Tensorflow 2.0 (github.com)

Article summary

TensorFlow 2.0 has been released, focusing on simplicity and ease of use, with updates such as easy model building with Keras and eager execution, robust model deployment, and powerful experimentation for research. The new version also features API simplification, reducing duplication and removing deprecated endpoints. TensorFlow 2.0 delivers Keras as the central high-level API used to build and train models. The release includes several breaking changes, including the removal of tf.contrib and the deprecation of certain APIs.

Main themes

• TensorFlow 2.0
• Keras API
• API Simplification
• Breaking Changes
• Machine Learning
• Deep Learning

What commenters say

• The transition to TensorFlow 2.0 will be difficult for many users due to the numerous breaking changes and the need to update existing code. Some users prefer PyTorch over TensorFlow due to its ease of use and simplicity. TensorFlow's complexity and global flags can cause annoying bugs and make it difficult to work with third-party libraries. Jax is a potential alternative to TensorFlow, offering a simpler and more efficient API. The serving story for PyTorch models is currently lacking, making TensorFlow a better choice for production environments. Anaconda is a useful tool for managing package versions and dependencies, but some users prefer using venv or Docker for creating hermetic environments. The installation of TensorFlow on Windows can be problematic, especially when trying to get GPU support working.