DIGITS Release 18.05
The NVIDIA application of DIGITS, release 18.05, is available.
Contents of DIGITS
This application contains the complete source of the version of DIGITS in /opt/digits. It is pre-built and installed into the /usr/local/python/ directory in the application.
The container also includes the following:
- Ubuntu 16.04 including Python 2.7
- 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) 7.1.2
- NCCL 2.1.15 (optimized for NVLink™ )
- NVCaffe™ 0.17.0
- Torch™ 7
- TensorFlow™ 1.7.0
Driver Requirements
Release 18.05 is based on CUDA 9, which requires NVIDIA Driver release 384.xx.
Key Features and Enhancements
This DIGITS container image version includes the following key
features and enhancements.
- NVIDIA DIGITS application version 18.05 is based on DIGITS version 6.1.1
- Added a “Blob format” option to include and set
store_blobs_in_old_format in
solver.prototxt when training with NVCaffe. If you set
the option to NVCaffe and backend framework is NVCaffe,
solver.prototxt will include
store_blobs_in_old_format=false. This option is the
default and the caffemodel files generated from the
training stage are not compatible with BVLC Caffe.
If you set the option to Compatible, DIGITS sets store_blobs_in_old_format=true in solver.prototxt and caffemodel files are compatible with BVLC Caffe.Note: The store_blobs_in_old_format attribute is not recognized by BVLC Caffe, therefore, you have to remove that line when reusing solver.prototxt and fine-tuning with BVLC Caffe.
- Updated resizing images algorithms to remove deprecated Python modules. Previous DIGITS called scipy.misc.imresize to perform a few image resizing steps. This API is marked as deprecated in scipy.misc. DIGITS now utilizes skimage.transform.resize.
- Several updates and improvements on CI test scripts
- Ubuntu 16.04 with April 2018 updates