TensorFlow Release 20.09

The NVIDIA container image of TensorFlow, release 20.09, is available on NGC.

Contents of the TensorFlow container

This container image includes the complete source of the NVIDIA version of 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 18.04
  Note:

  Container image 20.09-tf1-py3 and 20.09-tf2-py3 contains Python 3.6

• NVIDIA CUDA 11.0.3 including cuBLAS 11.2.0
• NVIDIA cuDNN 8.0.4
• NVIDIA NCCL 2.7.8 (optimized for NVLink™ )
• Horovod 0.19.5
• OpenMPI 3.1.6
• TensorBoard
  • 20.09-tf1-py3 includes version 1.15.0+nv
  • 20.09-tf2-py3 includes version 2.3.2
• MLNX_OFED
• OpenSeq2Seq at commit 8f040a49
  • Included only in 20.09-tf1-py3
• TensorRT 7.1.3
• DALI 0.25.1
• DLProf 0.15.0
  • Included only in 20.09-tf1-py3
• Nsight Compute 2020.1.2.4
• Nsight Systems 2020.3.2.6
• XLA-Lite
• Tensor Core optimized examples: (Included only in 20.08-tf1-py3)
  • U-Net Medical
  • SSD320 v1.2
  • Neural Collaborative Filtering (NCF)
  • BERT
  • U-Net Industrial Defect Segmentation
  • GNMT v2
  • ResNet-50 v1.5
• JupyterLab 1.2.14 including Jupyter-TensorBoard

Driver Requirements

Release 20.09 is based on NVIDIA CUDA 11.0.3, which requires NVIDIA Driver release 450 or later. However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 418.xx or 440.30. The CUDA driver's compatibility package only supports particular drivers. For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.

GPU Requirements

Release 20.09 supports CUDA compute capability 6.0 and higher. This corresponds to GPUs in the NVIDIA Pascal, Volta, Turing, and Ampere Architecture GPU 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 images version 20.09 are based onTensorflow 1.15.3 and 2.3.2
• The latest version of NVIDIA CUDA 11.0.3 including cuBLAS 11.2.0
• The latest version of NVIDIA cuDNN 8.0.4
• The latest version of TensorBoard 2.3.2
• The latest version of DALI 0.25.1
• The latest version of DLProf 0.15.0
• Ubuntu 18.04 with August 2020 updates

Announcements

• Python 2.7 is no longer supported in this TensorFlow container release.
• The TF_ENABLE_AUTO_MIXED_PRECISION environment variables are no longer supported in the tf2 container because it is not possible to automatically enable loss scaling in many cases in the tf 2.x API. Instead tf.train.experimental.enable_mixed_precision_graph_rewrite() should be used to enable AMP.
• Deep learning framework containers 19.11 and later include experimental support for Singularity v3.0.

NVIDIA TensorFlow Container Versions

The following table shows what versions of Ubuntu, CUDA, TensorFlow, and TensorRT are supported in each of the NVIDIA containers for TensorFlow. For older container versions, refer to the Frameworks Support Matrix.

Tensor Core Examples

The tensor core examples provided in GitHub focus on achieving the best performance and convergence by using the latest deep learning example networks and model scripts for training. Each example model trains with mixed precision Tensor Cores on Volta, therefore you can get results much faster than training without tensor cores. This model is tested against each NGC monthly container release to ensure consistent accuracy and performance over time. This container includes the following tensor core examples.

