Rare find

KerasHub. Ready-to-use AI models for text, images, and audio that work out of the box.

keras.io/keras_hub

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

July 2026
  • Master Version bump to 0.31.0.dev0 (#2859)
  • Release —v0.30.0
  • Release —v0.30.0.dev0
  • Switch CI to self-hosted runners (#2856)
  • Add Magistral Presets to Hub (#2845)
  • Support explicit head_dim on MistralBackbone (enables Magistral) (#2692)
  • Fix Zizmor Warnings (#2857)
  • Update keras version to 3.15 (#2852)
  • Fix Gemma3/3n/4 preprocessor crashes with Python-only tokenizer (#2773)
  • Enable LiteRT export tests on the torch backend (#2698)
  • Fix attention score returns in CachedMultiHeadAttention (#2799)
  • Register BLIP2 model presets to hub (#2805)
  • Add blip2 (#2699)
  • Bump the github-actions group with 2 updates (#2803)
  • Fix Zizmor security findings in GitHub Actions workflows (#2792)
  • Sync kaggle models to HF (#2802)
  • Increase Kokoro GPU timeout to 240 minutes for JAX, TensorFlow, and …
  • Enable python workflows for `ImageConverter`, `MaskedLMMaskGenerator`…
  • Register Qwen 3.5 MoE Model Presets to Hub (#2798)
  • Pin sentencepiece<0.2.2 to unblock CI test failures (#2800)
  • Add Microsoft Harrier Model to Hub (#2764)
  • Add BGE Embedding Model to Hub (#2755)
  • Add Qwen3.5 MoE Model to Hub (#2665)
  • Update Mixtral converter tests and lazy load sentencepiece in T5Gemma…
  • Update the kagglesdk version for gpu compatibility (#2788)
  • Pin kagglesdk==0.1.30 and surface kagglehub import errors (#2787)
  • Bump the github-actions group with 2 updates (#2786)
June 2026
  • test: add FileNotFoundError test for get_file preset utility (#2768)
  • Standardize rope_theta extraction and enhance converter tests (#2733)
  • Fix CLIPTokenizer inconsistency between Python and TF paths (#2732)