NVCaffe Release 18.02

The NVIDIA container image of Caffe, release 18.02, is available.

Contents of NVCaffe

This container image contains the complete source of the version of NVCaffe in /opt/caffe. It is pre-built and installed into the /usr/local/[bin,share,lib] directories in the container image. The container also includes the following:

• Ubuntu 16.04 including Python 2.7 environment
• NVIDIA CUDA 9.0.176 including:
  • CUDA^® Basic Linear Algebra Subroutines library™ (cuBLAS) 9.0.282 Patch 2 which is installed by default
  • cuBLAS 9.0.234 Patch 1 as a debian file. Installing Patch 1 by issuing the dpkg -i /opt/cuda-cublas-9-0_9.0.234-1_amd64.deb command is the workaround for the known issue described below.
• NVIDIA CUDA^® Deep Neural Network library™ (cuDNN) 7.0.5
• NCCL 2.1.2 (optimized for NVLink™ )

Driver Requirements

Release 18.02 is based on CUDA 9, which requires NVIDIA Driver release 384.xx.

Key Features and Enhancements

This NVCaffe release includes the following key features and enhancements.

• NVCaffe0.16.5 container image version 18.02 is based on Caffe Deep Learning Framework by the BVLC.
• Added the ImageDataParameter settings which belong to the ImageDataLayer's LayerParameter settings. More specifically:
  • Added the threads parameter which is the number of Data Transformer threads per GPU executed by the ImageDataLayer.
  • Added the cache parameter which ensures that the data is read once and put into the host memory.
• Latest version of cuBLAS
• Ubuntu 16.04 with January 2018 updates

Known Issues

cuBLAS 9.0.282 regresses RNN seq2seq FP16 performance for a small subset of input sizes. This issue should be fixed in the next update. As a workaround, install cuBLAS 9.0.234 Patch 1 by issuing the dpkg -i /opt/cuda-cublas-9-0_9.0.234-1_amd64.deb command.