All pull requests

Pull requests

• Align weight order between in-place and policy-built float8 layers

  awaiting reviewsize:M
  #23826·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  4

• Load a per-channel int4 layer saved with a custom weight quantizer

  awaiting reviewsize:S
  #23825·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  1

• Register GPTQConfig in keras.quantizers.deserialize

  awaiting reviewsize:S
  #23824·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  4

• Retrack the torch parameters of a Sequential after quantization

  awaiting reviewsize:M
  #23823·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  1

• Keep lora_alpha when Dense and Embedding build

  awaiting reviewsize:M
  #23822·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  1

• Serialize a DTypePolicyMap without changing its policies

  awaiting reviewsize:M
  #23821·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  6

• Keep the arguments of fake_quant_with_min_max_vars on symbolic inputs

  awaiting reviewsize:M
  #23820·JyotinderSingh opened Oct 2, 2026Collaborator·4 tasks done·

  ·
  6

• Serialize axis in CosineSimilarity and gain in Identity

  size:S
  #23819·GargiGupta-io opened Oct 2, 2026Contributor·4 tasks done·

  ·
  1

  2

• Refuse a vocabulary file path in from_config under safe mode

  size:M
  #23818·SABITHSAHEB opened Oct 2, 2026Contributor·3 of 4 tasks·

  ·
  1

  3

• Fix pearson_correlation returning NaN on zero-variance input

  size:S
  #23814·YashwanthNomula opened Oct 1, 2026·

  ·
  1

  5

• Add backend-agnostic implementation for nansum

  size:M
  #23813·samudraneel05 opened Oct 1, 2026Contributor·4 tasks done·

  ·
  1

  4

• Fix gradients of ops.hypot in TensorFlow backend

  size:M
  #23808·vruotolo opened Sep 30, 2026·4 tasks done·

  ·
  2

  2

• Support 2-D vectors in ops.cross on the torch backend

  size:S
  #23801·jayzuccarelli opened Sep 30, 2026·4 tasks done·

  ·
  1

  7

• Guard pearson_correlation against zero-variance rows (prevents NaN poisoning)

  size:S
  #23798·xiaou61 opened Sep 30, 2026·

  ·
  1

  2

• fix: unknowkn shape as per step execution for Tensorflow

  size:M
  #23796·maitry63 opened Sep 30, 2026Collaborator·4 tasks done·

  ·
  1

  3

• Enable pluggable backends.

  size:M
  #23795·hertschuh opened Sep 30, 2026Collaborator·4 tasks done·

  ·
  16

• Support Lanczos interpolation in torch backend resize

  size:Mstat:awaiting response from contributor
  #23792·laxmareddyp opened Sep 29, 2026Collaborator·4 tasks done·

  ·
  1

  10

• Harden RestrictedUnpickler against array-state confusion in npz_utils

  size:M
  #23790·rodry957 opened Sep 29, 2026·

  ·
  1

  13

• fix: Change exception from AssertionError to ValueError

  size:XS
  #23788·jjdev11 opened Sep 29, 2026·4 tasks done·

  ·
  1

  5

• reject resample bomb from untrusted WAV sample rate in audio_dataset

  size:M
  #23785·SABITHSAHEB opened Sep 29, 2026Contributor·2 of 4 tasks·

  ·
  3

• Keep beta in layer normalization for large inputs

  size:S
  #23782·vishwakt opened Sep 29, 2026·4 tasks done·

  ·
  1

  1

• Fix Layer.__delattr__ to untrack sub-layers (#23761)

  size:S
  #23781·Pushpak731 opened Sep 29, 2026·4 tasks done·

  ·
  1

  5

• restrict the world-writable /tmp .keras fallback dir to the owner

  size:Mstat:awaiting keras-eng
  #23774·SABITHSAHEB opened Sep 27, 2026Contributor·2 of 4 tasks·

  ·
  5

  

• Fix several Doc issues for mobilenet, mobilenet_v2, and mobilenet_v3

  size:Sstat:awaiting response from contributor
  #23765·ILCSFNO opened Sep 25, 2026Contributor·

  ·
  1

  5

  

• Fix symbolic std dtype for complex inputs

  size:Sstat:awaiting keras-eng
  #23749·aswanth-07 opened Sep 24, 2026Contributor·3 tasks done·

  ·
  1

  3