DALI Versions Save

A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.

v1.36.0

3 weeks ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added support for checkpointing in MXNet iterator and CPU TensorFlow plugin (#5334, #5315).
  • Added morphological operators (fn.experimental.dilate, fn.experimental.erode) (#5294).
  • Integrated nvImageCodec for decoding in fn.experimental.decoders (#5297, #5336, #5324, #5333, #5339).
  • Added fn.random_crop_generator operator (#5304).
  • Added support for multiple inputs and relative shapes and anchors in fn.multi_paste (#5331).

Fixed Issues

  • Fixed insufficient synchronization in MXNet iterator (#5364).
  • Fixed auto_reset argument handling in iterator plugins (#5340).
  • Fixed missing calls to nvml::Shutdown (#5317).
  • Limited a number of progressive scans for jpeg decoding (#5316).

Improvements

  • Propagate module and display name of the operator to backend (#5344)
  • Update dependencies (#5349)
  • Map backend exceptions into Python exception types (#5345)
  • Emphasise the optical flow is calculated at input resolution. (#5350)
  • Refactor custom ops classes to use python_op_factory as base (#5338)
  • Add origin stack trace capture for DALI operators (#5302)
  • Test fused decoder with two separate pipelines (#5343)
  • [Cutmix] Make fn.multi_paste more flexible, fix validation (#5331)
  • Enable checkpointing in TensorFlow plugin (CPU only) (#5334)
  • Copy out nvImageCodec conda package from the build (#5336)
  • Add error message when GPU is not available (#5329)
  • Enable build with statically linked nvimgcodec + hard dependency for dynamic linking (#5324)
  • Add tf_stack util to autograph (#5322)
  • Rewrite median blur to use nvcvop tools (#5327)
  • Add morphological operators and the nvcvop module (#5294)
  • Add OpSpec::ArgumentInputIdx (#5330)
  • Simplify workspace object. Ensure predictable argument order in OpSpec. (#5325)
  • Support checkpointing in MXNet iterator (#5315)
  • Set rpath at cmake level (do not wait for bundle-wheel) (#5323)
  • Interpolation modes documentation upgrade (#5314)
  • Update links in DALI documentation (#5321)
  • Integrate nvimagecodec (#5297)
  • Add naive_histogram custom operator to test suite (#4731)
  • Add RandomCropGenerator (#5304)
  • Use small videos in checkpointing tests (#5305)

Bug Fixes

  • Use synchronous copy to framework array in the absence of a stream (#5364)
  • Process TFRecord reader binding classes only when it is enabled (#5360)
  • Adjust stack formatting in backend to match Python (#5354)
  • Link test operators against nvml wrapper (#5355)
  • Fix range check in Workspace::SetInput (#5358)
  • Make async_pool immune to stream handle reuse. (#5348)
  • Coverity fixes for 1.36 (#5342)
  • Fix "auto_reset" argument handling (#5340)
  • Fix cupy tests (#5341)
  • Add nvimagecodec libs to DALI_EXCLUDES + test utils to dump mismatched images (#5339)
  • Fix warning about nvImageCodec version (#5333)
  • Silence warning about DOWNLOAD_EXTRACT_TIMESTAMP while fixing the cmake <3.24 builds (#5326)
  • Fix inconsistent calls to nvml::Init and nvml::Shutdown (#5317)
  • Limit the number of progressive scans for jpeg decoding (#5316)

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==1.36.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.36.0

For CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.36.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.36.0

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:

v1.35.0

1 month ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added support for PaddlePaddle and JAX iterators (#5279, #5282).
  • Added support for checkpointing of iterators with different last batch policies (#5298, #5278).
  • Added tutorial on running DALI with T5X (#5286).
  • Added do_not_convert decorator to address problems with parallel fn.external_source and conditional execution (#5263).

Fixed Issues

  • Fixed missing nvmlShutdown calls (#5311).
  • Fixed fn.readers.video handling of sequences bigger than 2GB (#5307).
  • Fixed fn.resize handling of samples larger than 2GB (#5306).
  • Fixed support for multi node JAX sharding (#5242).
  • Fixed handling of decorated callbacks and methods in fn.external_source (#5268).
  • Fixed insufficient synchronization when restoring random generators from a checkpoint (#5273).