• U-Net Medical model. The U-Net model is a convolutional neural network for 2D image segmentation. This repository contains a U-Net implementation as described in the paper U-Net: Convolutional Networks for Biomedical Image Segmentation, without any alteration. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).
• SSD320 v1.2 model. The SSD320 v1.2 model is based on the SSD: Single Shot MultiBox Detector paper, which describes an SSD as “a method for detecting objects in images using a single deep neural network”. Our implementation is based on the existing model from the TensorFlow models repository. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).
• Neural Collaborative Filtering (NCF) model. The NCF model is a neural network that provides collaborative filtering based on implicit feedback, specifically, it provides product recommendations based on user and item interactions. The training data for this model should contain a sequence of user ID, item ID pairs indicating that the specified user has interacted with, for example, was given a rating to or clicked on, the specified item. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).
• BERT model. BERT, or Bidirectional Encoder Representations from Transformers, is a new method of pre-training language representations which obtains state-of-the-art results on a wide array of Natural Language Processing (NLP) tasks. This model is based on BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding paper. NVIDIA's BERT is an optimized version of Google's official implementation, leveraging mixed precision arithmetic and Tensor Cores on V100 GPUS for faster training times while maintaining target accuracy. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).
• U-Net Industrial Defect Segmentation model. This U-Net model is adapted from the original version of the U-Net model which is a convolutional auto-encoder for 2D image segmentation. U-Net was first introduced by Olaf Ronneberger, Philip Fischer, and Thomas Brox in the paper: U-Net: Convolutional Networks for Biomedical Image Segmentation. This work proposes a modified version of U-Net, called TinyUNet which performs efficiently and with very high accuracy on the industrial anomaly dataset DAGM2007. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).
• GNMT v2 model. The GNMT v2 model is similar to the one discussed in the Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation paper. The most important difference between the two models is in the attention mechanism. In our model, the output from the first LSTM layer of the decoder goes into the attention module, then the re-weighted context is concatenated with inputs to all subsequent LSTM layers in the decoder at the current timestep. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).
• ResNet-50 v1.5 model. The ResNet-50 v1.5 model is a modified version of the original ResNet-50 v1 model. The difference between v1 and v1.5 is in the bottleneck blocks which requires downsampling, for example, v1 has stride = 2 in the first 1x1 convolution, whereas v1.5 has stride = 2 in the 3x3 convolution. The following features were implemented in this model; data-parallel multi-GPU training with Horovod, Tensor Cores (mixed precision) training, and static loss scaling for Tensor Cores (mixed precision) training. This model script is available on GitHub as well as NVIDIA GPU Cloud (NGC).

Known Issues

• A few models see performance regressions compared to the 20.08 release. Training WideAndDeep sees regressions of up to 30% on A100. In FP32 the TF1 Unet Industrial and Bert fine tuning training regress from 10-20%. Also the TF2 Unet Medical and MaskRCNN models regress by about 20% in some cases. These regressions will be addressed in a future release.
• There are several known performance regressions compared to 20.07. UNet Medical and Industrial on V100 and A100 GPUs can be up to 20% slower. VGG can be up to 95% slower on A100 and 15% slower on Turing GPUs. Googlenet can be up to 20% slower on V100. And ResNet50 inferencing can be up to 30% slower on A100 and Turing GPUs.

• An out-of-memory condition can occur in TensorFlow (TF1) 20.08 for some models (such as ResNet-50, and ResNext) when Horovod and XLA are both in use. In XLA, we added an optimization that skips compiling a cluster the very first time it is executed, which can help avoid unnecessary recompilations for models with dynamic shapes. On the other hand, for models like ResNet-50, the preferred compilation strategy is to aggressively compile clusters, as compiled clusters are executed many times. Per the "XLA Best Practices" section of the TensorFlow User Guide, running XLA with the following environment variable opts in to that strategy: TF_XLA_FLAGS=--tf_xla_enable_lazy_compilation=false

• There is a known performance regression of 10 to 30% compared to the 20.03 release when training the JoC V-Net Medical and U-Net Industrial models with small batch size on V100 and Turing GPUs. This will be addressed in a future release.
• There is a known performance regression of up to 60% when running inference using TF-TRT for SSD models with small batch size. This will be addressed in a future release.
• There is a known performance regression of up to 30% when training SSD models with fp32 data type on T4 GPUs. This will be addressed in a future release.
• There is a known issue where attempting to convert some models using TF-TRT produces an error "Failed to import metagraph". This issue is still under investigation and will be resolved in a future release.