{"full_name":"keras-team/keras-hub","owner":"keras-team","name":"keras-hub","shard":"1","url":"https://github.com/keras-team/keras-hub","description":"Pretrained model hub for Keras 3.","homepage":"https://keras.io/keras_hub/","topics":["deep-learning","keras","machine-learning","nlp","tensorflow","jax","llm","natural-language-processing","python","pytorch","cv"],"tag_pages":["cv","deep-learning","jax","keras","llm","machine-learning","natural-language-processing","nlp","python","pytorch","tensorflow"],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/34455048?v=4","created_at":"2020-05-28T23:03:54Z","pushed_at":"2026-11-16T03:03:36Z","metrics":{"stars":999,"forks":380,"watchers":29,"issues":1055,"prs":2144,"releases":83,"issues_open":169,"issues_closed":886,"prs_open":54,"prs_closed":408,"prs_merged":1682,"discussions":56},"growth":{"day":{"stars":0,"forks":1,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[998,998,998,999,999,999,999],"forks":[377,378,378,378,378,379,380],"watchers":[null,29,29,29,29,29,29],"issues_open":[null,169,167,167,168,168,169],"prs_open":[null,55,53,52,54,54,54]},"boards":{},"same_owner":[{"full_name":"keras-team/autokeras","name":"autokeras","shard":"1","boards":{}},{"full_name":"keras-team/keras","name":"keras","shard":"1","boards":{}}],"issues":[{"number":2101,"title":"KerasHub has mandatory dependency on tensorflow-text but it isn't mandatory","comments":16,"created_at":"2025-02-14T12:49:31Z"},{"number":1835,"title":"🚀 Contributing to KerasHub 🚀 ","comments":14,"created_at":"2024-09-17T17:00:31Z"},{"number":1030,"title":"Improve samplers documentation","comments":14,"created_at":"2023-05-11T11:38:48Z"},{"number":1630,"title":"Distributed training not working (batch size calculation)","comments":12,"created_at":"2024-05-16T15:29:01Z"},{"number":2392,"title":"Add Qwen Image & Qwen Image Edit models","comments":10,"created_at":"2025-09-08T18:17:10Z"},{"number":284,"title":"`char_to_token` in `keras_nlp.tokenizers.Tokenizer`","comments":9,"created_at":"2022-08-07T00:29:42Z"},{"number":2072,"title":"Add DoRA support in KerasHub","comments":8,"created_at":"2025-02-04T00:20:13Z"},{"number":2669,"title":"CONTRIBUTING_MODELS.md is outdated","comments":7,"created_at":"2026-04-11T05:35:15Z"}],"discussions":[{"number":2347,"title":"Plan for 3D variants of ImageNet models","comments":0,"category":"Q&A","created_at":"2025-07-27T09:20:06Z"},{"number":2291,"title":"Question regarding CNN usage for object detection.","comments":0,"category":"General","created_at":"2025-06-06T12:58:39Z"},{"number":2120,"title":"Interest in Contributing to KerasHub for GSoC 2025","comments":0,"category":"Q&A","created_at":"2025-02-28T02:51:43Z"},{"number":2119,"title":"Guidance on how to contribute (and GSOC 2025)","comments":0,"category":"Q&A","created_at":"2025-02-27T19:31:21Z"},{"number":2094,"title":"How to use custom input with backbones?","comments":0,"category":"Q&A","created_at":"2025-02-12T15:48:52Z"},{"number":1995,"title":"keras_cv.losses.IoULoss documentation not found","comments":0,"category":"Q&A","created_at":"2024-11-28T13:30:06Z"},{"number":1987,"title":"Getting error when trying to export GemmaCausalLM model to SavedModel format","comments":0,"category":"Q&A","created_at":"2024-11-18T17:30:58Z"},{"number":1708,"title":"AttributeError: module 'keras.saving' has no attribute 'register_keras_serializable'","comments":0,"category":"Q&A","created_at":"2024-07-25T09:45:13Z"}],"latest_release":{"name":"v0.33.0","tag":"v0.33.0","published_at":"2026-11-03T21:51:15Z","body":"# KerasHub v0.33.0 Release Notes\r\n\r\n<p align=\"center\">\r\n  <a href=\"https://pypi.org/project/keras-hub/0.33.0/\"><img alt=\"PyPI\" src=\"https://img.shields.io/pypi/v/keras-hub?label=keras-hub&color=d00000\"></a>\r\n  <a href=\"https://pypistats.org/packages/keras-hub\"><img