Commits

Branch selector

 jcaip/quantize_param_support

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 16, 2025

• fix imports

  jcaip

  committed
  92f9774

  Copy full SHA for 92f9774

• Merge branch 'main' into jcaip/quantize_param_support

  jcaip

  authored
  0fd52b2

  Copy full SHA for 0fd52b2

• Update compatibility matrix (#3178)

  Show description for 16c7d09
  andrewor14

  authored
  16c7d09

  Copy full SHA for 16c7d09

• Fix SyntaxWarning during installation / import (#3184)

  Show description for a9694a5
  jerryzh168

  authored
  a9694a5

  Copy full SHA for a9694a5

• add tests for value error

  jcaip

  committed
  7de9198

  Copy full SHA for 7de9198

Commits on Oct 15, 2025

• update check

  jcaip

  committed
  eb5c987

  Copy full SHA for eb5c987

• Revert "[Reland][CPU] Add ops for float8 linear" (#3179)

  Show description for 30082cb
  vkuzo

  authored
  30082cb

  Copy full SHA for 30082cb

• [mxfp8 moe training] add triton kernel for mxfp8 quantization along dim0 (#3128)

  Show description for 664124a
  danielvegamyhre

  authored
  664124a

  Copy full SHA for 664124a

• Simplify quantization README.md (#3162)

  Show description for fdd8fe7
  vkuzo

  authored
  fdd8fe7

  Copy full SHA for fdd8fe7

• fix ruff

  jcaip

  committed
  0eb5c61

  Copy full SHA for 0eb5c61

• fix typo

  jcaip

  committed
  7f8f775

  Copy full SHA for 7f8f775

• [moe training] update benchmarks using more reliable b200 gpu (#3174)

  Show description for ff16308
  danielvegamyhre

  authored
  ff16308

  Copy full SHA for ff16308

• [CI]Enable CI for TorchAO on XPU Using Docker (#3027)

  Show description for b633f89
  

  DiweiSun

  and

  chuanqi129

  authored
  b633f89

  Copy full SHA for b633f89

Commits on Oct 14, 2025

• fix default behavior

  jcaip

  committed
  b56bde9

  Copy full SHA for b56bde9

• [benchmarks] Add inference-only roofline for float8 (#3167)

  Show description for d43090a
  drisspg

  authored
  d43090a

  Copy full SHA for d43090a

• fixup missing callsites for _scale_e8m0 -> scale rename (#3173)

  Show description for ce9f009
  vkuzo

  authored
  ce9f009

  Copy full SHA for ce9f009

• Add parq utility to create an optimizer (#3165)

  Show description for 8878f30
  metascroy

  authored
  8878f30

  Copy full SHA for 8878f30

• call named_modules once per model prepare

  Show description for 41a6c11
  navsud

  authored
  41a6c11

  Copy full SHA for 41a6c11

• add learnablefakequantize to pt2e flow

  Show description for b1097de
  navsud

  authored
  b1097de

  Copy full SHA for b1097de

• [CPU] [FP8 SDPA] Enable FP8 SDPA pattern match (#3076)

  Show description for c5e7c18
  Valentine233

  authored
  c5e7c18

  Copy full SHA for c5e7c18

• [moe training] update readme (#3163)

  Show description for 838dceb
  danielvegamyhre

  authored
  838dceb

  Copy full SHA for 838dceb

Commits on Oct 13, 2025

• Remove LLaMa 2 from quantization README.md (#3161)

  Show description for e9c98c0
  vkuzo

  authored
  e9c98c0

  Copy full SHA for e9c98c0

• Small updates to main torchao README.md (#3160)

  Show description for 8482770
  vkuzo

  authored
  8482770

  Copy full SHA for 8482770

• rename MXTensor and NVFP4Tensor's to_dtype to dequantize (#3169)

  Show description for fe9be99
  vkuzo

  authored
  fe9be99

  Copy full SHA for fe9be99

• Rename NVFP4Tensor's _per_tensor_scale and _act_per_tensor_scale fields (#3168)

  Show description for ba4593f
  vkuzo

  authored
  ba4593f

  Copy full SHA for ba4593f

• rename NVFP4Tensor's _scale_e4m3 field to scale (#3166)

  Show description for 1c7ceea
  vkuzo

  authored
  1c7ceea

  Copy full SHA for 1c7ceea

• rename MXTensor's _scale_e8m0 to scale (#3164)

  Show description for 8a4f586
  vkuzo

  authored
  8a4f586

  Copy full SHA for 8a4f586

• update

  jcaip

  committed
  ef20a86

  Copy full SHA for ef20a86

• [BE] [moe training] bench script for single device moe layer (#3126)

  Show description for fb1450d
  danielvegamyhre

  authored
  fb1450d

  Copy full SHA for fb1450d

• fix parq

  jcaip

  committed
  041e00b

  Copy full SHA for 041e00b

• fix import

  jcaip

  committed
  addfc11

  Copy full SHA for addfc11

• Merge branch 'main' into jcaip/quantize_param_support

  jcaip

  authored
  fcac8f6

  Copy full SHA for fcac8f6

• update CR feedback

  jcaip

  committed
  05c9518

  Copy full SHA for 05c9518

• Remove config functions like `int4_weight_only` (#3145)

  Show description for 53a66f8
  andrewor14

  authored
  53a66f8

  Copy full SHA for 53a66f8

• [BE] [moe training] generic bench script for torchtitan models (#3124)

  Show description for 596da93
  danielvegamyhre

  authored
  596da93

  Copy full SHA for 596da93

Pagination

Previous

Next