Initializes multi-GPU-friendly python logger.

Safely get Weights&Biases logger from Trainer.

Args: year_string (str): must only contain {digits, '-', '+'}. Examples: '1988-90' will return [1988, 1989, 1990] '19

input_normalization (str): "z" for z-scaling (zero mean and unit standard deviation)

Args: config (DictConfig): A OmegaConf config (e.g. produced by hydra yaml config file parsing) Returns: A tuple of (model, d

Args: input_dim (int): the expected 1D input tensor dim output_activation_function (str, bool, optional): By default

Args: readout_which_output: Which graph part to return (default: edges), can be {EDGES, NODE

Args: results: The list that pl.Trainer() returns when predicting, i.e. results = trainer.predict(model, dat

N - the batch/data dimension size L - the number of levels (= number of layers + 1) Args: upwelling_flux: a (N, L) array downw

Args: config (DictConfig): A OmegaConf config (e.g. produced by hydra yaml config file parsing) Returns: A datamodule that yo

This is a prefix for naming the runs for a more agreeable logging.

Args: config (DictConfig): A OmegaConf config (e.g. produced by hydra yaml config file parsing) Returns: The model that you c

:param lon_plot: :param lat_plot: :param errors: Note that if plot_type is 2D, i.e is in ['heatmap', 'contour'], it has to have shape la

Returns: A (b, #layers, d) array, where d = #layer-feats + #global-feats (+ #levels-feats, if not drop_levels)