classes

Some useful helper classes.

`swem.utils.classes.ClfMetricTracker`	Class for keeping track of the metrics corresponding to a single label in a classification task.
`swem.utils.classes.KeyDependentDefaultdict`	Defaultdict where the default factory can depend on the key.

ClfMetricTracker

class swem.utils.classes.ClfMetricTracker(name: str | int | None = None, tp: int = 0, fp: int = 0, fn: int = 0)

Class for keeping track of the metrics corresponding to a single label in a classification task.

Parameters

name (str | int | None) – The label this instance is keeping track of.
tp (int) – Start value for true positives. Defaults to 0.
fp (int) – Start-value for false positives. Defaults to 0.
fn (int) – Start-value for false negatives. Defaults to 0.

Return type

None

support: Number of instances for the label that were encountered.

recall: Current value for the recall metric.

precision: Current value for the precision metric.

f1_score: Current value for the f1_score metric.

Examples

>>> from swem.utils.classes import ClfMetricTracker
>>> tracker = ClfMetricTracker(name="Label_1", tp=5, fp=1, fn=3)
>>> tracker
ClfMetricTracker(name='Label_1', tp=5, fp=1, fn=3)
>>> tracker.support
8
>>> tracker.recall
0.625
>>> tracker.precision
0.8333333333333334
>>> tracker.f1_score
0.7142857142857143

KeyDependentDefaultdict

class swem.utils.classes.KeyDependentDefaultdict(*args, **kwargs)

Defaultdict where the default factory can depend on the key.

Parameters: default_factory (callable) – Default factory called when a missing key is encountered. The call will be default_factory(key) so the callable should take exactly one argument.

Examples

>>> d = KeyDependentDefaultdict(lambda key: {"name": key})
>>> d["a"]["b"] = 1
>>> d
{'a': {'name': 'a', 'b': 1}}