Is there a fast training shortcut for unit testing in mmdetection similar to PyTorch Lightning's Fast Dev Run?

[ryan-minato]

Is there a shortcut similar to PyTorch Lightning's Fast Dev Run for mmdetection that allows for a quick round of training for unit testing without saving records and checkpoints?
If so, could you please provide relevant information and documentation?

like this: https://lightning.ai/docs/pytorch/stable/common/trainer.html#fast-dev-run

[narthchin]

You can follow the debug tricks here:
https://github.com/open-mmlab/mmengine/blob/main/docs/en/common_usage/debug_tricks.md
Set indices in your dataset config to use a relatively small number of samples for debugging, note that in order to use this feature, your dataset has to be inherited from BaseDataset. And for the checkpoints, you can simply remove CheckpointHook in your hook config.