alt=\"Downloads\" src=\"https://img.shields.io/pypi/dm/keras-hub?color=d00000\"></a>\r\n  <a href=\"https://pypi.org/project/keras-hub/\"><img alt=\"Python\" src=\"https://img.shields.io/pypi/pyversions/keras-hub\"></a>\r\n  <a href=\"https://github.com/keras-team/keras-hub/compare/v0.32.0...v0.33.0\"><img alt=\"Diff\" src=\"https://img.shields.io/badge/diff-v0.32.0...v0.33.0-blue\"></a>\r\n</p>\r\n\r\n<table align=\"center\" width=\"100%\">\r\n  <tr>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\"><h2>12</h2><sub>NEW MODEL FAMILIES</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\"><h2>29</h2><sub>NEW PRESETS</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\"><h2>52</h2><sub>PULL REQUESTS</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\"><h2>12</h2><sub>CONTRIBUTORS</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\"><h2>6</h2><sub>FIRST-TIME CONTRIBUTORS</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\"><h2>3</h2><sub>BACKENDS</sub></td>\r\n  </tr>\r\n</table>\r\n\r\n<table align=\"center\" width=\"100%\">\r\n  <tr>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🧱<br><b>ModernBERT</b><br><sub>8K-context encoder</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🧭<br><b>ModernBERT Embed</b><br><sub>sentence embeddings</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">📐<br><b>MPNet</b><br><sub>sentence embeddings</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">💎<br><b>EmbeddingGemma 2</b><br><sub>text · image · video · audio</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🌀<br><b>Diffusion Gemma</b><br><sub>26B MoE block diffusion</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🖼️<br><b>Mistral Small 3.x</b><br><sub>24B vision-language</sub></td>\r\n  </tr>\r\n  <tr>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">📱<br><b>Ministral 3</b><br><sub>3B · 8B · 14B edge VLMs</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🛠️<br><b>Devstral Small 2</b><br><sub>agentic coding</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🧠<br><b>Magistral Small</b><br><sub>reasoning</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🛡️<br><b>Shieldstral</b><br><sub>safety classifier</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">🎞️<br><b>SmolVLM2</b><br><sub>256M–2.2B image &amp; video</sub></td>\r\n    <td align=\"center\" valign=\"top\" width=\"16%\">✨<br><b>Muse Glimmer</b><br><sub>30B VLM + speculative drafter</sub></td>\r\n  </tr>\r\n</table>\r\n\r\n## Highlights\r\n\r\n### 🧩 New Models\r\n\r\nThis release is the largest model drop since the KerasHub rename: twelve new model families spanning text encoders, embedding models, vision-language models, and a new diffusion-based text generation paradigm.\r\n\r\n- **ModernBERT**: A modern BERT-style encoder with RoPE, alternating local/global attention and an 8K context window, for masked language modeling and text classification. ([#2518](https://github.com/keras-team/keras-hub/pull/2518))\r\n- **ModernBERT Text Embedder**: Sentence embedding models on the ModernBERT encoder, producing 768-dimensional vectors for retrieval and semantic search, with the Nomic and Alibaba GTE fine-tunes. ([#3150](https://github.com/keras-team/keras-hub/pull/3150))\r\n- **MPNet**: The sentence-transformers family of 768-dimensional sentence embedding models for semantic search, clustering and paraphrase mining. ([#3181](https://github.com/keras-team/keras-hub/pull/3181))\r\n- **EmbeddingGemma 2**: A 744M-parameter multimodal embedding model built on Gemma 4 that maps text, images, video and audio into a shared 768-dimensional space. ([#3144](https://github.com/keras-team/keras-hub/pull/3144))\r\n- **Diffusion Gemma**: A 26B-parameter MoE (3.8B active) instruction-tuned language model that gener","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}