TensorFlow Release 17.10

The NVIDIA container image of TensorFlow, release 17.10, is available.

TensorFlow container image version 17.10 is based on TensorFlow 1.3.0.

Contents of TensorFlow

This container image contains the complete source of the version of NVIDIA TensorFlow in /opt/tensorflow. It is pre-built and installed into the /usr/local/[bin,lib] directories in the container image.

To achieve optimum TensorFlow performance, for image based training, the container includes a sample script that demonstrates efficient training of convolutional neural networks (CNNs). The sample script may need to be modified to fit your application. The container also includes the following:

• Ubuntu 16.04
• NVIDIA CUDA^® 9.0
• NVIDIA CUDA^® Deep Neural Network library™ (cuDNN) 7.0.3
• NVIDIA^® Collective Communications Library ™ (NCCL) 2.0.5 (optimized for NVLink™ )

Driver Requirements

Release 17.10 is based on CUDA 9, which requires NVIDIA Driver release 384.xx.

Key Features and Enhancements

This TensorFlow release includes the following key features and enhancements.

• Added PNG image support to nvcnn.py.
• Fixed issue with batchnorm op that broke backwards compatibility in the previous release.
• Renamed the TF_ENABLE_TENSOR_OP_MATH (default=1) environment variable to TF_DISABLE_TENSOR_OP_MATH (default=0).
• Upgraded Bazel to version 0.5.4.
• Worked around hash mismatches in third-party source downloads.
• Enabled compilation flags -march=sandybridge -mtune=broadwell.
• Updated Eigen to the top of the tree and removed custom patches.
• Latest version of CUDA
• Latest version of cuDNN
• Latest version of NCCL
• Ubuntu 16.04 with September 2017 updates

Known Issues

There are no known issues in this release.