September 30, 2019
TensorFlow 2.0 released
Google released TensorFlow 2.0 on September 30, 2019 — making Keras the default APIApplication programming interface — a defined way for programs to talk to each other or to a service. and eager execution the default mode.
What it was for
TensorFlow 2.0 dropped the 1.x graph/session model for imperative `tf.function` training — simpler notebooks and production pipelines. It unified the fragmented TF ecosystem and responded to PyTorch's research dominance with a more approachable developer experience.
Companies
Why it's here
TensorFlow 2.0 was Google's answer to usability complaints that drove researchers to PyTorch.
Why it mattered
Keras integration and eager execution made TF competitive for interactive ML development.
What it solved
TensorFlow 1.x's graph boilerplate was too complex for most practitioners.
Media
- ImageTensorFlow
Google LLC, Public domain, via Wikimedia Commons
Related
- TensorFlow open sourcedNovember 9, 2015
- Bazel open-sourcedMarch 24, 2015