gigl.experimental.knowledge_graph_embedding.lib.config.parsing#

Functions#

build_modeling_and_resource_config_from_args(mode)

Build the modeling and resource config from command line arguments.

Module Contents#

gigl.experimental.knowledge_graph_embedding.lib.config.parsing.build_modeling_and_resource_config_from_args(mode)[source]#

Build the modeling and resource config from command line arguments.

Parses command line arguments to extract configuration URIs, loads the modeling configuration using Hydra, and loads the resource configuration. Handles both local and GCS URI sources, automatically downloading remote configs when needed.

Parameters:

mode (Literal['enumerating', 'training']) – The execution mode, either “enumerating” (for data preprocessing) or “training” (for model training). Determines fallback config paths.

Returns:

A tuple containing:

  • Applied task identifier for the current job

  • Hydra modeling configuration (training hyperparameters, model config, etc.)

  • Resource configuration wrapper (project info, compute resources, etc.)

Return type:

Tuple[AppliedTaskIdentifier, DictConfig, GiglResourceConfigWrapper]

Raises:

ValueError – If an unsupported mode is provided.