Logs the current training status. :param phase_runtime: The wall-clock time of the current phase. :type phase_runtime: float

Copies the dataset and returns the rootpath.

Evaluates the model on the given data set. :param data: string indicating the data set to evaluate on. Can be 'train' or 'val'.

Resets all metrics

Creates the fixval plots and logs them to wandb.

Returns the metrics for the current epoch. :return: dict containing the metrics :rtype: dict

Returns the model. :param reinit: If True, the model is reinitialized. :type reinit: bool :param model_path: Path to

Returns the optimizer. :param initial_lr: The initial learning rate :type initial_lr: float :return: The optimizer.

Controls the execution of the script.

Saves the model's state_dict to a file. :param model_identifier: Name of the file type. :type model_identifier: str :

Sets the seed for the current run. :param seed: seed to be used

Update config with default values recursively.

Update the list of extreme values (either largest or smallest) based on the new batch.

Initialize useful variables using config. :param config: wandb run config :type config: wandb.config.Config :param de

Constructor for the L1 loss function with shifted GEDI labels up to 10 meters in each direction. :param ignore_value: The value to ig

Forward pass of the loss function. :param predictions: The predictions of the network. :param labels: The labels of the data.

Applies the L1 loss :param out: output of the network :param target: target :return: l1 loss

Applies the Huber loss :param out: output of the network :param target: target :return Huber loss

Applies the L2 loss :param out: output of the network :param target: target :return L2 loss

Creates a density scatter plot visualization for the regression model :param process_variables: A function that takes in inputs, labels, and

Get a visualization function that plots the input and output of the model. :param process_variables: A function that processes the variables

Creates a boxplot visualization for the regression model :param process_variables: A function that takes in inputs, labels, and outputs and r