dlprog

Deep Learning Progress

A Python library for progress bars with the function of aggregating each iteration’s value. It helps manage the loss of each epoch in deep learning or machine learning training.

source and usage: misya11p/dlprog

class dlprog.Progress(n_iter=None, n_epochs=None, label=None, n_values=1, agg_fn='mean', momentum=None, width=40, leave_freq=1, unit=1, defer=False, note='', symbol='#', round=5, sep_label=': ', sep_values=', ', sep_note=', ', store_all_values=True, store_all_times=True)

Bases: object

Progress bar class. When the following attributes are None, the progress bar is not displayed correctly. Attributes defined in this constructor will be default values.

Parameters:

n_iter (int) – Number of iterations per epoch. Defaults to None.
n_epochs (int) – Number of epochs. Defaults to None.
label (Optional[Union[str, List[str]]]) – Label for progress bar. If you want to use multiple values, input labels as an iterable. Defaults to None.
n_values (int) – Number of values to be aggregated. If this number differs from the number of labels, the number of labels is used. Use this when you want to set None to label or when you want to use multiple values with the same label. Defaults to 1.
agg_fn (Union[str, Callable[[Number, Number], Number]]) – Aggregation function for epoch value with weight. Defaults to ‘mean’.
momentum (float) – Momentum for updating values. If None, the value is updated by the aggregation function. If set, the value is updated by the exponential moving average, agg_fn and weight are ignored. Defaults to None.
width (int) – Width of progress bar. Defaults to 40.
leave_freq (int) – Frequency of leaving the progress bar. If <= 0, none are left. Defaults to 1.
unit (int) – Unit of progress bar (epoch). Defaults to 1.
defer (bool) – If True, auto-step will be deferred until the next memo() call. Use when you want to update a note at the end of the epoch. Defaults to False.
note (str) – Note for progress bar. Defaults to ‘’.
symbol (str) – Symbol for progress bar. Defaults to ‘#’.
round (int) – Number of digits to round to. Default is 5.
sep_label (str) – Separator character for value and label. Defaults to ‘: ‘.
sep_values (str) – Separator character for values. Defaults to ‘, ‘.
sep_note (str) – Separator character for note. Defaults to ‘, ‘.
store_all_values (bool) – If True, store values of all iterations. Set to False if you want to reduce memory usage. Defaults to True.
store_all_times (bool) – If True, store times of all iterations. Set to False if you want to reduce memory usage. Defaults to True.

set_defaults(**kwargs): Modify default values.

reset(**kwargs): Reset all attributes.

now_values(no_unpacked=False)

Get current epoch values. Note that the value may differ from what is displayed on the progress bar.

Parameters:: no_unpacked (bool) – If False, return a unpacked value if n_values=1. Defaults to False.
Returns:: Current value if no_unpacked=True, else list of values.
Return type:: Union[Number, List[Number]]

start(**kwargs)

Start running. Initialize start time and epoch. You can set the attributes to be used at this runtime. If not set, the default value is used. Arguemnts is the same as the constructor.

Parameters:

n_iter (int) – Number of iterations per epoch. Defaults to None.
n_epochs (int) – Number of epochs. Defaults to None.
label (Optional[Union[str, List[str]]]) – Label for progress bar. If you want to use multiple values, input labels as an iterable. Defaults to None.
n_values (int) – Number of values to be aggregated. If this number differs from the number of labels, the number of labels is used. Use this when you want to set None to label. Defaults to 1.
agg_fn (Union[str, Callable[[Number, Number], Number]]) – Aggregation function for epoch value with weight. Defaults to ‘mean’.
momentum (float) – Momentum for updating values. If None, the value is updated by the aggregation function. If set, the value is updated by the exponential moving average, agg_fn and weight are ignored. Defaults to None.
width (int) – Width of progress bar. Defaults to 40.
leave_freq (int) – Frequency of leaving the progress bar. If <= 0, none are left. Defaults to 1.
unit (int) – Unit of progress bar (epoch). Defaults to 1.
defer (bool) – If True, auto-step will be deferred until the next memo() call. Use when you want to update a note at the end of the epoch. Defaults to False.
note (str) – Note for progress bar. Defaults to ‘’.
symbol (str) – Symbol for progress bar. Defaults to ‘#’.
round (int) – Number of digits to round to. Default is 5.
sep_label (str) – Separator character for value and label. Defaults to ‘: ‘.
sep_values (str) – Separator character for values. Defaults to ‘, ‘.
sep_note (str) – Separator character for note. Defaults to ‘, ‘.

update(value=None, weight=1, advance=1, auto_step=True, note=None)

Update progress bar and aggregate value.

Parameters:

value (Optional[Union[Number, List[Number]]]) – value. If None, only the progress bar advances. Defaults to None.
weight (Union[Number, List[Number]]) – weight of value. Defaults to 1.
advance (int) – Number of iterations to advance. Defaults to 1.
auto_step (bool) – If True, step() is called when the number of iterations reaches n_iter. Defaults to True.
note (Optional[str]) – Note for progress bar. Defaults to None.

step(bar_step=True, leave=True)

Step to the next epoch.

Parameters:: leave (bool) – If True, leave the progress bar. Defaults to True.

get_all_values(flatten=False)

Get values of all iterations.

Parameters:: flatten (bool) – If True, flatten to 1D sequence. else, 2D sequence of (epoch, value). Defaults to False.
Returns:: List[value] if flatten=True, else List[List[value]].
Return type:: Union[List[Numbers], List[List[Numbers]]]

get_all_times(flatten=False)

Get times of all iterations.

Parameters:

flatten (bool) – If True, flatten to 1D sequence. else, 2D sequence of (epoch, time). Defaults to False.

Returns:

List[time] if flatten=True, else: List[List[time]].

Return type:

List[Number]

memo(note=None, no_step=False)

Change note text for progress bar.

Parameters:

note (Optional[str]) – Text. If None, the note is not changed. Defaults to None.
no_step (bool) – If True, step() is not called when be deferred. Defaults to False.