh168

  authored
  b5ca5dcView commit details

  Copy full SHA for b5ca5dc

  Browse repository at this point

Commits on Feb 18, 2026

• Add FP8-INT4 checkpoint upload code (#3913)

  Show description for 9429ee2
  jerryzh168

  authored
  9429ee2View commit details

  Copy full SHA for 9429ee2

  Browse repository at this point

• Add AffineFakeQuantize for groupwise quantization support

  Show description for 3075bb6
  navsud

  authored
  3075bb6View commit details

  Copy full SHA for 3075bb6

  Browse repository at this point

• Assing meta["val"] to new bias nodes in fold_bn_weights_into_conv_node. (#3907)

  Show description for d5e4785
  abeakkas

  authored
  d5e4785View commit details

  Copy full SHA for d5e4785

  Browse repository at this point

Commits on Feb 17, 2026

• Allow prototype float8 tensor to be saved (#3909)

  Show description for 83b0e7c
  jerryzh168

  authored
  83b0e7cView commit details

  Copy full SHA for 83b0e7c

  Browse repository at this point

• Build: put x86 options in a dedicated function (#3892)

  Xia-Weiwen

  authored
  c893ddfView commit details

  Copy full SHA for c893ddf

  Browse repository at this point

• Align _emulated_mxfp8_scaled_grouped_mm_2d_2d API with 2d_3d variant (#3893)

  Show description for 396d78b
  lyprince

  authored
  396d78bView commit details

  Copy full SHA for 396d78b

  Browse repository at this point

Commits on Feb 16, 2026

• add mxfp8 and nvfp4 support to safetensors (#3668)

  liangel-02

  authored
  5906856View commit details

  Copy full SHA for 5906856

  Browse repository at this point

Commits on Feb 14, 2026

• Add plain int4_packing_format support for Float8DynamicActivationInt4WeightConfig (#3714)

  Show description for 9cd4107
  jerryzh168

  authored
  9cd4107View commit details

  Copy full SHA for 9cd4107

  Browse repository at this point

Commits on Feb 13, 2026

• Avoid MSLK on NVIDIA B200/GB200 for per-tensor scaled weights (fallback to TORCH) (#3786)

  Show description for 605ae1b
  lordaarush

  authored
  605ae1bView commit details

  Copy full SHA for 605ae1b

  Browse repository at this point

• Change PR check to check for `module:` labels instead (#3888)

  Show description for e915c07
  jerryzh168

  authored
  e915c07View commit details

  Copy full SHA for e915c07

  Browse repository at this point

• [X86] Update the call path to access CPUInfo properties (#3845)

  Show description for 019d12b
  Yuxingwang-intel

  authored
  019d12bView commit details

  Copy full SHA for 019d12b

  Browse repository at this point

• docs: fix broken cross-references and targets (#3881)

  Show description for cfbfa8b
  

  ArivunidhiA

  and

  Arivunidhi A

  authored
  cfbfa8bView commit details

  Copy full SHA for cfbfa8b

  Browse repository at this point

• Add support for `torch._higher_order_ops.scan` in pt2e quantization (#3882)

  Show description for a8a8ca9
  jerryzh168

  authored
  a8a8ca9View commit details

  Copy full SHA for a8a8ca9

  Browse repository at this point

• Delete output.png (#3890)

  

  Jesse Cai

  authored
  8b2b2e2View commit details

  Copy full SHA for 8b2b2e2

  Browse repository at this point

• Separately control the activation quantization granularity (#3524)

  Show description for 603e372
  cyxlily

  authored
  603e372View commit details

  Copy full SHA for 603e372

  Browse repository at this point

• [CPU] support scaled embedding bag pattern match for FP8 output (#3801)

  Show description for 41e02b5
  yanbing-j

  authored
  41e02b5View commit details

  Copy full SHA for 41e02b5

  Browse repository at this point

• docs: fix duplicate conclusion labels (#3870)

  Show description for 94df7db
  

  ArivunidhiA

  and

  Arivunidhi A

  authored
  94df7dbView commit details

  Copy full SHA for 94df7db

  Browse repository at this point

• [Reland][X86] add inductor lowering path for dequantize_float8 (#3859)

  Show description for 5b37faa
  Xia-Weiwen

  authored
  5b37faaView commit details

  Copy full SHA for 5b37faa

  Browse repository at this point