Janusz Lisiecki

Fix inverted mmap inside webdataset reader, cleans copy_read_data_

3

- fix the wrong ussage of mmap when the user is not asking for it inside webdataset - cleans up usage of copy_read_data_ ## Category: **Bug fix** (*non-breaking change which...

Make sure that the proper video stream index is used by the GPU decoder

6

- fixes the use of 0 index stream inside the GPU decoder by properly obtained index - related to https://github.com/NVIDIA/DALI/issues/5680 ## Category: **Bug fix** (*non-breaking change which fixes an issue*)...

Add an ability to rewind at the end of the video

3

- in some cases after seeking to the last key frame in the video leads to the premature end of decoding as the decoding doesn't start and yet DALI waits...

Turn off SHARP for DALI L2 tests

4

- due some problems with CI runners turn of SHARP that is not needed for the tests ## Category: **Other** (*e.g. Documentation, Tests, Configuration*) ## Description: - due some problems...

Add PyTorch DataLoader Evaluator plugin

28

- Introduces a lightweight diagnostic tool for identifying data loading bottlenecks in PyTorch training pipelines. - This change adds Loader Evaluator inside pytorch DALI plugin, a jupyter notebook tutorial, and...

Add dynamic mode RN50 pipeline to hw decoder bench

3

- Adds NDDRN50Pipeline class using nvidia.dali.experimental.dynamic API - Adds 'ndd_rn50' pipeline option to benchmark script - Supports NDD decoders module selection (experimental vs standard) ## Category: **New feature** (*non-breaking change...

Refactor package naming: move from flavor-based names to version-based dev differentiation

7

- Previously, weekly/nightly builds were differentiated by flavor suffix in the package name (e.g., nvidia-dali-weekly-cuda*, nvidia-dali-nightly-cuda*). This change moves to a unified naming schema where: - Package names are consistent...