Skip to content

Processes

ProgressTaskCallable

Bases: Protocol

Protocol for functions decorated with progress_task.

ResultProcess

Bases: PipelineProcess

ResultProcess is a class that handles the logging of images and their corresponding masks during the training, validation, and testing phases of a machine learning pipeline.

Attributes:

Name Type Description
progress_manager

An instance of the progress manager obtained from the manager.
total_epochs

The total number of epochs configured in wandb.
train_images_indices

A list of indices for the training images.
val_images_indices

A list of indices for the validation images.
test_images_indices

A list of indices for the test images.
datasets

The datasets containing training, validation, and test sets.
trainset

The training dataset.
valset

The validation dataset.
testset

The test dataset.

Methods: execute(): Executes the logging process for train, validation, and test images. _get_log_train_images() -> callable: Returns a function that logs the training images. _get_log_val_images() -> callable: Returns a function that logs the validation images. _get_log_test_images() -> callable: Returns a function that logs the test images. _log_image(image, mask, pred_mask, dataset, idx): Logs the image, predicted mask, and mask difference to wandb. skip() -> bool: Returns False indicating that this process should not be skipped.

__init__(controller, force, selected_images)

Initialize ResultProcess.

Parameters:

Name Type Description Default
controller Any

Controller/manager providing access to permanences.

 required
force bool

Whether to force execution.

 required
selected_images dict[str, list[int]]

Dictionary mapping dataset types to image indices.

 required