We are very happy to share a new message with you! TensorFlow version 1.10.0 has been officially released! Please read this article to learn about some main features and functional improvements, as well as some breakthrough changes!
â–ŒMajor changes
TensorFlow 1.10 pre-built binaries are built for NCCL 2.2, and NCCL is no longer included in the binary installation. The use of multiple GPUs and NCCL needs to upgrade NCCL to 2.2. For details, please refer to the update installation guide: Installing TensorFlow on Ubuntu and Install TensorFlow from Sources.
Starting from TensorFlow 1.11, Windows builds will use Bazel. Therefore, the official gave up support for cmake.
â–ŒMain features and improvements
tf.lite runtime supports complex64
tf.data integrates Bigtable by default
tf.estimator.train_and_evaluate can be evaluated without reloading the checkpoint
RunConfig supports restricting the communication between workers and PS by setting device_filters. In some cases, it can speed up training and simplify the cleaning of downtime equipment. If there are tasks that need to communicate between workers, you can customize session_options in RunConfig
Moved Distribution and Bijectors from tf.contrib.distributions to Tensorflow Probability (TFP). tf.contrib.distributions has been deprecated and will be removed before the end of 2018
Many new APIs have been added. It is recommended to use these new APIs, which may replace some existing APIs in the future. The new api is mainly added to the following modules: tf.debugging, tf.dtypes, tf.image, tf.io, tf.linalg, tf.manip, tf.math, tf.quantization, tf.strings
The release of this version is not only the effort of Google engineers, but also includes the contributions of many other developers:
Ag Ramesh, Alex Wiltschko, Alexander Pantyukhin, Amogh Mannekote, An Jiaoyang, Andrei Nigmatulin...
China leading manufacturers and suppliers of Single-Element Detector
Single-Element Detector,Hortwave Detector Unit,Swir Infrared Detector,Swir Ingaas Unit Detector
Ningbo NaXin Perception Intelligent Technology CO., Ltd. , https://www.nicswir.com