Improvements

  • Skip slow checkpointing tests when sanitizer is enabled (#5310)
  • Support checkpointing in PaddlePaddle iterator (#5279)
  • Bump protobuf version requirements in CMake. (#5312)
  • Dependency update (#5308)
  • Add SSL support to aarch64 deps containers (needed by CMake) (#5300)
  • Enable OpenSSL support (allow https downloading from cmake) (#5299)
  • Expose checkpointing in C API (#5287)
  • Enable AG-transformed code to show user code in exception (#5291)
  • Remove stage-related public APIs (from Pipeline and Executor) (#5244)
  • Add T5X tutorial (#5286)
  • Temporarily reduce the number of epochs in SBSA torch test (#5290)
  • Add a tool for correcting typos (#5193)
  • Support checkpointing in JAX iterator (#5282)
  • Expose do_not_convert decorator (#5263)
  • Fix Pipeline docs formatting (#5283)
  • Generalize iterator checkpointing tests (#5278)

Bug Fixes

  • Add missing calls to nvmlShutdown (#5311)
  • Fix bugs in iterator checkpointing, enable other last batch policies (#5298)
  • Fixes support for video sequences bigger than 2GB (#5307)
  • Fix resizing of volumes larger than 2G. (#5306)
  • Fix memory leak in C API test (#5303)
  • Fix support for multi node JAX sharding (#5242)
  • Fix TestPytorch (#5284)
  • Fix capitalized property descriptions (#5285)
  • Fix the lack of flavor in the conda DALI package name (#5281)
  • Allow source to be a decorated function or method (#5268)
  • Fix f-string in RN50 data pipeline test (#5280)
  • Remove old assumption from iterator docs (#5277)
  • Suppress leaks from inside dlopen (#5245)
  • Fix PaddlePaddle plugin docs (#5276)
  • Fix synchronization issues in RNG checkpointing utils (#5273)

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==1.35.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.35.0

or just:

pip install nvidia-dali-cuda120==1.35.0 pip install nvidia-dali-tf-plugin-cuda120==1.35.0

For CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.35.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.35.0

or just:

pip install nvidia-dali-cuda110==1.35.0 pip install nvidia-dali-tf-plugin-cuda110==1.35.0

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:

v1.34.0

2 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added support for CUDA 12.3 U2 (#5262)
  • Added support for checkpointing in fn.random_resized_crop (#5246)

Fixed Issues

  • Fixed synchronization problem when restoring GPU random operator checkpoints (#5273).
  • Fixed warnings on pipeline teardown in debug mode. (#5267)
  • Added check for reentrant version of CFITSIO for fits reader. (#5239)
  • Fixed scalar inputs handling in GPU fn.lookup_table. (#5257)
  • Added missing validation for bboxes in fn.ssd_random_crop (#5240)
  • Added validation that prevents running parallel externeral source without Python workers (#5238)

Improvements

  • Split conda built into core and python bindings (#5259)
  • Dependency update - 2024/01 (#5271)
  • Add framework attributes to DLFW iterator tests. (#5266)
  • Move to CUDA 12.3 U2 (#5262)
  • Add links to DALI success stories in README (#5247)
  • Add missing imports to AA simple examples in docs (#5243)
  • Support checkpointing in random_resized_crop (#5246)
  • Add linter GitHub Action (#5236)
  • Adjust the error message on failed IsDenseTensor check (#5237)
  • Format docs directory with black (#5214)

Bug Fixes

  • Fix PaddlePaddle plugin docs (#5276)
  • Fix synchronization issues in RNG checkpointing utils (#5273)
  • Fix missing Shutdown method warning in debug pipeline. (#5267)
  • Fix issues detected by Coverity (2024.01) (#5272)
  • Check if reentrant version of CFITSIO is used (#5239)
  • Fix LookupTable GPU for scalar inputs (#5257)
  • Add dimension check in ssd_random_crop (#5240)
  • Add validation preventing running PES without Python workers (#5238)

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==1.34.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.34.0

or just:

pip install nvidia-dali-cuda120==1.34.0 pip install nvidia-dali-tf-plugin-cuda120==1.34.0

For CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.34.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.34.0

or just:

pip install nvidia-dali-cuda110==1.34.0 pip install nvidia-dali-tf-plugin-cuda110==1.34.0

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:

v1.33.0

3 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Enhanced experimental support for checkpointing (saving and resuming DALI pipelines at arbitrary iteration) (#5232, #5195, #5166):
    • Support data readers checkpointing (#5198, #5213, #5184, #5182, #5165, #5181, #5180, #5183, #5162, #5139).
    • Support randomized GPU operators checkpointing (#5216, #5148).
    • Improved checkpointing documentation (#5230).
  • Improved Python annotations and signatures (#5217, #5159, #5167, #5154, #5188, #5158, #5150).
    • Added annotations for JAX and Pytorch iterators (#5129, #5197).
    • Improved PythonFunction annotations (#5207, #5149).
    • Improved data type annotations (#5179, #5153)
  • Improved JAX support:
    • Added pmap compatibility for JAX data_iterator (#5185).
    • Improved JAX, Flax and Paxml training examples (#5176, #5205).
  • Moved to CUDA 12.3U1 and enbaled GDS and nvJPEG2k support for the SBSA platform (#5209, #5170).
  • Added Python 3.11 support and experimental support for Python 3.12 (#5174)

Fixed Issues

  • Fixed fn.normalize handling of batch of empty samples (#5223).
  • Fixed infinite video decoder seek loop (#5218).
  • Fixed computation of maximal threads number for kernels in GPU fn.transpose and fn.normalize. (#5208)
  • Fixed handling of empty slices and slicing of empty inputs. (#5204)
  • Fixed scalar constant dimensionality inference (#5191)
  • Fixed sharding in Caffe reader (#5172)

Improvements

  • Mark missed operator stateless, assure checkpointing tests coverage (#5232)
  • Add improved checkpointing docs (#5230)
  • Support checkpointing in Numpy reader (#5198)
  • Set the maximum supported Python version to Tensorflow 2.13 in tests (#5234)
  • Dependency update 23/12 (#5231)
  • Support checkpointing in External Source (#5213)
  • Support checkpointing in other random operators (#5216)
  • Remove a redundant call to Executor::GetTensorQueueSizes() (#5225)
  • Fix links to release notes and docs archive (#5227)
  • Install black with jupyter formatting capabilities (#5226)
  • Fix the minimal Python version for TF 2.14 and 2.15, adds 2.13 (#5221)
  • Unify positional input name handling in docs and signatures (#5217)
  • Add launch_bounds to SliceNormalizeKernel_2D kernels (#5206)
  • Update test_RN50_external_source_parallel_train_ddp.py to work with the latest PyTorch (#5219)
  • Rework Flax and Paxml training tutorials (#5205)
  • Mark all remaining stateless operators (#5195)
  • Install black formatter in the aarch64 build (#5211)
  • Fix typos in docs and comments (#5194)
  • Bump up TensorFlow version in tests (#5175)
  • Enbale nvJPEG2k support for the SBSA platform (#5209)
  • Extend Python Function and plugin type annotations (#5207)
  • Update readme about the Black formatting (#5212)
  • Add "Format DALI with black" to blame ignore revs (#5210)
  • Add type annotations for JAX plugin (#5197)
  • Fix/simplify slice usage in examples. (#5203)
  • Re-enable Python linter (#5189)
  • Format DALI with black (#5169)
  • Adjust configuration for autoformatting with black (#5168)
  • Introduce TensorLike to signatures utilizing Array Interface (#5179)
  • Update nvJPEG2k to 0.7.5 version (#5202)
  • Add doc build artifacts to gitignore (#5187)
  • Update FFmpeg to 6.1 (#5186)
  • Add pmap compatibility for JAX data_iterator (#5185)
  • Support checkpointing in Nemo reader (#5184)
  • Support checkpointing in Webdataset reader (#5182)
  • Support checkpointing in mxnet and tfrecord (#5165)
  • Support checkpointing in Caffe reader (#5181)
  • Support checkpointing in experimental video reader (#5180)
  • Support checkpointing in sequence reader (#5183)
  • Enable Python 3.11 test and 3.12 experimental support (#5174)
  • Make sure that GDS on SBSA is not tested for drivers below cuda 12.2 (#5177)
  • Add checkpointing benchmarks (#5166)
  • Improve WarpAffine input documentation (#5178)
  • Refactor JAX basic training example (#5176)
  • Support checkpointing in Coco Reader (#5162)
  • Move to CUDA 12.3U1 and enable GDS support for SBSA (#5170)
  • Expose Tensors types stubs in nvidia.dali.tensors module (#5153)
  • Make the video reader warn instead of failing on unreadable files (#5163)
  • Implement checkpointing for random GPU operators (#5148)
  • Generate overloads for Multiple Input Sets in interface files (#5159)
  • Improve API submodules discoverability in some language engines (#5167)
  • Expose current pipeline in python functions (#5156)
  • Set signature for fn API (#5154)
  • Add file to ignore Flake8 commits in blame (#5161)
  • Suppress sanitizer leaks reported from xla and prevent hang in git clone (#5152)
  • Add checkpointing support to Video Reader (#5139)
  • Add type annotations for numba and pytorch plugins (#5129)
  • Improve Python Function signature annotation (#5149)
  • Dependency update 2023.11 (#5146)
  • Bump up CUDA version used for tests (#5100)

