keras-team/keras

Issues

• TextVectorization does not convert Cyrillic characters to lowercase

  #19668 opened 13 days ago by Ybisalt
  3

• Overriding `Layer.forward` unexpectedly changes the signature of `Layer.__call__` under torch backend

  #19730 opened 3 hours ago by LarsKue
  0

• `plot_model` does not work for all models in `keras.applications`

  #19717 opened 4 days ago by Star9daisy
  2

• loss and val_loss decreasing

  #19658 opened 16 days ago by ayatirfann
  2

• Multi-process batch size not calculated correctly

  #19726 opened 2 days ago by natbprice
  0

• Keras does not save weights properly

  #19693 opened 10 days ago by gcelano
  2

• `keras.Sequential` sometimes states misleading reason for failing to construct model

  #19701 opened 9 days ago by briango28
  2

• Resizing layer meets division by zero crash with jax backend

  #19725 opened 2 days ago by AGFACBNNR
  1

• CategoryEncoding generate output with wrong shape when output_mode='one_hot' and the last dim is 1

  #19724 opened 2 days ago by AGFACBNNR
  0

• CenterCrop meets a division-by-zero with jax backend

  #19723 opened 2 days ago by AGFACBNNR
  0

• softmax sliently generate a wrong output when the mask has an incompatible shape

  #19722 opened 2 days ago by AGFACBNNR
  0

• averagePooling2D calculates wrongly with torch backend

  #19721 opened 2 days ago by AGFACBNNR
  1

• Implement tool for saved Keras model file inspection, diff, and patching.

  #19705 opened 9 days ago by pmasousa
  5

• ValueError: (F1Score|FBetaScore) expects 2D inputs with shape (batch_size, output_dim).

  #19716 opened 4 days ago by msusol
  0

• More Customisation in utils.ProgBar

  #19684 opened 11 days ago by stellarpower
  6

• Request for a map function like map_fn in TF and vmap in Jax

  #19708 opened 8 days ago by yifengshao
  5

• FeatureSpace multiple output from one input

  #19697 opened 9 days ago by zippeurfou
  3

• NumPy 2.0 support

  #19691 opened 10 days ago by jakirkham
  3

• AttributeError raised: 'list' object has no attribute 'dtype' when running the official example of SparseCategoricalAccuracy, TopKCategoricalAccuracy, SparseTopKCategoricalAccuracy

  #19709 opened 7 days ago by Star9daisy
  2

• Unable to make two instances of the MobileNetV3 within the same model

  #19689 opened 7 days ago by DLumi
  2

• Potential bug in legacy h5 weights loading.

  #19694 opened 9 days ago by torzdf
  2

• Add backend-agnostic worker-process data loading

  #19692 opened 9 days ago by LarsKue
  8

• Enable Discussions Tab in Github

  #19696 opened 9 days ago by stellarpower
  1

• Feature request: keras.ops.linalg.lstsq

  #19678 opened 9 days ago by jonbarron
  4

• How to save Scikit-Learn-Keras Model into a Persistence File (pickle/hd5/json/yaml) with Keras 3

  #19650 opened 10 days ago by CAW9
  13

• Example Doubt

  #19683 opened 10 days ago by emi-dm
  3

• Multihead Attention Seed Specification

  #19688 opened 10 days ago by egehancosgun
  1

• Unusual behavior of `predict` for JAX backend

  #19639 opened 16 days ago by AakashKumarNain
  21

• Progress bar crash when empty dataset

  #19687 opened 11 days ago by stridge-cruxml
  1

• No module named 'keras.src.engine'

  #19677 opened 12 days ago by UmerElsaharty
  7

• Have the plan to support fp8 inference?

  #19671 opened 13 days ago by lingzhi98
  6

• RaggedTensor

  #19646 opened 18 days ago by markub3327
  4

• JAX array conversion failure in Keras model prediction

  #19674 opened 11 days ago by Qazalbash
  3

• TypeError when loading Keras 2 legacy models with nested outputs in Keras 3

  #19670 opened 12 days ago by torzdf
  2

• To Keras community: What interpretations do you have for these curves?

  #19676 opened 12 days ago by sondosaabed
  3

• StringLookup does not return expected dtype for multi_hot

  #19660 opened 12 days ago by Toku11
  1

• On JAX, Keras replaces any exception inside `call` method of `keras.Model` subclass with misleading error

  #19675 opened 12 days ago by burnpanck
  2

• Dice loss - incorrect

  #19637 opened 20 days ago by td-jakubl
  2

• The loss becomes neagative from positive values dring taining loop

  #19638 opened 20 days ago by yijianSU22
  4

• Shape error for some use cases of `binary_crossentropy`.

  #19662 opened 16 days ago by albertaillet
  3

• Regression: Cannot export model with bilinear UpSampling2D and dynamic inputs

  #19667 opened 13 days ago by pwuertz
  1

• Layernorm not supporting axis [-2, 3]

  #19642 opened 14 days ago by lllllllllaa
  2

• [Proposal] API tuning for use_cudnn in LSTM/other layers

  #19647 opened 18 days ago by LukeWood
  5

• ops.copy makes object (arrays only?) immutable

  #19665 opened 14 days ago by Mrutyunjay01
  1

• TypeError: Could not locate class 'adam'. Make sure custom classes are decorated with `@keras.saving.register_keras_serializable()`

  #19661 opened 14 days ago by mpetteno
  1

• Numeric diff of BatchNormalization layer when migrate from Keras2 to keras3

  #19649 opened 18 days ago by cliu-nuro
  5

• The torch backend for jnp.cholesky should use torch.linalg.cholesky, not torch.cholesky

  #19648 opened 18 days ago by jonbarron
  1

• LSTM layer with dropout does not use fast CuDNN implementation in Keras 3

  #19644 opened 18 days ago by giulatona
  1

• GSOC '24 Project?

  #19632 opened 21 days ago by Aryan-Deshpande
  1

• keras.ops.cross doesn't propagate input sizes

  #19631 opened 21 days ago by jonbarron
  1