CVE-2022-41883 TensorFlow is an open source platform for machine learning that has been patched for an issue where the executor crashes when given different input sizes.

CVE-2022-41883 TensorFlow is an open source platform for machine learning that has been patched for an issue where the executor crashes when given different input sizes.

We have also released TensorFlow with 2.11 as a public beta. You can download it from tensorflow/tensorflow/releases. The public beta of TensorFlow 2.11 is available for Windows and Linux. TensorFlow 2.11 is not yet ready for production use, as it is still a work in progress. It is recommended that you stick with TensorFlow 2.10 for now. TensorFlow 2.11 is a significant release that introduces significant new features and improvements. Some of the most notable changes in TensorFlow 2.11 include: Improved accuracy on larger datasets TensorFlow can now run on larger datasets, up to 4 million images, up from 10 million images. This may result in better accuracy on larger datasets, because as the dataset size grows, the system learns to accommodate the increased complexity.
To enable TensorFlow to handle larger datasets, we have enabled the use of a larger memory allocation. TensorFlow now allocates up to 32GB of memory, which is 2x the previous limit of 16GB.

TensorFlow now supports the Theano library

TensorFlow now supports Theano, a library which allows TensorFlow to take advantage of GPUs. This should allow better performance on NVIDIA GPU and on CPU-only systems with NVIDIA GPUs.

TensorFlow Object Detection API

The TensorFlow Object Detection API is a newly introduced capability to implement object detection in TensorFlow. It is a single, unified interface to facilitate development of object detection models. As with any other object detector, you can train and deploy the models with TensorFlow Object Detection API on your own machine or on a cloud.
The following features are supported by the Object Detection API:
• Feature extraction from images
• Simple image classification
• Convolutional neural networks for image classification

TensorBoard 2.11

TensorBoard 2.11 lets you explore and visualize your data visually with a new interface that supports an integrated global search, the ability to filter on data types, and support for time-series graphs. For more information about TensorFlow 2.11, please see the release notes at tensorflow/tensorflow/releases/tag/v2.11

TensorFlow Serving Support

TensorFlow has also been updated to use a more efficient implementation of the Inception V3 model. It now uses an updated architecture, which uses less memory and faster execution on larger datasets.
The new TensorFlow Serving Support enables downstream TensorFlow models to run on a GPU server without having to migrate data back and forth. This is important because it allows the models to remain running while they are learning, instead of having to shut down when they start training new layers.

TensorFlow with TensorDB

TensorFlow now supports TensorDB, a new data store that contains all the data and computation graphs.
One of the main features of TensorFlow 2.11 is support for TensorDB, which is a new way to store tensorflow graph in an embedded database. This means that you can use your own database or create your own datastore with graph-like semantics. You can run queries on those graphs to perform complex operations like joining more than one graph at once or performing row-level joins.


Subscribe to
Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.