Keras Hub Files

Summary:

New Models:

We've integrated a range of cutting-edge models, each designed to tackle specific challenges in their respective domains:

• Gemma 3 270M: Released Gemma 3 270M parameter model and instruction tuned, 18-layer, text-only model designed for hyper-efficient AI, particularly for task-specific fine-tuning.

• Qwen3: A powerful, large-scale multilingual language model, excelling in various natural language processing tasks, from text generation to complex reasoning.

• DeiT: Data-efficient Image Transformers (DeiT), specifically designed to train Vision Transformers effectively with less data, making high-performance visual models more accessible.

• HGNetV2: An advanced version of the Hybrid-Grouped Network, known for its efficient architecture in computer vision tasks, particularly optimized for performance on diverse hardware.

• DINOV2: A state-of-the-art Self-Supervised Vision Transformer, enabling the learning of robust visual representations without relying on explicit labels, ideal for foundation models.

• ESM & ESM2: Evolutionary Scale Modeling (ESM & ESM2), powerful protein language models used for understanding protein sequences and structures, with ESM2 offering improved capabilities for bioinformatics research.

Improvements & Enhancements

This update also includes several key improvements to enhance the platform's stability, compatibility, and flexibility:

• Python 3.10 Minimum Support: Updated the minimum supported Python version to 3.10, ensuring compatibility with the latest libraries and features.
• Gemma Conversion (Keras to SafeTensors): Added a new conversion script to effortlessly convert Gemma models from Keras format to Hugging Face's Safetensor format.
• Gemma3 Conversion Script: Added conversion script for Gemma3 models, streamlining their integration into the Hugging Face ecosystem.
• ViT Non-Square Image Support: Enhanced the Vision Transformer (ViT) model to now accept non-square images as input, providing greater flexibility for various computer vision applications.
• LLM Left Padding Method: Added support for left padding in our LLM padding methods, offering more control and compatibility for specific model architectures and inference requirements.

What's Changed

Complete list of all the changes included in this release. * register presets by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2268 * Fix batch preprocessing bug in Moonshine generation by @harshaljanjani in https://github.com/keras-team/keras-hub/pull/2266 * fix get_lora_target_names function by @divyashreepathihalli in https://github.com/keras-team/keras-hub/pull/2167 * implement of leftpadding by @pass-lin in https://github.com/keras-team/keras-hub/pull/2242 * make vit compatible with non square images by @sineeli in https://github.com/keras-team/keras-hub/pull/2255 * Bump up master version to 0.22.0.dev0 by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2277 * Fix keras-io integration test by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2280 * Add Qwen3 by @kanpuriyanawab in https://github.com/keras-team/keras-hub/pull/2249 * Add DeiT Model by @Sohaib-Ahmed21 in https://github.com/keras-team/keras-hub/pull/2203 * [HOTFIX] Add Docstring for QwenCausalLM by @kanpuriyanawab in https://github.com/keras-team/keras-hub/pull/2279 * Fix: Correct coverage tracking for keras_hub by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2283 * Update the sharded version number for Llama3 variants by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2294 * Support None for max_shard_size by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2261 * Sharded weights type error by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2296 * Fix PaliGemmaCausalLM example. by @hertschuh in https://github.com/keras-team/keras-hub/pull/2302 * Routine HF sync by @divyashreepathihalli in https://github.com/keras-team/keras-hub/pull/2303 * Incorrect condition on sliding_window_size by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2289 * Bump the python group with 2 updates by @dependabot[bot] in https://github.com/keras-team/keras-hub/pull/2282 * Modify TransformerEncoder masking documentation by @sonali-kumari1 in https://github.com/keras-team/keras-hub/pull/2297 * Fix Gemma3InterleaveEmbeddings JAX inference error by ensuring indices are int32 by @pctablet505 in https://github.com/keras-team/keras-hub/pull/2305 * Update preset versions for Mixtral,Qwen-MoE and Mistral models by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2307 * Fix Mistral conversion script by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2306 * Bump the python group with 6 updates by @dependabot[bot] in https://github.com/keras-team/keras-hub/pull/2317 * Qwen3 causal lm by @kanpuriyanawab in https://github.com/keras-team/keras-hub/pull/2311 * Fix JAX GPU tests by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2319 * support flash-attn at torch backend by @pass-lin in https://github.com/keras-team/keras-hub/pull/2257 * Add HGNetV2 to KerasHub by @harshaljanjani in https://github.com/keras-team/keras-hub/pull/2293 * Qwen3 presets register by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2325 * Register HGNetV2 presets by @laxmareddyp in https://github.com/keras-team/keras-hub/pull/2326 * Safetensors conversion by @Bond099 in https://github.com/keras-team/keras-hub/pull/2290 * Add DINOV2. by @james77777778 in https://github.com/keras-team/keras-hub/pull/2328 * Refactor CLIP and update SD3. by @james77777778 in https://github.com/keras-team/keras-hub/pull/2316 * add DINOv2 preset details by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2336 * Fix dtype issues on JAX CPU in SD3 tests. by @james77777778 in https://github.com/keras-team/keras-hub/pull/2338 * Revert "Fix dtype issues of JAX CPU in SD3. (#2338)" by @divyashreepathihalli in https://github.com/keras-team/keras-hub/pull/2344 * Resolve preset comparison bug in glue load model method by @emmanuel-ferdman in https://github.com/keras-team/keras-hub/pull/2345 * Removes unnecessary call to torch.no_grad() by @JyotinderSingh in https://github.com/keras-team/keras-hub/pull/2353 * Add Esm by @pass-lin in https://github.com/keras-team/keras-hub/pull/2244 * Fix float16 issue in SD3 when using JAX CPU. by @james77777778 in https://github.com/keras-team/keras-hub/pull/2354 * update python to 3.10 and Keras minimum version to 3.8 by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2292 * register DeiT presets by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2348 * Fix path for presets to link it to API docs in keras.io by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2357 * Fix for llama3.1 instruct models by @pctablet505 in https://github.com/keras-team/keras-hub/pull/2355 * Add & register ESM presets by @sachinprasadhs in https://github.com/keras-team/keras-hub/pull/2356 * Add Gemma 3 conversion script by @abheesht17 in https://github.com/keras-team/keras-hub/pull/2358 * Remove exact matching of outputs from Gemma 3 conversion notebook by @abheesht17 in https://github.com/keras-team/keras-hub/pull/2359

New Contributors

• @Sohaib-Ahmed21 made their first contribution in https://github.com/keras-team/keras-hub/pull/2203
• @sonali-kumari1 made their first contribution in https://github.com/keras-team/keras-hub/pull/2297
• @Bond099 made their first contribution in https://github.com/keras-team/keras-hub/pull/2290
• @emmanuel-ferdman made their first contribution in https://github.com/keras-team/keras-hub/pull/2345

Full Changelog: https://github.com/keras-team/keras-hub/compare/v0.21.1...v0.22.0

For detailed documentation and usage examples/guides, please refer to our updated guides on https://keras.io/keras_hub/