plmodules

BasicDetectionModule

Bases: pl.LightningModule

:class:PytorchLightning.LightningModule to use for binary semantic segmentation tasks.

Parameters:

Name	Type	Description	Default
model	nn.Module	underlying PyTorch model.	required
lr	float	learning rate.	required
wd	float	weight decay for AdamW optimizer.	required
scheduler_func	Optional[Callable]	Function that takes an optimizer as input and returns a scheduler dict formatted for PytorchLightning.	None
seg_metrics	Optional[Sequence[Metric]]	list of :class:torchmetrics.Metric segmentation metrics to compute on validation.	None
det_metrics	Optional[Sequence[Metric]]	list of :class:torchmetrics.Metric detection metrics to compute on validation.	None

BasicSegmentationModule

Bases: pl.LightningModule

:class:pytorch_lightning.LightningModule to use for binary semantic segmentation tasks.

Parameters:

Name	Type	Description	Default
model	nn.Module	underlying PyTorch model.	required
loss	nn.Module	loss function.	required
lr	float	learning rate.	required
wd	float	weight decay for AdamW optimizer.	required
scheduler_func	Optional[Callable]	Function that takes an optimizer as input and returns a scheduler dict formatted for PytorchLightning.	None
metrics	Optional[Sequence[Metric]]	list of :class:torchmetrics.Metric metrics to compute on validation.	None

get_scheduler_func(name, total_steps, lr)

Get a function that given an optimizer, returns the corresponding scheduler dict formatted for PytorchLightning <https://www.pytorchlightning.ai/>_.

Parameters:

Name	Type	Description	Default
name	str	name of the scheduler. Can either be "one-cycle", "cosine-anneal", "reduce-on-plateau" or "none".	required
total_steps	int	total number of training iterations, only useful for "one-cycle" or "cosine-anneal".	required
lr	float	baseline learning rate.	required

Returns:

Type	Description
Callable[[Optimizer], Optional[Dict[str, Union[str, _LRScheduler]]]]	Function that takes an optimizer as input and returns a scheduler dict formatted
Callable[[Optimizer], Optional[Dict[str, Union[str, _LRScheduler]]]]	for PytorchLightning.