TensorFlow Release 19.02

The NVIDIA container image of TensorFlow, release 19.02, is available.

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 as a system Python module.

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
  Note:

  Container image 19.02-py2 contains Python 2.7; 19.02-py3 contains Python 3.5.

• NVIDIA CUDA 10.0.130 including CUDA^® Basic Linear Algebra Subroutines library™ (cuBLAS) 10.0.130
• NVIDIA CUDA^® Deep Neural Network library™ (cuDNN) 7.4.2
• NVIDIA Collective Communications Library (NCCL) 2.3.7 (optimized for NVLink™ )
• Horovod 0.15.1
• OpenMPI 3.1.3
• TensorBoard 1.12.2
• MLNX_OFED 3.4
• OpenSeq2Seq v18.12 at commit 59c70e7
• TensorRT 5.0.2
• DALI 0.6.1 Beta
• Jupyter and JupyterLab:
  • Jupyter Client 5.2.4
  • Jupyter Core 4.4.0
  • JupyterLab 0.35.4
  • JupyterLab Server 0.2.0

Driver Requirements

Release 19.02 is based on CUDA 10, which requires NVIDIA Driver release 410.xx. However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384. For more information, see CUDA Compatibility and Upgrades.

GPU Requirements

Release 19.02 supports CUDA compute capability 6.0 and higher. This corresponds to GPUs in the Pascal, Volta, and Turing families. Specifically, for a list of GPUs that this compute capability corresponds to, see CUDA GPUs. For additional support details, see Deep Learning Frameworks Support Matrix.

Key Features and Enhancements

This TensorFlow release includes the following key features and enhancements.

• TensorFlow container image version 19.02 is based on TensorFlow 1.13.0-rc0.
• Latest version of DALI 0.6.1 Beta
• Latest version of TensorBoard 1.12.2
• Added Jupyter and JupyterLab software in our packaged container.
• Latest version of jupyter_client 5.2.4
• Latest version of jupyter_core 4.4.0
• Ubuntu 16.04 with January 2019 updates

Accelerating Inference In TensorFlow With TensorRT (TF-TRT)

For step-by-step instructions on how to use TF-TRT, see Accelerating Inference In TensorFlow With TensorRT User Guide.

Key Features And Enhancements

• The following operators can now be converted from TensorFlow to TensorRT: ExpandDims, Reshape, Sigmoid, Sqrt, Square, Squeeze, StridedSlice and Tanh. For more information, see Supported Ops.

• You can manually insert quantization ranges (generated during quantization-aware training) to the graph, and then TF-TRT can use them during INT8 inference. That means calibration is not required with this feature. For more information, see INT8 Quantization.

Deprecated Features

• Support for TensorRT 3 has been removed.

Announcements

TensorRT 3.x is not longer supported, therefore, models that were accelerated using TensorRT 3.x will no longer run. If you have a production model that was accelerated with TensorRT 3.x, you will need to convert your model with TensorRT 5.x or later again.

For more information, see the Note in Serializing A Model In C++ or Serializing A Model In Python.

Known Issues

• Horovod and XLA cannot be used together due to a known issue in upstream TensorFlow. We expect this to be resolved in an upcoming release.
• There is a known performance regression with TensorFlow 1.13.0-rc0 for some networks when run with small batch sizes. As a workaround, increase the batch size.
• If using or upgrading to a 3-part-version driver, for example, a driver that takes the format of xxx.yy.zz, you will receive a Failed to detect NVIDIA driver version. message. This is due to a known bug in the entry point script's parsing of 3-part driver versions. This message is non-fatal and can be ignored. This will be fixed in the 19.04 release.