TensorFlow 2.4.0 Now Available

TensorFlow is an end-to-end open source platform for machine learning. It has a comprehensive, flexible ecosystem of tools, libraries, and community resources that lets researchers push the state-of-the-art in ML and developers easily build and deploy ML-powered applications.

The newest version of TensorFlow brings major features, improvements, bug fixes and other changes.

Major Features and Improvements

• tf.distribute introduces experimental support for asynchronous training of models via the tf.distribute.experimental.ParameterServerStrategy API. See the tutorial to learn more.
• MultiWorkerMirroredStrategy is now a stable API and is no longer considered experimental. Some of the major improvements involve handling peer failure and many bug fixes. Check out the detailed tutorial on Multi-worker training with Keras.
• Introduces experimental support for a new module named tf.experimental.numpy which is a NumPy-compatible API for writing TF programs. See the detailed guide.
• Adds support for TensorFloat-32 on Ampere based GPUs. TensorFloat-32, or TF32 for short, is a math mode for NVIDIA Ampere based GPUs and is enabled by default.
• A major refactoring of the internals of the Keras Functional API has been completed, that should improve the reliability, stability, and performance of constructing Functional models.
• Keras mixed precision API tf.keras.mixed_precision is no longer experimental and allows the use of 16-bit floating point formats during training, improving performance by up to 3x on GPUs and 60% on TPUs. Please see below for additional details.
• TensorFlow Profiler now supports profiling MultiWorkerMirroredStrategy and tracing multiple workers using the sampling mode API.
• TFLite Profiler for Android is available. See the detailed guide.
• TensorFlow pip packages are now built with CUDA11 and cuDNN 8.0.2.

Download TensorFlow 2.4.0 on the GitHub page.

Sources: Press materials received from the company and additional information gleaned from the company’s website.