Verified project record

keras-team/keras

A high-level deep learning framework that provides a unified API for building, training, and running models across JAX, TensorFlow, PyTorch, and OpenVINO backends. It is installable via standard Python package managers and designed to scale from local development to datacenter-scale training.

Project overview

The framework allows development and inference across multiple backends while avoiding framework lock-in.

Project type: Model Development · Model Runtime
Deployment: Refer to project documentation
License: Apache-2.0

Best for

Developers, AI engineers, and researchers who need to build, train, and run deep learning models across JAX, TensorFlow, PyTorch, and OpenVINO backends without framework lock-in.
Users who want to scale model training from a laptop to large clusters of GPUs or TPUs.

Key capabilities

Enables building, training, and running models interchangeably across JAX, TensorFlow, PyTorch, and OpenVINO backends.
Provides a high-level UX and easy-to-debug runtimes to ship deep learning solutions faster.
Scales model training from a laptop to large clusters of GPUs or TPUs.
Allows writing custom components (e.g., layers, models, metrics) that can be used in low-level workflows in any supported framework.

Limitations and risks

Windows users must use WSL2 to run the framework natively.
The OpenVINO backend is inference-only and does not support training.
The backend must be configured before importing the package and cannot be changed after the package has been imported.
Requires a clean Python environment for each backend to avoid CUDA version mismatches.

Getting started

Installation is considered easy and can be done via a standard Python package manager using pip. The first success path requires running 'pip install keras --upgrade' and installing backend package(s) such as tensorflow, jax, or torch.

Alternatives and comparisons

Provides an easy-to-use, unified API for state-of-the-art text, vision, audio, video, and multimodal models, lowering compute costs by sharing trained checkpoints and allowing movement across PyTorch/JAX/TF2.0 frameworks.
Provides a unified, efficient framework for zero-code fine-tuning of 100+ LLMs and VLMs via CLI and Web UI.
A web UI and code-based library that lets users efficiently run and train text, audio, embedding, and vision models locally with faster speeds and lower VRAM usage.

Project comparisons

Evidence and sources

README: Keras 3 is a multi-backend deep learning framework, with support for JAX, TensorFlow, PyTorch, and OpenVINO (for inference-only). Effortlessly build and train models for computer…
Release: v3.15.0
README: pip install keras --upgrade
README: To use `keras`, you should also install the backend of choice: `tensorflow`, `jax`, or `torch`. Additionally, The `openvino` backend is available with support for model inference…
README: Keras 3 is compatible with Linux and macOS systems. For Windows users, we recommend using WSL2 to run Keras. To install a local development version:

Reviewed at: 2026-07-17T05:50:57Z

Open original GitHub project