Choose when you need a mature, low-level framework with a full ecosystem and production tooling; heavier than Keras.
An Open Source Machine Learning Framework for Everyone
- 198.0k
- C++
- Apache-2.0
deep learning framework
Keras 3 is a multi-backend deep learning framework for building and training neural networks with a high-level API, supporting JAX, TensorFlow, PyTorch, and OpenVINO backends.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Choose when you need a mature, low-level framework with a full ecosystem and production tooling; heavier than Keras.
An Open Source Machine Learning Framework for Everyone
Choose when you need dynamic, imperative programming with strong research community and lower-level control.
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Choose for Rust projects that need a modern deep learning framework with eager execution and portability.
Burn is a next generation tensor library and Deep Learning Framework that doesn't compromise on flexibility, efficiency and portability.
Choose for large-scale distributed training with an alternative parallel architecture; less mature and smaller community.
OneFlow is a deep learning framework designed to be user-friendly, scalable and efficient.
Choose for C++ projects needing a standalone ML library with high performance.
A C++ standalone library for machine learning
Choose for JVM-based projects needing an engine-agnostic DL API.
An Engine-Agnostic Deep Learning Framework in Java
Choose for Julia projects wanting a differentiable programming DL framework.
Relax! Flux is the ML library that doesn't make you tensor
Choose when working directly in JAX and wanting a light, functional NN library instead of Keras's full API.
JAX-based neural network library
Choose for Nim projects needing a tensor library with deep learning support and embedded/GPU targets.
A fast, ergonomic and portable tensor library in Nim with a deep learning focus for CPU, GPU and embedded devices via OpenMP, Cuda and OpenCL backends
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Choose for Apple's optimized CV models and training recipes, but it's narrower than Keras.
CoreNet: A library for training deep neural networks
An older DL framework; choose if you need its distributed training features but note it's been stale for over a year.
MegEngine 是一个快速、可拓展、易于使用且支持自动求导的深度学习框架
Rust DL framework with compile-time shape checking; unmaintained since 2.1y.
Deep learning in Rust, with shape checked tensors and neural networks
An older C# DL library; archived and unusable for production today.
An easy to use C# deep learning library with CUDA/OpenCL support
An older Rust ML library with multi-backend support; unmaintained.
Powerful Powerful Machine Learning library with GPU, CPU and WASM backends
A previous-generation high-level training interface on TensorFlow; choose for legacy TF code.
A Neural Net Training Interface on TensorFlow, with focus on speed + flexibility
These projects were analysed and named keras among their alternatives. The relationship is not symmetric — how keras rates them is a separate judgement, made when keras is analysed in its own right.
calls keras “when you want the high-level Keras API directly in Python (on top of TensorFlow) rather than TF.js.”
A WebGL accelerated JavaScript library for training and deploying ML models.
calls keras “When you want a backend-agnostic high-level NN API instead of CoreNet's task-specific model zoo.”
CoreNet: A library for training deep neural networks