NVIDIA/DALI v1.37.0 on GitHub

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

Added support for running JAX defined augmentations in the iterator and pipeline. (#5406, #5426, #5432)
Improved error reporting with a stack trace pointing to the offending operation in user code. (#5357, #5396)
Added CPU fn.random.choice operator. (#5380, #5387)
Added support for CUDA 12.4. (#5353, #5410)
Improved iterators checkpointing (#5374, #5375, #5371, #5356)
Optimized fn.resize operator for better GPU utilization (#5382)
Added option to skip bboxes in fn.random_bbox_crop with the fraction of area within the crop below user-provided threshold. (#5368)

Fixed Issues

Fixed handling of special values of the stream field in CUDA Array Interface v3 (#5425).
Fixed insufficient synchronization around scratch memory in nvImageCoded-based decoders (fn.experimental.decoders.*). (#5408)
Fixed readers saving incorrect checkpoint when restored and saved back in the same epoch. (#5378)

Improvements

Add JAX-defined augmentation examples (#5426)
Extend context and name propagation in errors (#5396)
Add experimental jax operator (#5406)
Enable Bandit security scan (#5402)
Reworks links in the RST documentation (#5413)
Refactor to remove duplicated logic in traverse_directories utility function (#5419)
Update DALI deps version (#5417)
Changes to dali/util/numpy (#5416)
Add libcurl-devel (#5412)
Move to CUDA 12.4 U1 (#5410)
Separate excutor interface and implementation files. (#5411)
Make the video reader use cudaVideoDeinterlaceMode_Adaptive only for non-progressive videos (#5392)
Skip AutoAug test when sanitizers are on (#5403)
Unpin typing_extensions in tests (#5405)
Dependency update 03-2024 (#5397)
Review Bandit reported vulnerabilities (#5398)
Support checkpointing in JAX decorators (#5374)
Workaround ASAN bug ignoring RPATH (#5388)
Update supported TensorFlow version (#5386)
Disable more video tests on selected machines (#5385)
Extend fn.random.choice to support n-D inputs (#5387)
Add random choice CPU operator for 0D samples (#5380)
Resize: Optimize block sizes, use dynamic amount of shared mem. (#5382)
Support checkpointing in JAX peekable iterator (#5375)
Increase DALI TF Plugin loading timeout (#5381)
Improve iterator checkpointing (#5371)
Improve logs when the DALI TF plugin loading process fails (#5379)
Add option to prune bboxes based on % area in Crop ROI (#5368)
Improve op deprecation and deprecate sequence reader (#5372)
Fix typo in nvcuvid error (#5373)
Optimize sanitizer operator tests (#5352)
Introduce operator origin stack trace in the error message (#5357)
Make ExternalContext more flexible (#5356)
Enable CUDA 12.4 build (#5353)

Bug Fixes

Add nose as a dependency to iterators tests (#5433)
Disable jax_function notebook conversions for unsupported Python3.8 (#5432)
Improve handling of CUDA Array Interface v3 (#5425)
Fix debug build (#5414)
Add stream synchronization before decode for nvImageCodec <= 0.2 (#5408)
Fix Loader checkpointing bug (#5378)
Fix pixelwise_masks support when the ratio is on in the coco reader (#5407)
Fix storage of non-POD random distributions. (#5395)
Fix nvImageCodec version check. (#5399)
Fix bug in checkpointing C API (#5390)
Add nose to the package list for TL1_separate_executor. (#5393)
Use host sync allocation for nvImageCodec <= 0.2 (#5391)
Remove temporary lock file from wheel (#5384)
Disable type annotation tests in sanitizer build (#5383)
Fix CUDA 12.4 with ASAN (#5370)
Skip video tests on M60 (#5369)
Enable eager mode tests, fix mixed ops and improve coverage (#5367)

Breaking API changes

There are no breaking changes in this DALI release.

Deprecated features

No features were deprecated in this release.

Known issues:

The following operators: experimental.readers.fits, experimental.decoders.video, experimental.inputs.video, and experimental.decoders.image_random_crop do not currently support checkpointing.
The video loader operator requires that the key frames occur, at a minimum, every 10 to 15 frames of the video stream.
If the key frames occur at a frequency that is less than 10-15 frames, the returned frames might be out of sync.
Experimental VideoReaderDecoder does not support open GOP.
It will not report an error and might produce invalid frames. VideoReader uses a heuristic approach to detect open GOP and should work in most common cases.
The DALI TensorFlow plugin might not be compatible with TensorFlow versions 1.15.0 and later.
To use DALI with the TensorFlow version that does not have a prebuilt plugin binary shipped with DALI, make sure that the compiler that is used to build TensorFlow exists on the system during the plugin installation. (Depending on the particular version, you can use GCC 4.8.4, GCC 4.8.5, or GCC 5.4.)
In experimental debug and eager modes, the GPU external source is not properly synchronized with DALI internal streams.
As a workaround, you can manually synchronize the device before returning the data from the callback.
Due to some known issues with meltdown/spectra mitigations and DALI, DALI shows best performance when running in Docker with escalated privileges, for example:
- privileged=yes in Extra Settings for AWS data points
- --privileged or --security-opt seccomp=unconfined for bare Docker.

Binary builds

NOTE: DALI builds for CUDA 12 dynamically link the CUDA toolkit. To use DALI, install the latest CUDA toolkit.

CUDA 11.0 and CUDA 12.0 builds use CUDA toolkit enhanced compatibility. 
They are built with the latest CUDA 11.x/12.x toolkit respectively but they can run on the latest, 
stable CUDA 11.0/CUDA 12.0 capable drivers (450.80 or later and 525.60 or later respectively).
However, using the most recent driver may enable additional functionality. 
More details can be found in enhanced CUDA compatibility guide.

Install via pip for CUDA 12.0:
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda120==1.37.0
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.37.0

or just:

pip install nvidia-dali-cuda120==1.37.0
pip install nvidia-dali-tf-plugin-cuda120==1.37.0

For CUDA 11:
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.37.0
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.37.0

or just:

pip install nvidia-dali-cuda120==1.37.0
pip install nvidia-dali-tf-plugin-cuda120==1.37.0

Or use direct download links (CUDA 12.0):

Or use direct download links (CUDA 11.0):

FFmpeg source code:

This software uses code of FFmpeg licensed under the LGPLv2.1 and its source can be downloaded here

Libsndfile source code:

https://developer.download.nvidia.com/compute/redist/nvidia-dali/libsndfile-1.2.2.tar.gz

NVIDIA/DALI v1.37.0 DALI v1.37.0 on GitHub