Bug Fixes

  • Patch libtiff for CVE-2023-6277 (#5224)
  • Coverity fixes: fix fn.normalize handling of batch of empty samples, fix broken assertion in copy_with_stride (#5223)
  • Avoids infinite video decoder seek loop (#5218)
  • Fix and move caching from MaxThreadsPerBlock to MaxThreadsPerBlockStatic. (#5208)
  • Remove redundant overload signature for operators without inputs (#5188)
  • Fix hanlding of empty slices and slicing of empty inputs. (#5204)
  • Fix slice/stack/cat/scalar usage in RNNT pipeline tests. (#5201)
  • Fix Python3.11 tests (#5196)
  • Fix scalar constant dimensionality & hide constant op documentation (#5191)
  • Fix sharding in Caffe reader (#5172)
  • Fix the check for dynamically generated number of outputs (#5158)
  • Improve symbol visibility when importing (#5150)

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, random_resized_crop, 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==1.33.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.33.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.33.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.33.0

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:

v1.32.0

4 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added Python signatures/type hints to the DALI Python API (#5096, #5039, #5112, #5118, #5124, #5143).
  • Added experimental support for checkpointing DALI pipelines at arbitrary iterations (fn.readers.file, CPU fn.random generators, and stateless operators) (#5085, #5088, #5103, #5114, #5113, #5142, #5128, #5144).
  • Added support for CUDA 12.3 (#5106).

Fixed Issues

  • Fixed a potential crash on process teardown when using the fn.python_function in the DALI pipeline (#5138).
  • Removed unused arguments from fn.fast_resize_crop_mirror. The operator was deprecated in favor of fn.resize_crop_mirror (#5123).
  • Fixed a potential crash on process teardown when using CPU fn.resize in the DALI pipeline (#5133).
  • Fixed constructing tensors from stream-aware __cuda_array_interface__ v3 (#5125).
  • Fixed the crop_pos_z handling for a fixed crop window in the fn.crop operator (#5119).
  • Fixed releasing Python tensors without GIL in fn.external_source. The problem led to crashes when using fn.external_source in no_copy or parallel mode with conditional execution enabled in the pipeline (#5101).

Improvements

  • Improve checkpointing docs (#5142)
  • Add type hints to python-defined ops, run and tfrecord APIs (#5118)
  • Allow mid-epoch checkpointing in FileReader (#5113)
  • Mark more stateless operators (#5114)
  • Load ASAN during the build with sanitizers (#5121)
  • Add default arg values to JAX decorator (#5115)
  • Improve docs of variance and stddev (#5130)
  • Update python op tutorial (#5120)
  • Generalize __module__ handling and hide private modules docs (#5112)
  • Add mid-epoch checkpointing to Loader (#5103)
  • Add ViT data processing pipeline to hw_bench_script (#5110)
  • Add JAX Getting Started Tutorial (#5095)
  • Generate type hints for fn and ops APIs (#5096)
  • Move to CUDA 12.3 (#5106)
  • Add the BUILD ID to the Xavier wheel name (#5099)
  • Add Fast-Forward to Loader base (#5088)
  • Update installation guide to move https://pypi.nvidia.com/ or just pypi (#4815)
  • Remove Python 3.7 support and replace defaults with Python 3.8 (#5089)
  • Add better error for VFR check (#5092)
  • Move the snapshot queue from loader to reader (#5085)
  • Refactor ops and fn APIs (#5039)

