All pull requests

Pull requests

• docs: add a technical guide to static weight decode fusion

  topic: documentationtopic: operatortopic: quantization
  #8809·justinchuby opened Nov 13, 2026Member·1 task·

  ·

  · 1.24
  3

• fix(reference): honor scan axes and directions in Scan

  #8808·xgbtxy opened Nov 13, 2026·1 task done·

  ·
  1

  1

• Update Optimizations SIG roadmap

  #8806·andife opened Nov 13, 2026Member·

  ·
  1

• Add Probabilistic Programming WG to roadmap

  #8805·andife opened Nov 13, 2026Member·

  ·
  3

• test(cpp): fix function_verify_test build with exceptions disabled

  #8804·andife opened Nov 13, 2026Member·1 task·

  ·
  1

• fix(inliner): rename device-sharding tensor references

  module: inlinertopic: bug fixtopic: test
  #8803·justinchuby opened Nov 13, 2026Member·1 task·

  ·
  2

• feat(defs): add opset 29 MelWeightMatrix function body

  module: inlinermodule: reference implementationmodule: schematopic: operatortopic: spec clarification
  #8802·justinchuby opened Nov 12, 2026Member·1 task·

  ·
  4

• fix(defs): stabilize opset 29 math function bodies

  module: inlinermodule: reference implementationmodule: schematopic: operator
  #8801·justinchuby opened Nov 12, 2026Member·1 task done·

  ·

  · 1.24
  4

• test(backend): cover bfloat16 matmul and batch broadcasting

  #8797·luke-opus opened Nov 11, 2026Contributor·1 task done·

  ·
  1

• fix(reference): do not store omitted node outputs under the empty name

  #8795·sohailshk opened Nov 11, 2026·1 task·

  ·

• fix(tools): fix fill value handling in replace_initializer_by_constant_of_shape

  #8794·sohailshk opened Nov 11, 2026·1 task done·

  ·

• fix(reference): treat 1-D FeatureVectorizer inputs as [1, C]

  #8793·Iamsohungrynow opened Nov 11, 2026·1 task done·

  ·
  3

• fix(reference): keep intercepts optional in LinearRegressor and LinearClassifier

  #8792·Iamsohungrynow opened Nov 11, 2026·1 task done·

  ·

• Fix Hardmax opset 12 to 13 conversion semantics

  #8791·heathsanchez opened Nov 11, 2026·1 task·

  ·
  5

• fix(reference): return float outputs from ML Scaler, LinearRegressor and FeatureVectorizer

  #8790·sohailshk opened Nov 11, 2026·1 task done·

  ·

• fix(reference): stack Loop scan outputs on a new leading axis

  #8789·sohailshk opened Nov 11, 2026·1 task done·

  ·

• fix(reference): compute NegativeLogLikelihoodLoss in float32 for float16/bfloat16

  #8788·bhaskargurram-ai opened Nov 11, 2026Contributor·1 task done·

  ·
  1

• feat(quantization): support multi-axis blocked quantization

  #8787·jburo1 opened Nov 10, 2026Contributor·1 task done·

  ·
  2

• fix(reference): compute LpPool in float32 for float16/bfloat16

  #8786·AP24110011695 opened Nov 10, 2026·

  ·
  4

• Generate operator registrations from spec files

  topic: better engineering
  #8785·cbourjau opened Nov 10, 2026Contributor·2 of 3 tasks·

  ·
  14

• Move declarative operator schemas to spec files

  topic: better engineering
  #8784·cbourjau opened Nov 10, 2026Contributor·2 of 3 tasks·

  ·
  4

• RFC: Markdown Operator Schema Definition and Code Generation

  rfctopic: better engineering
  #8783·cbourjau opened Nov 10, 2026Contributor·1 task done·

  ·
  5

• feat(reference): implement Multinomial sampling

  #8781·hulkbig opened Nov 10, 2026Contributor·1 task·

  ·

  · 1.24
  2

• fix(reference): filter complete StringNormalizer elements

  #8779·hulkbig opened Nov 9, 2026Contributor·1 task·

  ·
  3

• fix(reference): broadcast QLinearMatMul parameters per row

  #8778·hulkbig opened Nov 9, 2026Contributor·1 task·

  ·
  1