# callbacks

Python API: `luxonis_train.callbacks`

Callbacks that run at points in the training loop.

An entry of `trainer.callbacks` in the config names a callback of the
[CALLBACKS](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/registry.md)
registry. Its `params` go to the constructor. The run does not build an entry whose `active` is false.

A run gets these callbacks without an entry in the config:

 * [GracefulInterruptCallback](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/graceful_interrupt.md)
   and
   [FailOnNoTrainBatches](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/fail_on_no_train_batches.md).
 * [LuxonisRichProgressBar](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/luxonis_progress_bar.md)
   when `rich_logging` is true, and
   [LuxonisTQDMProgressBar](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/luxonis_progress_bar.md)
   when it is false.
 * [TrainingManager](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/training_manager.md)
   and
   [LuxonisModelSummary](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/luxonis_model_summary.md).
 * A `ModelCheckpoint` on the lowest validation loss, and a second one on the main metric when the config has one.
 * [AIMETCallback](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/aimet_callback.md)
   when `exporter.aimet.active` is true.
 * A `GradientAccumulationScheduler` when `trainer.accumulate_grad_batches` is set and no such callback is in the list.

With `trainer.smart_cfg_auto_populate`, the config also adds the checkpoint upload, testing, and conversion callbacks when they
are missing.
[ConvertOnTrainEnd](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/convert_on_train_end.md)
replaces separate export and archive callbacks when it is configured explicitly.

## Child Pages

 * [aimet_callback](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/aimet_callback.md)
 * [archive_on_train_end](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/archive_on_train_end.md)
 * [convert_on_train_end](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/convert_on_train_end.md)
 * [ema](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/ema.md)
 * [export_on_train_end](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/export_on_train_end.md)
 * [fail_on_no_train_batches](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/fail_on_no_train_batches.md)
 * [gpu_stats_monitor](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/gpu_stats_monitor.md)
 * [graceful_interrupt](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/graceful_interrupt.md)
 * [gradcam_visualizer](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/gradcam_visualizer.md)
 * [luxonis_model_summary](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/luxonis_model_summary.md)
 * [luxonis_progress_bar](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/luxonis_progress_bar.md)
 * [metadata_logger](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/metadata_logger.md)
 * [needs_checkpoint](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/needs_checkpoint.md)
 * [test_on_train_end](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/test_on_train_end.md)
 * [training_manager](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/training_manager.md)
 * [training_progress_callback](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/training_progress_callback.md)
 * [upload_checkpoint](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/callbacks/upload_checkpoint.md)