# typing

Python API: `luxonis_train.typing`

The type aliases shared across the package.

## Attributes

### AttachIndexType

The outputs of the input node that a node reads.

The value is `"all"` for every output, or an integer index for one output. A tuple of two integers selects a range, and a third
integer sets the step of the range. `None` means that the index is not set.
[BaseNode.get_attached](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/base_node.md)
applies the index.

### Labels

The label tensors of a sample or a batch, keyed by task and label.

A key has the form `"<task_name>/<label>"`, where `<task_name>` is the dataset task. Examples are `"detection/boundingbox"` and
`"detection/metadata/id"`.

### Packet

A dictionary that maps output names to values or to lists of values.

The values of one packet are all tensors or all sizes. A node reads packets of tensors and returns a packet of tensors, for
example `{"features": [f1, f2]}`. A packet of sizes holds the shapes of the outputs of a node.

### T

`Tensor` or `Size`.

### View

`"train"`, `"val"`, or `"test"`.

[LuxonisModel](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/core/core.md)
builds one loader for each view.