Bug Fixes

  • Fix Python Functions ops signature (#5143)
  • Fix stateless tests (#5144)
  • Fix exit sequence. (#5138)
  • Fix data_iterator docs (#5140)
  • Reimplement (Fast)ResizeCropMirror in terms of Resize. Remove dead code. (#5123)
  • Ensure that the default host resource is destroyed after ResamplingFilters CPU instance. (#5133)
  • Fix CVE-2023-45853 in zlib (#5116)
  • Fix stub generation during DALI build (#5124)
  • Improve checkpointing error message (#5128)
  • Fix constructing tensors from cuda_array_interface v3. (#5125)
  • Fix lack of crop_pos_z handling for fixed crop window (#5119)
  • Fix CVE-2022-33065 in libsnd (#5105)
  • Add DALI package dependencies to the custom build test (#5122)
  • Make sure that the relase of python memory from the DALI tensor happens inside GIL (#5101)
  • Relax UpdatePropertiesFromSamples check constraints (#5098)

Breaking API changes

  • DALI 1.31 was the final release that supported Python 3.7.

Deprecated features

  • The operator fn.fast_resize_crop_mirror was deprecated in favour of fn.resize_crop_mirror.

Known issues:

  • 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==1.32.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.32.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.32.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.32.0

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:

v1.31.0

5 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Preliminary experimental support for pipeline checkpointing. (#5061, #5057)
  • Added data_iterator and peekable_data_iterator decorators for simplified JAX iterators definitions. (#5050, #5049)
  • Added the "Training neural network with DALI and Pax" tutorial. (#5060)

Fixed Issues

  • The fn.permute_batch operator can now be used with the conditional execution (if expressions). (#5063)
  • Fixed support for videos with different bit depths in the video decoder. (#5055)
  • Input operators with multiple outputs can be fed with data by the operator name. (#5066)

Improvements

  • Expose checkpointing in Python pipeline (#5061) 
  • Update deps: RapidJSON, OpenCV (#5079) 
  • Fix coverity issues 10/23 (#5083) 
  • Add Pax tutorial (#5060) 
  • Add Efficientnet pipeline to hw_decoder_bench (#5076) 
  • Add JAX iterator decorator (#5050) 
  • Update libwep library to remediate CVE-2023-5129 (#5075) 
  • Replace optional stream in SaveState with AccessOrder  (#5062) 
  • Add implicit scope to batch_permutation (#5063) 
  • Fix enumeration formatting in conditionals docs (#5067) 
  • Update DALI key visual (#5069) 
  • Deprecate Python 3.7 starting DALI 1.31 (#5068) 
  • Extend HW image decoder bench script to support multiple GPUs (#5065) 
  • Remove the avformat_find_stream_info call from the video loader when not needed (#5047) 
  • Add ability to serialize/deserialize Checkpoint (#5057) 
  • Remove dali::any in favor of std::any. (#5058) 
  • Disable container overflow sanitizer all the time (#5053) 
  • Replace PaddlePaddle ResNet50 example with one from the DeepLearningExamples (#5048) 
  • Make the ResNet50 example compatible with TensorFlow 2.13 (#5045) 
  • Reorganize JAX plugin (#5049) 
  • Replace GPU dltensor per-sample copying kernel with a batched one (#5038) 
  • September dependency update (#5043) 

Bug Fixes

  • Add user-friendly message about missing numpy (#5081) 
  • Set external input by op name instead of tensor name (#5066) 
  • Fix the support of videos with different bith dept in the video reader (#5055) 
  • Set layout from argument in external source (#5064) 
  • Update JAX version to 0.4.13 (#5056) 
  • Extend Numba compatibility checks. Skip Numba GPU tests on incompatibe systems. (#5054) 

Breaking API changes

There are no breaking changes in this DALI release.

