module emote.mixins.logging

Classes

class LoggingMixin:

A Mixin that accepts logging calls. Logged data is saved on this object and gets written by a Logger. This therefore doesn't care how the data is logged, it only provides a standard interface for storing the data to be handled by a Logger.

Methods

def __init__(self, *default_window_length) -> None

Arguments:

  • default_window_length(int)
def log_scalar(self, key, value) -> None

Use log_scalar to periodically log scalar data.

Arguments:

  • key(str)
  • value(float | torch.Tensor)
def log_windowed_scalar(self, key, value) -> None

Log scalars using a moving window average. By default this will use default_window_length from the constructor as the window length. It can also be overridden on a per-key basis using the format windowed[LENGTH]:foo/bar. Note that this cannot be changed between multiple invocations - whichever length is found first will be permanent.

Arguments:

  • key(str)
  • value(float | torch.Tensor | Iterable[torch.Tensor | float])
def log_image(self, key, value) -> None

Use log_image to periodically log image data.

Arguments:

  • key(str)
  • value(torch.Tensor)
def log_video(self, key, value) -> None

Use log_scalar to periodically log scalar data.

Arguments:

  • key(str)
  • value(Tuple[np.ndarray, int])
def log_histogram(self, key, value) -> None

def state_dict(self) -> None

def load_state_dict(
    self,
    state_dict,
    load_network,
    load_optimizer,
    load_hparams
) -> None