gunz_ml.integrations package

Submodules

gunz_ml.integrations.hydra module

Provides helper functions for working with Hydra configurations.

This module includes utilities for initializing and validating Hydra configurations and for resolving OmegaConf objects into standard Python dicts.

gunz_ml.integrations.hydra.init_hydra_and_check_config(cfg: DictConfig, script_name: str | None = None, check_script_name: bool = True, check_paths: bool = True, allow_unresolved_keys: bool = False) → HydraConfig[source]

Initializes and validates the Hydra configuration for an experiment.

Parameters:

  • cfg (DictConfig) – The configuration object provided by Hydra.

  • script_name (str | None, optional) – The expected name of the main script. If None, it’s inferred from the Hydra config. Defaults to None.

  • check_script_name (bool, optional) – If True, validates that the experiment config matches the script name. Defaults to True.

  • check_paths (bool, optional) – If True, ensures that all paths in cfg.path exist. If False, it creates them instead. Defaults to True.

  • allow_unresolved_keys (bool, optional) – If False, raises an error if any keys in the config are missing (i.e., have a value of ‘???’). Defaults to False.

Returns:

The active Hydra configuration object.

Return type:

HydraConfig

Raises:

  • ValueError – If the experiment configuration is invalid for the current script.

  • RuntimeError – If allow_unresolved_keys is False and missing keys are found.

gunz_ml.integrations.hydra.resolve_cfg(cfg: DictConfig | None, default_to_empty_dict: bool = False) → dict | None[source]

Resolves an OmegaConf DictConfig object into a standard Python dictionary.

Parameters:

  • cfg (DictConfig | None) – The configuration object to resolve.

  • default_to_empty_dict (bool, optional) – If True, returns an empty dict if cfg is None. If False, returns None. Defaults to False.

Returns:

The resolved configuration as a dictionary, or None.

Return type:

dict | None

gunz_ml.integrations.lightning module

Provides helper functions and constants for PyTorch Lightning integration.

This module includes utilities for managing common PyTorch Lightning warnings and defines standardized status enums for logging purposes.

class gunz_ml.integrations.lightning.FinalizeStatus(value)[source]

Bases: StrEnum

Enumeration for the final status of a run or trial.

FAILED = 'failed'
FINISHED = 'finished'
SUCCESS = 'success'
gunz_ml.integrations.lightning.ignore_pl_warnings(dataloader_num_workers: bool = True, slurm_srun: bool = True, mixed_precision: bool = True)[source]

Suppresses common, often noisy, warnings from PyTorch Lightning globally.

Parameters:

  • dataloader_num_workers (bool, optional) – If True, suppresses the warning about using a small number of workers in the DataLoader. Defaults to True.

  • slurm_srun (bool, optional) – If True, suppresses the warning about the srun command being available on the system. Defaults to True.

  • mixed_precision (bool, optional) – If True, suppresses the historical usage warning for 16-bit mixed precision. Defaults to True.

gunz_ml.integrations.lightning.suppress_pl_warnings(dataloader_num_workers: bool = True, slurm_srun: bool = True, mixed_precision: bool = True)[source]

A context manager to temporarily suppress common PyTorch Lightning warnings.

Parameters:

  • dataloader_num_workers (bool, optional) – If True, suppresses the warning about using a small number of workers in the DataLoader. Defaults to True.

  • slurm_srun (bool, optional) – If True, suppresses the warning about the srun command being available on the system. Defaults to True.

  • mixed_precision (bool, optional) – If True, suppresses the historical usage warning for 16-bit mixed precision. Defaults to True.

gunz_ml.integrations.mlflow module

gunz_ml.integrations.optuna module

gunz_ml.integrations.timm module

gunz_ml.integrations.toml module