save_task_node_classification

gli.io.save_task_node_classification(name, description, feature, target, num_classes, train_set=None, val_set=None, test_set=None, train_ratio=0.8, val_ratio=0.1, test_ratio=0.1, num_samples=None, task_id=1, save_dir='.')

Save the node classification task information into task json and data files.

Parameters:

name (str) – The name of the dataset.
description (str) – The description of the task.
feature (list of str) – The list of feature names to be used in the task. The features could be node attributes, edge attributes, or graph attributes. For homogeneous graphs, the feature names should be in the format of “Node/{node_attr_name}”, “Edge/{edge_attr_name}”, or “Graph/{graph_attr_name}”. For heterogeneous graphs, the feature names should be in the format of “Node/{node_type}/{node_attr_name}”, “Edge/{edge_type}/{edge_attr_name}”, or “Graph/{graph_type}/{graph_attr_name}”. The node/edge/graph_attr_name, and the node/edge/graph_type should be the ones declared in the metadata.json file.
target (str) – The attribute name as prediction target in the task. For a node classification task, the attribute should be a node attribute. For homogeneous graphs, the attribute name should be in the format of “Node/{node_attr_name}”. For heterogeneous graphs, the attribute name should be in the format of “Node/{node_type}/{node_attr_name}”. The node_attr_name and the node_type should be the ones declared in the metadata.json file.
num_classes (int) – The number of classes in the task.
train_set (list/array of int or list of lists/2-d array of int) – A list of training node IDs or a list of list training node IDs. In the latter case, each inner list is the node IDs of one data split. If not None, train_ratio, val_ratio, and test_ratio will be ignored while val_set and test_set must present. If None, the task json file will store train_ratio, val_ratio, and test_ratio and random splits will be generated at run time. Default: None.
val_set (list/array of int or list of lists/2-d array of int) – A list of validation node IDs or a list of list validation node IDs. See train_set for more details. Default: None.
test_set (list/array of int or list of lists/2-d array of int) – A list of test node IDs or a list of list test node IDs. See train_set for more details. Default: None.
train_ratio (float) – The ratio of training nodes. See train_set for more details. Default: 0.8.
val_ratio (float) – The ratio of validation nodes. See train_set for more details. Default: 0.1.
test_ratio (float) – The ratio of test nodes. See train_set for more details. Default: 0.1.
num_samples (int) – The total number of nodes in the dataset. This needs to be provided if train_set, val_set, and test_set are not provided. Default: None.
task_id (int) – The task ID. This is needed when there are multiple tasks of the same task type are defined on the dataset. Default: 1.
save_dir (str) – The directory to save the task json and data files. Default: “.”.

Raises:

ValueError – If task_type is not “NodeRegression” or “NodeClassification”.
ValueError – If description is not a string.
ValueError – If feature is not a list of strings.
ValueError – If elements in feature do not correspond to node/edge/graph attributes.
ValueError – If target is not a string.
ValueError – If target does not correspond to a node/graph attribute.
ValueError – If num_classes is not None for regression tasks.
ValueError – If train_set, val_set, and test_set are not provided and num_samples is not provided.
ValueError – If train_set, val_set, and test_set are not provided and train_ratio, val_ratio, and test_ratio do not sum up to 1.
ValueError – If train_set, val_set, and test_set are not provided at the same time.
ValueError – If train_set, val_set, and test_set are provided but they are not lists or numpy arrays.
ValueError – If train_set, val_set, and test_set contain multiple splits but the split ratio of different splits is different.

Returns:

The dictionary of the content in the task json file.

Return type:

dict

Example

train_set = [0, 1]
val_set = [2, 3]
test_set = [4, 5]

# Save the task information.
save_task_node_classification(
    name="example_dataset",
    description="A node classification task for the example dataset.",
    feature=["Node/DenseNodeFeature", "Node/SparseNodeFeature"],
    target="Node/NodeLabel",
    num_classes=4,
    train_set=train_set,
    val_set=val_set,
    test_set=test_set)
# This function will save the task information into a json file named
# `task_node_classification_1.json` and one numpy data file storing the
# data splits, `train_set`, `val_set`, and `test_set`. The json file
# will look like the following.

{
    "description": "A node classification task for the example dataset.",
    "type": "NodeClassification",
    "feature": [
        "Node/DenseNodeFeature",
        "Node/SparseNodeFeature"
    ],
    "target": "Node/NodeLabel",
    "num_classes": 4,
    "train_set": {
        "file": "example_dataset__task_node_classification_1__4dcac617700f69a6dec06c2b5f75a246.npz",
        "key": "train_set"
    },
    "val_set": {
        "file": "example_dataset__task_node_classification_1__4dcac617700f69a6dec06c2b5f75a246.npz",
        "key": "val_set"
    },
    "test_set": {
        "file": "example_dataset__task_node_classification_1__4dcac617700f69a6dec06c2b5f75a246.npz",
        "key": "test_set"
    }
}