Search results

🗓️ SIG Keras Community Meetings: Schedule, Invites, and Notes

AnnouncementOfficial updates from the Keras team.Official updates from the Keras team.

Status: Open.

#22410 In keras-team/keras;

· laxmareddyp opened on Mar 12, 2026

TextVectorization fails on G4 on Colab (all fine on H100, or A100 and earlier)

Status: Open.

#22406 In keras-team/keras;

· jchwenger opened on Mar 12, 2026

Incorrect Symbolic Output Shape for Discretization Layer with output_mode="count"/"multi_hot"

Status: Open.

#22403 In keras-team/keras;

· Isa-Fay opened on Mar 12, 2026

Torch backend cast+multiply masking pattern causes unnecessary memory allocations

type:Bug

Status: Open.

#22386 In keras-team/keras;

· MarcosAsh opened on Mar 8, 2026

CLIPBackbone text_encoder: predict_on_batch() output differs from tf.function on GPU (TF backend)

type:Bug

Status: Open.

#22380 In keras-team/keras;

· griffinstalha opened on Mar 7, 2026

Feature Request: Add Grain support for keras.layers.TextVectorization

type:featureThe user is asking for a new feature.The user is asking for a new feature.

Status: Open.

#22378 In keras-team/keras;

· alenowak opened on Mar 7, 2026

[Feature] Modernize benchmark infrastructure for multi-backend comparison

keras-team-review-pendingPending review by a Keras team member.Pending review by a Keras team member.

type:featureThe user is asking for a new feature.The user is asking for a new feature.

Status: Open.

#22307 In keras-team/keras;

· MarcosAsh opened on Feb 27, 2026

With torch backend jit_compile=True causes AssertionError

type:Bug

Status: Open.

#22282 In keras-team/keras;

· innat opened on Feb 25, 2026

[Feature] Add unified device utilities API (keras.utils.device_utils) for backend-agnostic device and memory management

keras-team-review-pendingPending review by a Keras team member.Pending review by a Keras team member.

type:featureThe user is asking for a new feature.The user is asking for a new feature.

Status: Open.

#22278 In keras-team/keras;

· amadhan882 opened on Feb 24, 2026

Torch backend: LSTM model training is terribly slow

backend:torch

keras-team-review-pendingPending review by a Keras team member.Pending review by a Keras team member.

type:bug/performance

Status: Open.

#22275 In keras-team/keras;

· DLumi opened on Feb 24, 2026

Torch backend: LSTM fails to use cuDNN when criteria are satisfied

type:Bug

Status: Open.

#22274 In keras-team/keras;

· DLumi opened on Feb 24, 2026

Passing a plain callable to add_weight(constraint=...) raises ValueError in Keras 3 (torch backend)

type:Bug

Status: Open.

#22221 In keras-team/keras;

· griffinstalha opened on Feb 19, 2026