PyTorch Release 17.09
The NVIDIA container image of PyTorch, release 17.09, is available.
PyTorch container image version 17.09 is based on PyTorch 0.2.0.
Contents of PyTorch
This container image contains the complete source of the version of PyTorch in /opt/pytorch
. It is pre-built and installed in the pytorch-py35
Conda™ environment in the container image.
The container also includes the following:
- Ubuntu 16.04
- NVIDIA CUDA® 9.0
- NVIDIA CUDA® Deep Neural Network library™ (cuDNN) 7.0.2
- NVIDIA® Collective Communications Library ™ (NCCL) 2.0.5 (optimized for NVLink™ )
Driver Requirements
Release 17.09 is based on CUDA 9, which requires NVIDIA Driver release 384.xx.
Key Features and Enhancements
This PyTorch release includes the following key features and enhancements.
- Supports Tensor Core operations for convolutions and GEMMs on Volta hardware
- The
examples
directory contains examples of ImageNet and LSTM training scripts that use FP16 data, as well as show how to train with FP16 - Matrix multiplication on FP16 inputs uses Tensor Core math when available
- A custom batch normalization layer is implemented to use cuDNN for batch normalization with FP16 inputs
- Latest version of CUDA
- Latest version of cuDNN with support for Tensor Core math when available
- Latest version of NCCL
- Ubuntu 16.04 with August 2017 updates
Known Issues
There are no known issues in this release.