Skip to main content
Version: 0.94

snorkelflow.client.nodes.add_node_hierarchy

snorkelflow.client.nodes.add_node_hierarchy(application, input_node_uid, hierarchy_config)

Initiates a hierarchical classification DAG from an initial application.

Parameters​

NameTypeDefaultInfo
applicationUnion[str, int]The name or UID of the application.
input_node_uidintThe input node from where hierarchy will begin.
hierarchy_configDict[str, Dict]A dictionary config for the labels in the hierarchy.

Returns​

Dict[str – Dictionary detailing nodes created in the DAG. Semantics of the output are as follows - 1. Initial_Model -> uid of the model node of the initial application 2. Label str -> Dictionary which details nodes created for the branch a. Label_Filter: uid of the LabelFilter node which segregates datapoints as per a particular label for downstream classification, b. Column_Filler: uid of the ColumnFiller node which replaces existing gts with UNK to allow new labels, c. Model: uid of the Model node which classifies these datapoints into child labels 3. Concat -> uid of the Concat node which consolidates all labels at the end

Return type​

Any]

Examples​

>>> sf.add_node_hierarchy(
>>>     APP_NAME,
>>>     model_node_uid,
>>>     {
>>>         "1": {"1.1": {}, "1.2": {}},
>>>         "2": {"2.1": {}, "2.2": {}},
>>>     }
>>> )
{
    'nodes_output_dict':
    {'Initial_Model': {'Model': 525},
    '1': {'Label_Filter': 531, 'Column_Filler': 532, 'Model': 533},
    '2': {'Label_Filter': 536, 'Column_Filler': 537, 'Model': 538},
    '1.1': {'Label_Filter': 534},
    '1.2': {'Label_Filter': 535},
    '2.1': {'Label_Filter': 539},
    '2.2': {'Label_Filter': 540},
    'Concat': {'node_uid': 541, 'op_version_uid': 345}}
}