8 Pull requests merged by 5 people

• PyDataset now implements __iter__.

  #21330 merged May 28, 2025

• Revert "Fixed issue with dot_product_attention when using TPU. "

  #21329 merged May 27, 2025

• add ops.layer_normalization

  #21327 merged May 27, 2025

• api_gen now excludes backend code.

  #21321 merged May 27, 2025

• improve RMSNorm preformance when torch backend

  #21325 merged May 26, 2025

• Updated random_grayscale.py compute_output_spec

  #21326 merged May 26, 2025

• Implement hanning function in keras.ops

  #21318 merged May 22, 2025

• Modified confusion_metrics.py

  #21315 merged May 22, 2025

3 Pull requests opened by 2 people

• [OpenVINO backend] Update `ndim` behavior

  #21317 opened May 22, 2025

• Fix progbar bug

  #21331 opened May 28, 2025

• fix csv logger bug

  #21332 opened May 28, 2025

9 Issues closed by 4 people

• User guide to use Rematerialization

  #20964 closed May 29, 2025

• keras.ops.print: `tf.print`; `jax.debug.print`; …

  #21137 closed May 29, 2025

• possible bug in `keras.utils.Sequence`

  #21151 closed May 28, 2025

• Error loading history file (tf/keras 2.15)

  #20342 closed May 28, 2025

• Significant performance regression

  #21316 closed May 24, 2025

• Keras `__init__.py` structure isn't readable by a static type checker

  #19779 closed May 24, 2025

• Sequential errors for multiple batch dimensions, even when all inner layers are compatible

  #21158 closed May 24, 2025

• ERROR: pip's dependency resolver does not currently take into account all the packages that are installed

  #21162 closed May 24, 2025

• _int8_build() bug from keras-nightly

  #21272 closed May 22, 2025

4 Issues opened by 4 people

• Type of model.input is KerasTensor in 3.9 and list in 3.10

  #21328 opened May 27, 2025

• Documentation update guide required

  #21322 opened May 23, 2025

• SpectralNormalization layer fails on multiple batch dimensions

  #21320 opened May 22, 2025

• Inconsistent broadcasting behaviour in layers.Rescaling

  #21319 opened May 22, 2025

38 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Add memory usage monitor callback

  #21245 commented on May 29, 2025 • 5 new comments

• [OpenVINO Backend] support ops.where

  #21295 commented on May 26, 2025 • 1 new comment

• TF+XLA+Mixed Precision: Keras fails to compute gradients

  #20931 commented on May 27, 2025 • 0 new comments

• loss_weights depending on epoch number

  #20294 commented on May 27, 2025 • 0 new comments

• Is this Bug?

  #20984 commented on May 27, 2025 • 0 new comments

• Inconsistencies with the behavior of bias initializers, leading to poor performance in some cases

  #20873 commented on May 28, 2025 • 0 new comments

• Add `on_epoch_begin` to `utils.Sequence`

  #19749 commented on May 28, 2025 • 0 new comments

• Distributed training with tensorflow doesn't work with Keras 3.7 and above

  #21172 commented on May 28, 2025 • 0 new comments

• Inconsistent execution results by the PyTorch backend

  #20235 commented on May 28, 2025 • 0 new comments

• layers.GRU returns wrong shaped output with GPU

  #20173 commented on May 28, 2025 • 0 new comments

• Progress bar showing wrong step number

  #20861 commented on May 28, 2025 • 0 new comments

• ValueError: File not found: filepath=Backend\C4__256g_000040000.keras. Please ensure the file is an accessible `.keras` zip file.

  #20259 commented on May 28, 2025 • 0 new comments

• Enormous memory usage after batched forward passes with TensorFlow 2.16.1 (CPU)

  #19500 commented on May 28, 2025 • 0 new comments

• MLX Backend

  #19571 commented on May 29, 2025 • 0 new comments

• OpenVINO: linspace

  #21209 commented on May 26, 2025 • 0 new comments

• Fix stateful RNN to raise ValueError on batch size mismatch

  #21249 commented on May 22, 2025 • 0 new comments

• Keras <> NNX integration

  #21252 commented on May 28, 2025 • 0 new comments

• [Keras 3 OpenVINO Backend]: Support numpy.logspace operation #30114

  #21253 commented on May 26, 2025 • 0 new comments

• numpy.meshgrid implementation

  #21292 commented on May 25, 2025 • 0 new comments

• Cannot generate heatmaps using Gradcam library

  #20475 commented on May 22, 2025 • 0 new comments

• keras.layers.Normalization#adapt() does not support keras.utils.PyDataset as input.

  #21300 commented on May 22, 2025 • 0 new comments

• sample_weight with multiple, named losses

  #20992 commented on May 23, 2025 • 0 new comments

• Error: ``model`` must be a callable, a Keras Model instance or None

  #20893 commented on May 23, 2025 • 0 new comments

• Issue error you're encountering stems from an issue with using masking in the Embedding layer for POS Tagging Model

  #20773 commented on May 23, 2025 • 0 new comments

• Attention layer does not accept output of previous layers in functional API

  #20318 commented on May 23, 2025 • 0 new comments

• Potential bug in legacy h5 weights loading.

  #19694 commented on May 23, 2025 • 0 new comments

• Sharing weights across layers in keras 3 [feature request]

  #18821 commented on May 23, 2025 • 0 new comments

• Keras predict/predict_on_batch giving different answers than predict_step/__call__()

  #19403 commented on May 23, 2025 • 0 new comments

• Special optimization for torch backend

  #21313 commented on May 23, 2025 • 0 new comments

• StringLookup does not work on torch.Tensor

  #21255 commented on May 23, 2025 • 0 new comments

• Alias `KerasTensor` as `Tensor`

  #21210 commented on May 23, 2025 • 0 new comments

• Inconsistent behaviours between backends on ops.tile

  #20914 commented on May 23, 2025 • 0 new comments

• Error after upgrade from 3.6 to 3.7

  #20686 commented on May 23, 2025 • 0 new comments

• Stateful metrics produce incorrect output in distributed context

  #21054 commented on May 23, 2025 • 0 new comments

• SavedModel format doesn't support masking in Keras 3

  #21133 commented on May 24, 2025 • 0 new comments

• sklearn wrappers save checkpoints without required attributes

  #21143 commented on May 27, 2025 • 0 new comments

• This method creates a model with a 100% memory leak loop using model. fit()

  #20552 commented on May 27, 2025 • 0 new comments

• Evaluation of Time Series Forecasting Discrepancies in the Keras Example

  #18597 commented on May 27, 2025 • 0 new comments