pipeline_config

Functions

`adapt_runtime_hydra_config`	Fill runtime Hydra fields from the unified config boundary.
`canonical_stage_name`
`load_runtime_hydra_config`	Reconstruct the instantiated Hydra config used by GPU-heavy stages.
`normalize_pipeline_config`	Return the strict Puzzletron stage-runner config shape.
`pipeline_config_from_path`	Load a Hydra YAML and attach runtime metadata for stage handlers.
`rebase_authored_pipeline_config`	Bind an intentionally derived worker config as its own authored view.

adapt_runtime_hydra_config(hydra_cfg, config)

Fill runtime Hydra fields from the unified config boundary.

This adapter exists only at the stage boundary. It keeps the new canonical Puzzletron config clean while allowing the GPU-heavy stage implementations to keep their established Hydra section names.

Parameters:

hydra_cfg (DictConfig)
config (dict[str, Any])

Return type:

DictConfig

canonical_stage_name(stage)

Parameters:: stage (str)
Return type:: str

load_runtime_hydra_config(config)

Reconstruct the instantiated Hydra config used by GPU-heavy stages.

Parameters:: config (dict[str, Any])
Return type:: DictConfig

normalize_pipeline_config(config)

Return the strict Puzzletron stage-runner config shape.

The heavy implementations still consume the Hydra runtime config. This normalizer is the boundary where older experiment YAMLs and the new experiment/model/search_space shape are folded into one manifest-friendly dictionary.

Parameters:: config (DictConfig | dict[str, Any])
Return type:: dict[str, Any]

pipeline_config_from_path(config_path, *, overrides=None, num_nodes=1, node_index=0)

Load a Hydra YAML and attach runtime metadata for stage handlers.

Parameters:

config_path (str | Path)
overrides (list[str] | None)
num_nodes (int)
node_index (int)

Return type:

dict[str, Any]

rebase_authored_pipeline_config(config)

Bind an intentionally derived worker config as its own authored view.

Parameters:: config (dict[str, Any])
Return type:: dict[str, Any]