PyTorch Release 19.03

The NVIDIA container image for PyTorch, release 19.03, is available on NGC.

Contents of the PyTorch container

This container image contains the complete source of the version of PyTorch in /opt/pytorch. It is pre-built and installed in Conda default environment (/opt/conda/lib/python3.6/site-packages/torch/) in the container image. The container also includes the following:

• Ubuntu 16.04 including Python 3.6 environment
• NVIDIA CUDA 10.1.105 including cuBLAS 10.1.105
• NVIDIA cuDNN 7.5.0
• NVIDIA NCCL 2.4.3 (optimized for NVLink™ )
• APEX
• OpenMPI 3.1.3
• TensorRT 5.1.2
• DALI 0.7 Beta
• Tensor Core optimized examples:
  • Tacotron 2 and WaveGlow v1.1
  • SSD300 v1.1
  • Neural Collaborative Filtering (NCF)
  • Transformer
  • ResNet50 v1.5
  • GNMT v2
• 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 19.03 is based on CUDA 10.1, which requires NVIDIA Driver release 418.xx+. However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410. 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 19.03 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 PyTorch release includes the following key features and enhancements.

• PyTorch container image version 19.03 is based on PyTorch commit 81e025d from March 9th, 2019
• Latest version of NVIDIA CUDA 10.1.105 including cuBLAS 10.1.105
• Latest version of NVIDIA cuDNN 7.5.0
• Latest version of NVIDIA NCCL 2.4.3
• Latest version of DALI 0.7 Beta
• Latest version of TensorRT 5.1.2
• Added the Tacotron 2 and WaveGlow v1.1 and SSD300 v1.1Tensor Core examples
• Ubuntu 16.04 with February 2019 updates

Tensor Core Examples

These examples focus on achieving the best performance and convergence from NVIDIA Volta Tensor Cores by using the latest deep learning example networks 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.

• An implementation of the Tacotron 2 and WaveGlow v1.1 model. This text-to-speech (TTS) system is a combination of two neural network models: a modified Tacotron 2 model from the Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions paper and a flow-based neural network model from the WaveGlow: A Flow-based Generative Network for Speech Synthesis paper.
• An implementation of the SSD300 v1.1 model. The SSD300 v1.1 model is based on the SSD: Single Shot MultiBox Detector paper. The main difference between this model and the one described in the paper is in the backbone. Specifically, the VGG model is obsolete and is replaced by the ResNet50 model.
• An implementation of the Neural Collaborative Filtering (NCF) model. The NCF model focuses on providing recommendations, also known as collaborative filtering; with implicit feedback. The training data for this model should contain binary information about whether a user interacted with a specific item. NCF was first described by Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu and Tat-Seng Chua in the Neural Collaborative Filtering paper.
• An implementation of the Transformer model architecture. The Transformer model is based on the optimized implementation in Facebook's Fairseq NLP Toolkit and is built on top of PyTorch. The original version in the Fairseq project was developed using Tensor Cores, which provides significant training speedup. Our implementation improves the performance and is tested on a DGX-1V 16GB.
• An implementation of the ResNet50 model. The ResNet50 v1.5 model is a slightly modified version of the original ResNet50 v1 model that trains to a greater accuracy.
• An implementation of the 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.

Known Issues

• Persistent batch normalization kernels have been disabled due to a known bug during validation. Batch normalization provides correct results and work as expected from users, however, this may cause up to 10% regression in time to solution performance on networks using batch normalization.
• 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.