`nexuml.tracking.logger`¶

nexuml.tracking.logger ¶

Logger factory for NexuML experiment tracking.

get_temp_artifact_root ¶

get_temp_artifact_root() -> Path

Return the directory used for temporary artifact staging.

staged_artifact_path ¶

staged_artifact_path(
    artifact_name: str, *, prefix: str = "nexuml_artifact_"
) -> Iterator[Path]

Yield a temporary filesystem path for building an artifact before logging it.

create_loggers ¶

create_loggers(
    logging_spec: "LoggingSpec | None",
    run_name: str | None = None,
) -> list[Any]

Build a list of Lightning loggers from a LoggingSpec.

Returns an empty list (disables logging) if spec is None. Each backend is imported lazily so missing optional deps don't crash the import.

Parameters:

Name	Type	Description	Default
`logging_spec`	`'LoggingSpec \| None'`	LoggingSpec instance or None.	required
`run_name`	`str \| None`	Override run name; falls back to logging_spec.run_name.	`None`

Returns:

Type	Description
`list[Any]`	List of Lightning logger instances ready for Trainer(logger=...).

iter_loggers ¶

iter_loggers(logger_obj: Any) -> list[Any]

Normalize Lightning logger containers into a flat list.

Returns:

Type	Description
`list[Any]`	Flat list of non-None logger instances.

log_artifact ¶

log_artifact(
    logger_obj: Any,
    source_path: str | Path,
    artifact_path: str | None = None,
) -> None

Log or copy a file artifact to all configured logger backends.

Raises:

Type	Description
`FileNotFoundError`	If source_path does not exist.

log_text_artifact ¶

log_text_artifact(
    logger_obj: Any,
    text: str,
    artifact_name: str,
    artifact_path: str | None = None,
) -> None

Write text content to a temporary file and log it as an artifact.

log_image ¶

log_image(
    logger_obj: Any,
    tag: str,
    image: Any,
    step: int | None = None,
    artifact_path: str | None = None,
) -> None

Log an image to all configured logger backends.

Parameters:

Name	Type	Description	Default
`logger_obj`	`Any`	Lightning logger or list of loggers.	required
`tag`	`str`	Name/tag for the image (e.g. `"val/reconstruction"`). Slashes are used as subdirectory separators for file-based backends.	required
`image`	`Any`	Image data — numpy array (HWC, uint8 or float [0,1]), PIL Image, or torch tensor (CHW, uint8 or float [0,1]).	required
`step`	`int \| None`	Global step for time-series backends (TensorBoard).	`None`
`artifact_path`	`str \| None`	Optional subdirectory within the run's artifact store (MLflow / file-based backends only).	`None`

print_service_info ¶

print_service_info(
    trainer_loggers: list[Any],
    scenario_name: str,
    log_dir: str | Path = ".experiments",
    logging_spec: Any | None = None,
    tuning_spec: Any | None = None,
    data_backend: str | None = None,
    training_backend: str | None = None,
) -> None

Print actionable commands for all active monitoring services.

Called once at the start of training so the user knows how to inspect metrics, tune hyperparameters, etc.

nexuml.tracking.logger¶