Deprecated features

  • Python 3.7 support is deprecated starting from DALI 1.31.

Known issues:

  • 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==1.31.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.31.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.31.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.31.0

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:

v1.30.0

6 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added support for running custom CPU and GPU Python operators (fn.*python_function) inside DALI asynchronous pipelines (#4965, #5038).
  • Improved support for GPU Numba operator (plugin.numba.fn.experimental.numba_function) (#4000).
  • Improved (fn.crop_mirror_normalize) performance (#4993, #4992).
  • Added support for strides in subscript operator (#5007).
  • Added support for video in predefined automatic augmentations (#5012).
  • Added case insensitive mode in fn.readers.webdataset (#5016).
  • Moved to CUDA 12.2U2 (#5027).
  • Added Flax training examples (#5004, #4978).

Fixed Issues

  • Fixed GPU fn.readers.numpy global shuffling (#5034).
  • Fixed finalization of custom operator plugins during pipeline shutdown (#5036).
  • Fixed synchronization issue in fn.resize operator family that could result in distorted outputs in initial iterations (#4990).

Improvements

  • Replace GPU dltensor per-sample copying kernel with a batched one (#5038)
  • September dependency update (#5043)
  • Make download_pip_packages.sh resilient to errors (#5044)
  • Move to CUDA 12.2U2 (#5027)
  • Clean up and refactor code around Multiple Input Sets (#5035)
  • Move to the upstream CV-CUDA 0.4 (#5032)
  • Revert "Make nesting conditionals supported only for Python 3.7+" (#5031)
  • Move all remaining video files to LFS (#5025)
  • Refactor custom op wrappers into separate files of ops module (#5028)
  • Add pipeline checkpointing to the Executor (#5008)
  • Refactor ops into a submodule (#5018)
  • Add checkpointing support to ImageRandomCrop (#4999)
  • Replace deprecated fluid APIs to recommended APIs of Paddle (#5020)
  • fix: CMakeLists.txt typo (#5006)
  • Support video in predefined automatic augmentations (#5012)
  • Extend GPU numba support (#4000)
  • Add opt-in support for case insensitive webdataset (#5016)
  • Add optimized variant of CMN for HWC to HWC pad FP16 case (#4993)
  • Added Stride to Subscript and Slice Kernel (#5007)
  • Add optimized variant of CMN for HWC to HWC case (#4992)
  • Add multiple GPU code to Flax example (#5004)
  • Pin inputs to decoder operators as well (#5003)
  • Add checkpointing support to stateless operators used in EfficientNet (#4977)
  • Use a different way to ensure that the right version of libabseil is used in conda (#4991)
  • Make samples' descriptors copy in resize op fully asynchronous (#4989)
  • Remove mentions of experimental from conditional tutorial. (#4988)
  • Enable python operators in async pipelines (#4965)
  • Make sure that the right version of libabseil is used in conda (#4987)
  • Coverity fixes - 08.2023 (#4970)
  • CPU fn.random operators checkpointing (#4961)
  • Add Flax training example (#4978)
  • Make error reporting more verbose for rand augment tests (#4958)

Bug Fixes

  • Propagate to conda build packages required for DALI installation (#5041)
  • Fix wheel predownload (#5023)
  • Fix GPU numpy reader global shuffling (#5034)
  • Change the way the input operators are traversed during the pipeline shutdown (#5036)
  • Fix issues detected by Coverity as of 2023.09.04. (#5030)
  • Fix CUDA block sizes in Numba GPU tests. (#5026)
  • Change Loader to make checkpoints at the end of an epoch (#5019)
  • Disable Flax tutorial test (#5015)
  • Fix resize processing cost calculation (#5009)
  • Fix abs diff computation in check_batch test utility (#4957)
  • Fix sync in Resize operator family (#4990)

Breaking API changes

There are no breaking changes in this DALI release.

Deprecated features

No features were deprecated in this release.

