CS5720 - Week 9
Slide 161 of 180

Deep Learning Framework Landscape

TensorFlow

Google Brain Team

  • Production-ready deployment
  • Keras high-level API
  • TensorFlow Lite for mobile
  • Extensive ecosystem

PyTorch

Facebook AI Research

  • Dynamic computation graphs
  • Pythonic and intuitive
  • Strong research community
  • TorchScript for production

JAX

Google Research

  • NumPy-compatible
  • XLA compilation
  • Functional programming
  • Automatic differentiation

MXNet

Apache Software Foundation

  • Hybrid programming
  • Multi-language support
  • Efficient memory usage
  • AWS integration

PaddlePaddle

Baidu

  • Industrial applications
  • Chinese NLP support
  • Model compression
  • Federated learning

Fast.ai

fast.ai

  • Built on PyTorch
  • Best practices built-in
  • Educational focus
  • High-level abstractions

Evolution of Deep Learning Frameworks

2007
Theano - First major deep learning framework
2015
TensorFlow released by Google
2016
PyTorch released by Facebook
2018
JAX emerges from Google Research
2020
PyTorch becomes research favorite
← Previous Next →
Prepared by Dr. Gorkem Kar