2 Releases published by 2 people

• v3.11.0 Keras 3.11.0

  published Jul 29, 2025

• v3.11.1 Keras 3.11.1

  published Jul 31, 2025

12 Pull requests merged by 8 people

• [OpenVINO backend] support repeat

  #21433 merged Aug 3, 2025

• Fix: Keep use_cudnn in nested RNNs in Bidirectional layer

  #21534 merged Aug 1, 2025

• Bump github/codeql-action from 3.29.2 to 3.29.5 in the github-actions group

  #21539 merged Aug 1, 2025

• Add opset_version in export_onnx.

  #21541 merged Aug 1, 2025

• Version bump 3.11.1

  #21535 merged Jul 31, 2025

• Implement isin function in keras.ops

  #21523 merged Jul 31, 2025

• Bug fix with serialization when walking through operations.

  #21532 merged Jul 30, 2025

• Update flax

  #21527 merged Jul 29, 2025

• [OpenVINO backend] fix openvino model exported names to match keras names

  #21526 merged Jul 29, 2025

• Fix a few typos in comments

  #21525 merged Jul 28, 2025

• Add VectorizedMap op class.

  #21516 merged Jul 28, 2025

• update python version

  #21517 merged Jul 28, 2025

6 Pull requests opened by 6 people

• [OpenVINO backend] Support numpy.deg2rad

  #21524 opened Jul 28, 2025

• TESTING ONLY (TPU test)

  #21528 opened Jul 29, 2025

• Implement isneginf function in keras.ops

  #21538 opened Aug 1, 2025

• Bump the python group with 6 updates

  #21540 opened Aug 1, 2025

• Add public register_call_context_args method to Layer

  #21543 opened Aug 1, 2025

• [OpenVINO backend] fix __getitem__ and convert_to_tensor issues

  #21545 opened Aug 2, 2025

3 Issues closed by 3 people

• How to specify the device on which the model resides

  #21537 closed Aug 4, 2025

• LossScaleOptimizer crash under jit_compile=True with MirroredStrategy and mixed_precision

  #21366 closed Jul 31, 2025

• TorchModuleWrapper serialization issue

  #20860 closed Jul 30, 2025

6 Issues opened by 5 people

• OpenVINOKerasTensor does not implement the full expected API of KerasTensor

  #21544 opened Aug 2, 2025

• Keras stuck on set_random_seed with PyTorch backend on Metal (Applied Silicon)

  #21542 opened Aug 1, 2025

• Multi-backend support for timeseries_dataset_from_array()

  #21536 opened Jul 31, 2025

• Bidirectional layer loses use_cudnn value in nested RNN layers

  #21533 opened Jul 31, 2025

• Custom call context args are not supported for the common methods for model calling (e.g. fit, predict, evaluate, ...)

  #21530 opened Jul 30, 2025

• Cannot load weight of pre-trained models

  #21529 opened Jul 30, 2025

19 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.

• Fix Discretization layer graph mode bug

  #21514 commented on Aug 4, 2025 ? 3 new comments

• [OpenVINO backend] Support numpy.logaddexp

  #21522 commented on Jul 28, 2025 ? 0 new comments

• docs: Keras 3 evaluate() and compiled metrics clarification

  #21506 commented on Jul 28, 2025 ? 0 new comments

• Simulated OpenVINO Backend for Testing Unmerged PR Features with Memory Profiling

  #21500 commented on Aug 3, 2025 ? 0 new comments

• Fix TypeError in keras.ops.searchsorted by using shape instead of len

  #21472 commented on Aug 1, 2025 ? 0 new comments

• It worth not hiding exceptions

  #21432 commented on Aug 4, 2025 ? 0 new comments

• Running TPU tests on linux-x86-ct6e-44-1tpu

  #21425 commented on Jul 29, 2025 ? 0 new comments

• [OpenVINO Backend] support ops.slice_update

  #21362 commented on Aug 3, 2025 ? 0 new comments

• Fix TorchModuleWrapper serialization issue

  #20869 commented on Jul 30, 2025 ? 0 new comments

• Outdated code

  #21092 commented on Aug 4, 2025 ? 0 new comments

• ??? Keras Development Roadmap

  #19519 commented on Aug 1, 2025 ? 0 new comments

• Specifying ONNX opset with export method

  #21511 commented on Aug 1, 2025 ? 0 new comments

• `Discretization` layer returns inconsistent result in graph mode.

  #21468 commented on Jul 31, 2025 ? 0 new comments

• Keras Fails to load quantized model

  #21378 commented on Jul 31, 2025 ? 0 new comments

• Model.quantize() increases memory allocated.

  #21518 commented on Jul 30, 2025 ? 0 new comments

• TypeError with single-output dictionary targets - "Expected float32, but got my_output_0 of type 'str'"

  #21513 commented on Jul 30, 2025 ? 0 new comments

• Mixed precision on AMD GPUs results in lower performance

  #21520 commented on Jul 29, 2025 ? 0 new comments

• NaN metrics when training under tf.distribute.MirroredStrategy

  #21467 commented on Jul 29, 2025 ? 0 new comments

• keras.layers.Flatten() not updating shape information after transformation

  #21502 commented on Jul 29, 2025 ? 0 new comments