Known issues:

  • 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==1.30.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.30.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.30.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.30.0

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:

v1.29.0

7 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added GPU fn.experimental.median_blur operator. (#4950, #4975)
  • Improved JAX support:
    • Added support for jax.Sharding to dali.plugin.jax.DALIGenericIterator (#4969).
    • Improved examples and tutorials (#4973, #4956, #4944, #4937).
  • Optimized the HWC to CHW transposition variant of the fn.crop_mirror_normalize operator (#4972).
  • Moved to CUDA 12.2U1 (#4966)

Fixed Issues

  • Fixed layout broadcasting in arithmetic expressions (#4951).
  • Added missing layout propagation in fn.reductions (#4947).

Improvements

  • Trim CV-CUDA to expose only median blur to reduce the binary size (#4985)
  • Add optimized variant of CMN for HWC to CHW case (#4972)
  • Enable CV-CUDA build for xavier (#4976)
  • Update DALI_deps version (#4971)
  • Add automatic parallelization JAX example (#4973)
  • Exclude median_blur test from xavier tests (#4975)
  • Move to CUDA 12.2 U1 (#4966)
  • Add basic jax.Sharding support for the iterator (#4969)
  • Enable cv-cuda in conda build (#4968)
  • Fix wheel bundling with cvcuda for debug builds (#4959)
  • Fix Getting Started link in README (#4962)
  • Add multigpu JAX tutorial (#4956)
  • Add median blur operator (#4950)
  • Fix updated linter errors (#4960)
  • Support checkpointing in FileReader (#4954)
  • Add CV-CUDA as a subproject (#4949)
  • Remove the direct use of cuda_for_dali auxiliary namespace. (#4953)
  • Checkpointing classes (#4946)
  • Make sure that lossless support is disabled when it fails to initialize (#4934)
  • Add L3 short test for RN50 training (#4614)
  • DALI_deps update 13 Jul 2023 (#4945)
  • Add JAX tutorial tests (#4944)
  • Update OpenCV 4.7.0 to 4.8.0, patch for CVE-2023-1999 (#4941)
  • Fix L1 Jupyter Conda Job (#4942)
  • Update the TensorFlow version used in tests (#4940)
  • Add basic JAX tutorial (#4937)

Bug Fixes

  • Checkpoint after running epoch (#4983
  • Propagate layout in fn.reductions (#4947)
  • Fix layout broadcasting arithm ops (#4951)
  • Fix coverity issues - July 2023 (#4948)

Breaking API changes

There are no breaking changes in this DALI release.

Deprecated features

No features were deprecated in this release.

Known issues:

  • 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==1.29.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.29.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.29.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.29.0

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:

v1.28.0

8 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added CUDA 12.2 support (#4930, #4938, and #4939).
  • Added cudaMallocAsync support (#4900, #4923, and #4921).
  • Improved JAX multiprocessing support (#4929, #4927, #4919, #4906, and #4920).
  • Added DALIRaggedIterator, a DALI Pytorch plugin iterator that supports non-uniform tensors (#4911).

Fixed Issues

No major fixes are included in this release.

Improvements

  • Fix OpticalFlow test premature exit on sm < 8 (#4933)
  • Remove dependency on forked libcudacxx (#4938)
  • Add JAX multinode multigpu tests (#4929)
  • Adding handling of non-uniform tensors in DALI Pytorch plugin (#4911)
  • Reworks supported Python versions (#4924)
  • Disable cudaMemPoolReuseAllowOpportunistic in cudaMallocAsync for <r470.60 (#4931)
  • Move to CUDA 12.2 (#4930)
  • Remove template from tensor rule-of-five for c++20 compat (#4928)
  • Add JAX container test job (#4927)
  • Extends guards against intercepting by asan certain functions (#4925)
  • Fix CUDA_remove_toolkit_include_dirs CMake function (#4922)
  • Add alignment to cuda_malloc_async_memory_resource. (#4923)
  • Add source_info to the tensors produced by video readers (#4916)
  • Add JAX multigpu sharding tests (#4919)
  • Add basic JAX multi process test (#4906)
  • Add libabseil as a runtime DALI dependency in conda (#4907)
  • Remove pinning Cython version from PyThon SSD test (#4913)
  • Add a memory resource based on cudaMallocAsync (#4900)

Bug Fixes

  • Fix memory_resource compilation in conda build (#4939)
  • Disable JAX iterator tests in ASAN build (#4920)
  • Fix number of devices for JAX multigpu test (#4921)
  • Remove unnecessary cudaDeviceSynchronize from memory resource perf test. (#4908)
  • Fix broken assertion in sequence operator (#4905)

Breaking API changes

  • DALI 1.27 was the final release that supported Python 3.6.

