GOOD.data.dataset_manager

A module that is consist of a dataset loading function and a PyTorch dataloader loading function.

Functions

create_dataloader(dataset, config)

Create a PyG data loader.

load_dataset(name, config)

Load a dataset given the dataset name.

read_meta_info(meta_info, config)

GOOD.data.dataset_manager.create_dataloader(dataset, config: Union[CommonArgs, Munch])[source]

Create a PyG data loader.

Parameters

  • loader_name –

  • dataset – A GOOD dataset.

  • config – Required configs: config.train.train_bs config.train.val_bs config.train.test_bs config.model.model_layer config.train.num_steps(for node prediction)

Returns

A PyG dataset loader.

GOOD.data.dataset_manager.load_dataset(name: str, config: Union[CommonArgs, Munch]) → dir[source]

Load a dataset given the dataset name.

Parameters

  • name (str) – Dataset name.

  • config (Union[CommonArgs, Munch]) – Required configs: config.dataset.dataset_root config.dataset.domain config.dataset.shift_type config.dataset.generate

Returns

A dataset object and new configs

  • config.dataset.dataset_type

  • config.model.model_level

  • config.dataset.dim_node

  • config.dataset.dim_edge

  • config.dataset.num_envs

  • config.dataset.num_classes

GOOD.data.dataset_manager.read_meta_info(meta_info, config: Union[CommonArgs, Munch])[source]