Deci-AI/super-gradients: 3.6.0

What's Changed

• Fix numpy deprecation warning on creating ragged array when using _pad_image by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1632
• Allow setting per-layer learning rates" by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1612
• Add skip_resize for model.predict by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1605
• Add documentation for DatasetAdapter by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1591
• Notebooks update by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1634
• Add custom dataset notebook by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1592
• Added support of multiple test loaders in train_from_config by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1641
• Improve LR and WD assignment message on training start by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1639
• [Feature] lazy implementation of video predictions by @hakuryuu96 in https://github.com/Deci-AI/super-gradients/pull/1621
• Add automatic model transition to CUDA if FP16 export is requested by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1643
• Fix Invalid syntax at convert_recipe_to_code.py by @seunghalee1226 in https://github.com/Deci-AI/super-gradients/pull/1642
• Prevent modification of the input argument (targets) in DetectionVisualization.visualize_batch by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1652
• Fix Data Adapter for Segmentation by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1654
• Add notebook for fine-tuning YoloNAS with QAT by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1638
• New SG version to reflect bugfixes in DG by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1657
• Introduce DetectionSample dataclass by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1635
• Use onnxsim and not onnx-simplifier (Same package but different name) by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1668
• OhemLoss: bug fix by @danielafrimi in https://github.com/Deci-AI/super-gradients/pull/1669
• updated condition to work on rank0 since only for it context.sg_logger exists by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1683
• Segmentation sample + albumentation support for segmentation by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1678
• [Enhancement] Implement Distance-Based Detection Matching in DetectionMetrics by @DimaBir in https://github.com/Deci-AI/super-gradients/pull/1463
• fixed LegacySegmentationTransformMixin docs by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1685
• Change default set_device value, which currently may launch DDP by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1698
• Fix multigpu=None with device=cpu by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1701
• Adjust bbox thickness and text to bbox size by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1687
• Finetune training hyperparameter by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1695
• Bugfixes in DistanceBasedDetectionMetrics and few improvements in DetectionMetrics by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1696
• Ensure that ImagePermute is present in processings by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1702
• Fix ExtremeBatchDetectionVisualizationCallback when multiscale collate fn in use by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1697
• fix dataset plot and improve the plotting by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1686
• Update predict + detection TL notebooks and update makefile/readme by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1637
• Feature/sg 1277 depth estimation dataset support by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1705
• A proposal API for checking the input to be compatible with the model by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1704
• predict() support for segmentation models and more by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1708
• Add pose model by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1593
• Feature/sg 1246 remove deprecations of 3.6.0 by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1714
• Fixed bug in DetectionMixup implementation that prevents YoloXTrainingStageSwitchCallback to work as intended by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1719
• Feature/sg 1277 Depth Estimation Metrics Support by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1709
• Fixed typo in variable name in exception message by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1726
• Fixed pipelines returning ImagesPredictions instead of ImagePrediction for single image by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1724
• Ensure that class names are trivial container in DetectionDataset by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1729
• removed entries from averaged state dict by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1718
• Update pre-trained models badge URL by @gasparitiago in https://github.com/Deci-AI/super-gradients/pull/1728
• PPYoloELoss not does not require reg_max parameter by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1727
• bring the tests back by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1742
• added gha trigger to release and put the unless condition under the same 'when' condition as the release by @Yonatan-Kaplounov in https://github.com/Deci-AI/super-gradients/pull/1744
• Removed leftover albumentations transform from regseg recipe by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1745
• PPYoloECollate to support random_aspect_ratio parameter by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1737
• Feature/sg 1196 fix breaking change test by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1720
• Allow passing additional params to .predict() for detection models by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1738
• Feature/sg 1271 add keypoint aug by @Louis-Dupont in https://github.com/Deci-AI/super-gradients/pull/1710
• Add support for resolving strings in PhaseCallback by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1747
• added albumentations tutorial notebook by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1743
• Improve YoloNAS model flexibility by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1749
• Added albumentations by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1750
• Fixed an issue where itertools.product in PPYoloECollateFN needed list() by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1754
• Fix import issue in convert_recipe_to_code by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1751
• Improve DetectionMetrics implementation to properly handle score_thresholds by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1752
• allow removing anti_alias and adding downsample_mode to update code by @NatanBagrov in https://github.com/Deci-AI/super-gradients/pull/1741
• Fix potential unwanted simplification of ONNX graph when exported model with random weights by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1757
• fixed by casting training params and one hot dtype fixed by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1758
• Added test_enum_factory unit test and fixed implementation of TypeEnumFactory by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1765
• changes logic from only one to exactly one parameter for step lr by @NatanBagrov in https://github.com/Deci-AI/super-gradients/pull/1756
• LR assignement tutorial by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1759
• updated returned dtype from predict by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1748
• Added notebook showing how to use YoloNAS with TensorRT by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1767
• Bugfix/infra 2429 fix ci by @ranrubin in https://github.com/Deci-AI/super-gradients/pull/1778
• Fixes for 3.6.0 release by @shaydeci in https://github.com/Deci-AI/super-gradients/pull/1779
• Bump SG version by @BloodAxe in https://github.com/Deci-AI/super-gradients/pull/1774

New Contributors

• @seunghalee1226 made their first contribution in https://github.com/Deci-AI/super-gradients/pull/1642
• @DimaBir made their first contribution in https://github.com/Deci-AI/super-gradients/pull/1463
• @gasparitiago made their first contribution in https://github.com/Deci-AI/super-gradients/pull/1728
• @Yonatan-Kaplounov made their first contribution in https://github.com/Deci-AI/super-gradients/pull/1744

Full Changelog: https://github.com/Deci-AI/super-gradients/compare/3.4.1...3.6.0