Deep Learning Frameworks

Back to Training at Scale

The software frameworks used to build, train, and deploy neural networks.

Key Frameworks

  • PyTorch — dynamic computation graphs, Pythonic, dominant in research and increasingly in production
  • JAX — NumPy on accelerators, functional transformations (vmap, jit, grad), used at Google/DeepMind
  • TensorFlow — static graphs (eager mode available), TensorFlow Serving, production-focused
  • Keras — high-level API, now multi-backend (TF, JAX, PyTorch)

deep-learning frameworks pytorch jax tensorflow