torchwrench.extras.tensorboard module¶

class torchwrench.extras.tensorboard.TensorboardEvent[source]¶

Bases: TypedDict

dtype : str¶

float_val : NotRequired[List[float]]¶

step : int¶

string_val : NotRequired[str]¶

tag : str¶

value : str | float¶

wall_time : float¶

torchwrench.extras.tensorboard.get_tfevents_duration(fpath: str | Path, verbose: int = 0) → float[source]¶

Return time elapsed between first and last log in a tensorboard event file.

torchwrench.extras.tensorboard.load_tfevents(fpath: str | Path, *, cast_float_and_str: bool = True, ignore_underscore_tags: bool = True, verbose: int = 0) → list[TensorboardEvent][source]¶

Args:

fpath: File path to a tensorboard event file. cast_float_and_str: Cast string to floats and store result in ‘value’ field. defaults to True. ignore_underscore_tags: Ignore event when tag starts with an underscore. defaults to True. verbose: Verbose level. Higher value means more log messages. defaults to 0.

torchwrench.extras.tensorboard.load_tfevents_files(paths_or_patterns: str | Path | Iterable[str | Path], *, cast_float_and_str: bool = True, ignore_underscore_tags: bool = True, verbose: int = 0) → dict[str, list[TensorboardEvent]][source]¶

Args:

paths_or_patterns: Path or glob patterns to multiple files. cast_float_and_str: Cast string to floats and store result in ‘value’ field. defaults to True. ignore_underscore_tags: Ignore event when tag starts with an underscore. defaults to True. verbose: Verbose level. Higher value means more log messages. defaults to 0.