[Example Request] NVIDIA DALI LightningDataModule

[austinmw]

📓 New <Tutorial/Example>

Is this a request for a tutorial or for an example?
Either really.

What is the task?
Object detection

Is this example for a specific model?
Any detection model

Is this example for a specific dataset?
Any LightningDataModule that uses nvidia DALI. Maybe COCO?

I have a LightningDataModule that uses NVIDIA DALI. The train and val dataloaders return a single item, batch, which is a dictionary that contains the following items:

batch['images'] - a batch of preprocessed tensor images, for example of size (batch_size, 3, 416, 416)
batch['bboxes'] - a padded tensor of bboxes, with format x,y,x2,y2, for example of size (batch_size, 200, 4)
batch['labels'] - a padded tensor of class labels, for example of size (batch_size, 200)

I'm not sure how to overwrite training_step and validation_step to incorporate a dali datamodule like this. Especially, what is records in (xb, yb), records = batch, and what do I need to do with self.convert_raw_predictions?

---

Don't remove
Main issue for examples: #39