Deprecated features

No features were deprecated in this release.

Known issues:

  • 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==1.28.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.28.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.28.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.28.0

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:

v1.27.0

9 months ago

Key Features and Enhancements

This DALI release includes the following key features and enhancements:

  • Added O_DIRECT support mode support to fn.readers.tfrecord (#4820).
  • Added JAX integration (#4867, #4883, #4853).
  • Added the GPU backend for fn.experimental.readers.fits images that are stored in the FITS format (#4752).

Fixed Issues

  • Assured deterministic outputs for multiple instances of auto_augment pipelines that are built with the same seeds (#4885).
  • Fixed the blocking option in the external source operator (#4874).
  • Fixed the returning empty pixel mask for COCO samples with no objects (#4856).
  • Fixed the handling of unsupported images by image decoders in fn.experimental.decoders (#4846).

Improvements

  • Update deps 23/06 (#4902)
  • Add O_DIRECT support to the TFRecord reader (#4820)
  • Relax the gast version requirement (#4896)
  • Add DALI iterator for JAX (#4867)
  • Fix coverity issues (#4897)
  • Add deprecation warning for Python3.6 (#4895)
  • Use memory pool for large host allocs (#4886)
  • Improve the feed_input documentation regarding prefetching (#4875)
  • Support nesting data structures in conditionals (#4880)
  • Add JAX multi GPU tests (#4883)
  • Move the mention of the EfficientNet example to a box (#4882)
  • Update the Protobuf version to 23.01 and adjust the build system to it (#4861)
  • Add basic JAX integration (#4853)
  • Limit the version of typing_extensions for the TensorFlow test (#4863)
  • Add GPU implementation for Fits reader (#4752)
  • Disable Numba CPU tests on AARCH64. (#4862)
  • Update readme text and code highlighting (#4858)
  • Disable NUMBA CPU test for runs with memory sanitizer (#4854)
  • Adjust numpy reader tests for nose2 (#4851)
  • Update support for Numba 0.57 (#4845)
  • Move to CUTLASS 3.1 (#4841)
  • Add a test that triggers a failure in Python (#4836)
  • Improve VA reservation robustness (#4826)
  • fix: bad relative path (#4822)

Bug Fixes

  • Skip DLPack CPU export test for incompatible Numpy (#4904)
  • Fix parsing numpy header (#4903)
  • Remove outdated info from iterators docs. (#4899)
  • Bugfix (async_pool): Store original alignment in 'padded_'. (#4898)
  • Fix the augmentation coalescing in AA (#4887)
  • Skip tests for incompatible env (#4894)
  • Make nesting conditionals supported only for Python 3.7+ (#4888)
  • Fix DALI FW iterator reset for DROP last batch policy (#4881)
  • Assure same operator initialization order in the AA graph (#4885)
  • Fix the lack of support for the blocking option in the external source operator (#4874)
  • Disable container overflow errors (#4878)
  • Fix the wrong assignment of the default values in build_helper.sh (#4871)
  • Disable JAX support for unsupported Python versions (#4870)
  • Disable FITS test when not building with CFITSIO support. Fix build without libTIFF. (#4866)
  • Fix layout propagation in jpeg compression distortion (#4864)
  • Fix returning empty pixel mask for COCO samples with no objects (#4856)
  • Bugfix in imgcodec: filter should happen after set decode result (#4846)
  • Don't run image decoder tests in test discovery stage. (#4833)

Breaking API changes

There are no breaking changes in this DALI release.

Deprecated features

DALI 1.27 is the final release that will support Python 3.6.

Known issues:

  • 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==1.27.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda120==1.27.0

or for CUDA 11: pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-cuda110==1.27.0 pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/ nvidia-dali-tf-plugin-cuda110==1.27.0

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: