A set of notebooks as a guide to the process of fine-grained image classification of birds species, using PyTorch based deep neural networks.
Issue with Github Actions uploading correct package release to PYPI. So new version and local build will be pushed to PYPI.
Minor updates to Torchscript model file output handling. Some changes to how CLI arguments are handled by scripts.
Issues with auto build to PYPI, with module not being found installing from PYPI. So, will be manually building and pushing to PYPI for now. Removed build on publish release for now.
Issue with package on previous release. Very minor change to output from PYPI Github build package and publish pipeline. No functional changes to package.
Patch on args.py module of cub_tools to allow training scripts to be executed without the need for a configuration YAML file. Training arguments will be the default, and can be overridden purely command line argument key value pairs after the --opts command line flag.
E.g.
python train_script.py --opts MODEL.NAME resnet34 MODEL.MODEL_LIBRARY pytorchcv
Added updates to documentation, including installation notes to package. Added conda_dependencies yml file for environment creation.
Changes to the auto publishing github pipeline. Publishing of a new release now results in the package being published to PYPI as an updated package release.
Tagged at v2.5.0 for first official PyPI package release.
Minor changes to setup.py package requirements and Triton deployment scripts.
Latest release with various small fixes. Stamped version for use with external AzureML development.