File size: 4,482 Bytes
cee9920
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d16fc41
cee9920
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d16fc41
cee9920
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
937d2a1
 
cee9920
937d2a1
cee9920
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""TODO: Add a description here."""

import evaluate
import datasets
import re
from statistics import median
import numpy as np
import ast

# TODO: Add BibTeX citation
_CITATION = """\
@InProceedings{huggingface:module,
title = {A great new module},
authors={huggingface, Inc.},
year={2020}
}
"""

# TODO: Add description of the module here
_DESCRIPTION = """\
This metric aims to evaluate regression tasks done by LMs.
"""


# TODO: Add description of the arguments of the module here
_KWARGS_DESCRIPTION = """
Calculates Accuracy and Blue-1 between generations and gold answers in a MCQ context.
Args:
    generations: list of predictions to score. Each predictions
        should be a string generated by a LM model.
    golds: list of reference for each prediction. Each
        reference should be a list of floats.
Returns:
    precision:  ,
    recall:
Examples:
    Here is an exemple on how to use the metric:

    >>> metric = evaluate.load("rfr2003/regression_evaluate")
    >>> results = metric.compute(generations=['[150, 0]'], golds=[183, 177, 146, 85, 70, 78, 55, 17, 0, -1, -1])
    >>> print(results)
    {'precision': 4.0, 'recall': 344.0, 'macro-mean': 174.0, 'median macro-mean': 174.0}
"""


@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
class regression_evaluate(evaluate.Metric):
    """TODO: Short description of my evaluation module."""

    def _info(self):
        # TODO: Specifies the evaluate.EvaluationModuleInfo object
        return evaluate.MetricInfo(
            # This is the description that will appear on the modules page.
            module_type="metric",
            description=_DESCRIPTION,
            citation=_CITATION,
            inputs_description=_KWARGS_DESCRIPTION,
            # This defines the format of each prediction and reference
            features=datasets.Features({
                'generations': datasets.Value('string'),
                'golds': datasets.Sequence(datasets.Value('float32')),
            }),
        )

    def _download_and_prepare(self, dl_manager):
        """Optional: download external resources useful to compute the scores"""
        pass

    def _calculate_pre_rec(self, gens, golds, d):
        dists = []
    
        for gold in golds:
            g_dist = []
            for gen in gens:
                g_dist.append(d(gold, gen))
            if len(g_dist) == 0:
                g_dist.append(100) #penalty if the model doesnt generate anything
            dists.append(g_dist)
    
        dists = np.array(dists)
        precision = np.min(dists, axis=0).sum()
    
        recall = np.min(dists, axis=1).sum()
    
        return precision, recall

    def _compute(self, generations, golds):
        assert len(generations) == len(golds)
        assert isinstance(golds, list)
    
        precisions, recalls, means_pre_rec = [], [], []
    
        for gen, gold in zip(generations, golds):
            f_gold = list(set([float(g) for g in gold]))
            
            f_ans = re.findall(r'\d+(?:,\d+)*(?:\.\d+)?|\d+(?:\.\d+)?', gen)
                
            f_ans = list(set([float(a.replace(',', '')) for a in f_ans])) #get rid of duples values
    
            precision, recall = self._calculate_pre_rec(f_ans, f_gold, lambda x,y: abs(x-y))
    
            precisions.append(precision)
            recalls.append(recall)
            means_pre_rec.append((precision+recall)/2)
            
    
        macro_prec = np.mean(precisions).item()
        macro_rec = np.mean(recalls).item()
    
        metrics = {}
        metrics.update({ 
            'precision': np.mean(precisions).item(),
            'recall': np.mean(recalls).item(),
            'macro-mean': np.mean(means_pre_rec).item(),
            'median macro-mean': median(means_pre_rec)
        })
    
        return metrics