DALI Release 0.7 Beta

The DALI 0.7 is a beta release. Hence, for all the features, the functionality and performance will likely be limited.

Using DALI 0.7 Beta

The DALI 0.7 can be used with the 19.03 NVIDIA GPU Cloud (NGC) optimized container for MXNet, PyTorch, and TensorFlow.

To upgrade to DALI 0.7 beta from an older version of DALI, follow the installation instructions in the DALI Quick Start Guide.

Refer to the DALI Developer Guide for usage details.

Note: The internal DALI C++ API used for operators implementation, and the C++ API that enables using DALI as a library from native code, are not yet officially supported. Hence these APIs may change in the next release without advance notice.

Key Features and Enhancements

This DALI release includes the following key features and enhancements.

  • Added new operators:
    • HostDecoderRandomCrop
    • Element Extract for sequences
  • Added the API stub for incoming Optical Flow video operator.
  • Introduced Python 3.7 based whl build.
  • Added Docker-based "one click" build script.
  • Added the ability to fine-tune the internal DALI buffers using per-operator presize hints to stage the output queues.
  • Added step, stride and shuffling in SequenceReader.
  • Added a new, internal test API.
  • Added the ability to custom handle CPU affinity using the DALI_AFFINITY_MASK environment variable.

Breaking API Changes

  • None.

Deprecated Features

  • Dropped the Python 3.4 based whl build.

Known Issues

  • The new video reader operator requires NVIDIA VIDEO CODEC SDK support in the platform. The NVIDIA GPU Cloud (NGC) optimized containers lack this functionality in the default configuration prior to the 19.01 version. To enable the functionality, run the container with the "video" capability enabled, as below: 
    -e "NVIDIA_DRIVER_CAPABILITIES=compute,utility,video"
  • There is no clear distinction in the documentation between the operators supporting video sequences and the operators supporting images.