DALI Release 0.6 Beta

The DALI 0.6 is a beta release. Hence, for all the features, the functionality and performance will likely be limited.

Key Features and Enhancements

This DALI release includes the following key features and enhancements.
  • Added the CPU variant of DALI TensorFlow operator. Now the DALI pipeline can be scheduled on the CPU also.
  • Added the TensorFlow operator API to allow defining an arbitrary number of outputs and their types.
  • Added full DALI detection pipeline sample for SSD.
  • Added clear mapping between operators and the supported device types in the API documentation.
  • Created initial support for the video-based pipeline - NCCL functionality is available as VideoReader.
  • Added the below new operators:
    • BoxEncoder
    • VideoReader
    • SequenceCrop
    • SequenceReader
  • Added new variants of the below existing operators:
    • For the GPU:
      • RandomBBoxCrop
      • BbFlip
      • Slice
    • For the CPU:
      • CropMirrorNormalize, and
      • BBoxPaste

Breaking API Changes

  • DALI TensorFlow operator has a new API - check the examples for the reference.

  • PyTorch and MXNet python iterators API have changed - check the examples for the reference.

Using DALI 0.6 Beta

The 19.01 NVIDIA GPU Cloud (NGC) optimized container for MXNet, PyTorch, and TensorFlow includes an older version of DALI. To upgrade to DALI 0.6 beta, follow the installation instructions in the DALI Quick Start Guide.

Refer to the DALI Developer Guide for usage details.

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, resulting in problem with loading libnvcuvid.so. To enable it, run the container with the ‘video’ capability enabled, as below:
    -e "NVIDIA_DRIVER_CAPABILITIES=compute,utility,video"