Helper Module for Deep Learning.

Module that provides logging utilities.

class pynet.history.History(name, verbose=0)[source]¶

Track training progress by following the some metrics.

__init__(name, verbose=0)[source]¶

Initilize the class.

Parameters

name: str

the object name.

verbose: int, default 0

control the verbosity level.

get_total_time()[source]¶

Returns the total period between the first and last steps.

classmethod load(file_name)[source]¶

log(step, **kwargs)[source]¶

Record some metrics at a specific step.

Example:

state = History() state.log(1, loss=1., accuracy=0.)

If logging the same metrics for one specific step, new values overwrite older ones.

Parameters

step: int or uplet

The step name: we can use a tuple to log the fold, the epoch or the step within the epoch.

kwargs

The metrics to be logged.

save(outdir, fold, epoch)[source]¶

property steps¶

Returns a list of all steps.

summary()[source]¶