We detailed the most common TensorFlow 1 concepts that were deprecated in the new version. Many smaller modules and paradigms were also redesigned in TensorFlow 2. When migrating a project, we recommend having a thorough look at the documentation of both versions. To ensure that a migration went well and the TensorFlow 2 version works as expected, we recommend that you log both inference metrics (such as latency, accuracy, or average precision) and training metrics (such as the number of iterations before convergence), and compare their values between the old and new versions.
As it is open source and backed by an active community, TensorFlow is constantly evolving—integrating new features, optimizing others, improving the developer experience, and more. While this may sometimes require some additional effort, upgrading to the latest version as soon as possible will provide you with the best environment to develop more performant recognition applications.
