Coverage for addmo/s3_model_tuning/scoring/abstract_scorer.py: 84%

1from abc import ABC, abstractmethod 

2 

3from addmo.s3_model_tuning.config.model_tuning_config import ModelTunerConfig 

4from addmo.s3_model_tuning.models.abstract_model import AbstractMLModel 

5from addmo.s3_model_tuning.scoring.metrics.abstract_metric import AbstractMetric 

6from addmo.s3_model_tuning.scoring.metrics.metric_factory import MetricFactory 

7from addmo.s3_model_tuning.scoring.validation_splitting.splitter_factory import ( 

8 SplitterFactory, 

9) 

10 

11 

12class Scoring: 

13 """This class is used to score the model on the test period. 

14 Currently I dont see any customizations that couldnt be implemented directly into the custom 

15 metric class. Hence, this class is not designed as abstract class.""" 

16 

17 def __init__(self, scoring_metric: str): 

18 self.metric: AbstractMetric = MetricFactory.metric_factory(scoring_metric) 

19 

20 def score_test(self, model: AbstractMLModel, x, y): 

21 """Returns a positive float value. The higher the better. 

22 x and y include only test period. The model is already trained.""" 

23 return self.metric(model, x, y) 

24 

25 

26class ValidationScoring(ABC): 

27 def __init__(self, config: ModelTunerConfig): 

28 self.metric: AbstractMetric = MetricFactory.metric_factory( 

29 config.validation_score_metric, config.validation_score_metric_kwargs 

30 ) 

31 self.splitter = SplitterFactory.splitter_factory(config) 

32 

33 @staticmethod 

34 @abstractmethod 

35 def score_validation(model: AbstractMLModel, x, y): 

36 """Returns a positive float value. The higher the better. 

37 x and y include train and evaluation period. The model will be trained on the corresponding 

38 train period.""" 

39 pass