Luxonis-train

New Contributors
* HonzaCuhel made their first contribution in https://github.com/luxonis/luxonis-train/pull/134

**Full Changelog**: https://github.com/luxonis/luxonis-train/compare/v0.2.0-beta...v0.2.1-beta

What's Changed
* Fix binary segmentation export by HonzaCuhel in https://github.com/luxonis/luxonis-train/pull/134
* Keypoint M predefined model by JSabadin in https://github.com/luxonis/luxonis-train/pull/136
* README.md update - link to the tutorials by HonzaCuhel in https://github.com/luxonis/luxonis-train/pull/135
* Custom optimizer by JSabadin in https://github.com/luxonis/luxonis-train/pull/132
* fix: training strategy docs by JSabadin in https://github.com/luxonis/luxonis-train/pull/137
* FOMO detection head by JSabadin in https://github.com/luxonis/luxonis-train/pull/133
* Refactor Archiver and add BaseHead by sokovninn in https://github.com/luxonis/luxonis-train/pull/139
* Feat: Added new rule in config auto-population and updated documentation. by JSabadin in https://github.com/luxonis/luxonis-train/pull/140
* Feat/confusion matrix metric by JSabadin in https://github.com/luxonis/luxonis-train/pull/138
* Better pretrained weight loading by klemen1999 in https://github.com/luxonis/luxonis-train/pull/142

**Full Changelog**: https://github.com/luxonis/luxonis-train/compare/v0.1.0-beta...v0.2.0-beta

What's Changed
* Removed GitHub version of `luxonis-ml` from `requirements-dev.txt` by kozlov721 in https://github.com/luxonis/luxonis-train/pull/109
* Add export output names specification for individual nodes by sokovninn in https://github.com/luxonis/luxonis-train/pull/112
* Add support for multi-label visualization by JSabadin in https://github.com/luxonis/luxonis-train/pull/110
* SequentialLR Scheduler by JSabadin in https://github.com/luxonis/luxonis-train/pull/111
* Add OHEM losses by sokovninn in https://github.com/luxonis/luxonis-train/pull/117
* Add model EMA (Exponential Moving Average) by sokovninn in https://github.com/luxonis/luxonis-train/pull/114
* Add dynamic aliases for predefined models by sokovninn in https://github.com/luxonis/luxonis-train/pull/118
* Mixed Precision Training by JSabadin in https://github.com/luxonis/luxonis-train/pull/113
* Add gradient visualization during validation by sokovninn in https://github.com/luxonis/luxonis-train/pull/115
* Add tests action with a manual trigger by sokovninn in https://github.com/luxonis/luxonis-train/pull/123
* Fix tests action by sokovninn in https://github.com/luxonis/luxonis-train/pull/124
* Force reinstall of luxonis-ml in tests action by sokovninn in https://github.com/luxonis/luxonis-train/pull/126
* Config smart auto populate by JSabadin in https://github.com/luxonis/luxonis-train/pull/127
* Updated docs action by kozlov721 in https://github.com/luxonis/luxonis-train/pull/120
* feat: add n_validation_batches option to trainer by JSabadin in https://github.com/luxonis/luxonis-train/pull/125
* Updated how we access config version from luxonis-ml by klemen1999 in https://github.com/luxonis/luxonis-train/pull/122
* Unsupervised Anomaly Detection by JSabadin in https://github.com/luxonis/luxonis-train/pull/116
* fix: call tracker.close() by JSabadin in https://github.com/luxonis/luxonis-train/pull/119
* Run inference on the same device as used in the Trainer by sokovninn in https://github.com/luxonis/luxonis-train/pull/121
* Pretrained weights for detection-light and segmentation-light by klemen1999 in https://github.com/luxonis/luxonis-train/pull/128
* add test for cli commands by JSabadin in https://github.com/luxonis/luxonis-train/pull/129
* added `DDRNet` to archive by kozlov721 in https://github.com/luxonis/luxonis-train/pull/130

What's Changed - Highlights

- Full support for multi-input and multi-task models
- Multi Input Support by CaptainTrojan in 36
- Full Custom Task Support by kozlov721 in 53
- Add video inference by sokovninn in 84
- New Keypoint Heads and Losses by JSabadin in 40
- Generate NN archive from training configs by klemen1999 in 17
- Tuning - Augmentation Subsets Support by jkbmrz in 35
- Train-Only Heads by JSabadin in 83
- Code Cleanup and Improved Tests by kozlov721in 69
- Tensor Core Float16 Precision by kozlov721in 24