PyTorch Release 18.06
The NVIDIA container image of PyTorch, release 18.06, is available.
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-py3.6 Conda™ environment in the container image.
The container also includes the following:
- Ubuntu 16.04 including Python 3.6 environment
- NVIDIA CUDA 9.0.176 (see Errata section and 2.1) including CUDA® Basic Linear Algebra Subroutines library™ (cuBLAS) 9.0.333 (see section 2.3.1)
- NVIDIA CUDA® Deep Neural Network library™ (cuDNN) 7.1.4
- NCCL 2.2.13 (optimized for NVLink™ )
- Caffe2 0.8.1
Driver Requirements
Release 18.06 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.
- PyTorch container image version 18.06 is based on PyTorch 0.4.0 upstream main branch post commit 0e9613c.
- Improved data loader pipeline in the ImageNet example, see /opt/pytorch/examples/imagenet within the container.
- Data loader pipeline now uses pillow-simd and jpeg-turbo.
- Improved FP16 support, specifically, reductions like sum() are now more accurate when using FP16.
- Improved distributed performance, specifically, gradient communication can now overlap with gradient computation in backwards().
- Compatibility changes, specifically, Magma 1 is no longer supported.
- Ubuntu 16.04 with May 2018 updates