File size: 9,351 Bytes
f9e8d8c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aef20d5
f9e8d8c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
53e45ca
 
 
f9e8d8c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
74a7ec7
f9e8d8c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
# 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 ast 
import re
import heapq


# 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 metrics aims to evaluate path planning abilities of LMs.
"""


# TODO: Add description of the arguments of the module here
_KWARGS_DESCRIPTION = """
Calculates how good the path predicted by the model is.
Args:
    predictions: list of predictions to score. Each predictions
        should be a string containing coordinates in the format (x, y).
    references: list of reference for each prediction. Each
        reference should be a string with tokens separated by spaces.
Returns:
    compliance_ratio: 1 if the model is able to produce a well formated python list,
    feasible_ratio: A path is feasible if it doesn'collide with obstacles, get out of the grid and doesn't move diagonally. This metric is the number of feasible paths over the total number of reachable paths.
    success_ratio: A path is successful if it is feasible and goes from the starting point to the end(s) point(s). This metric is the number of successful paths over the total number of reachable paths.
    optimal_ratio: A path is optimal if it is successful and its length is equal or shorter to the golden one. This metric is the number of optimal paths over the total number of reachable paths.
    distance: Sum of the distance from the last position predicted and the objectives for a feasible path not successful.
    unreachable_acc: Number of times the model was able to detect an unreachable path over the total of unreachable paths. -1 if all the goals are reachable.
Examples:

    >>> my_new_module = evaluate.load("rfr2003/path_planning_evaluate")
    >>> results = my_new_module.compute(generations=['[(0,0), (0,1), (1,1)]', '[(0,0), (1,0), (1,1)]', '[(0,0), (1,0), (1,1), (0,1)]', '(0,0'], golds=[[(0,0), (0,1), (1,1)], [(0,0), (0,1), (1,1)], [(0,0), (0,1)], []], obstacles=[[(1,0)], [(1,0)], [], []], ends=[[(1,1)], [(1,1)], [(0,1)], [(0,1)]], n=[2, 2, 2, 2])
    >>> print(results)
    {'compliance_ratio': 0.75, 'success_ratio': 0.6666666666666666,'optimal_ratio': 0.3333333333333333, 'feasible_ratio': 0.6666666666666666, 'distance': 0, 'unreachable_acc': 1.0}
"""


@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
class Path_Planning_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.Sequence(datasets.Value('int64'))),
                'obstacles': datasets.Sequence(datasets.Sequence(datasets.Value('int64'))),
                'ends': datasets.Sequence(datasets.Sequence(datasets.Value('int64'))),
                'n': datasets.Value('int64'),
            }),
        )

    def _download_and_prepare(self, dl_manager):
        """Optional: download external resources useful to compute the scores"""
        # TODO: Download external resources if needed
        pass

    def _a_star(self, n, obs, start, end):
        obs = set(obs)
        actions = [(-1, 0), (1, 0), (0, -1), (0, 1)]
    
        def heuristic(position):
            return abs(position[0] - end[0]) + abs(position[1] - end[1])
    
        visited = set()
        heap = []
        heapq.heappush(heap, (0, start, []))
    
        while heap:
            cost, current, path = heapq.heappop(heap)
            if current == end:
                return len(path)
    
            if current in visited:
                continue
    
            visited.add(current)
    
            for action in actions:
                neighbor = (current[0] + action[0], current[1] + action[1])
    
                if 0 <= neighbor[0] < n and 0 <= neighbor[1] < n:
                    if neighbor not in obs:
                        new_cost = cost + 1
                        new_path = path + [current]
                        heapq.heappush(heap, (new_cost + heuristic(neighbor), neighbor, new_path))
    
        return 100
    
    def _calculate_path_dist(self, n, obs, start, unvisited):
        dist = 0
    
        for end in unvisited:
            dist += self._a_star(n, obs, start, end)
    
        return dist
    
    def _evaluate_path(self, gen_path, gold_path, ends, obs, n):
        metrics = {'correct': False, 'optimal': False, 'detect_unreachable': False, 'feasible': False, 'is_unreachable': False, 'distance': 0}
        if len(gold_path) == 0:
            metrics['is_unreachable'] = True
            if len(gen_path) == 0:
                metrics['detect_unreachable'] = True
            return metrics
        else:
            if len(gen_path) == 0:
                return metrics
            
        start = gold_path[0]
    
        obst_set = set(obs)
    
        for i, (x, y) in enumerate(gen_path):
            if not (0 <= x < n and 0 <= y < n):
                return metrics #not feasible, correct or optimal
                
            if (x, y) in obst_set:
                return metrics #not feasible, correct or optimal
    
            if i > 0 and (abs(x-gen_path[i][0]) + abs(y-gen_path[i][1]) > 1):
                return metrics #not feasible, correct or optimal
    
        metrics['feasible'] = True
    
        unvisited = set()
        
        if gen_path[0] == start and gen_path[-1] in ends:
            metrics['correct'] = True
            for e in ends:
                if e not in gen_path:
                    metrics['correct'] = False
                    unvisited.add(e)
    
        if metrics['correct']:
            if len(gen_path) <= len(gold_path):
                metrics['optimal'] = True  
        else:
            metrics['distance'] = self._calculate_path_dist(n, obs, gen_path[-1], unvisited)
            
        return metrics

    def _compute(self, generations, golds, obstacles, ends, n):
        assert len(generations) == len(golds) == len(obstacles) == len(ends) == len(n)
        assert isinstance(golds, list)
    
        correct, total_reach, total_unreach = 0, 0, 0
    
        pattern = r"\(\s*(\d*)\s*,\s*(\d*)\s*\)"
    
        metrics = {'compliance': 0, 'correct': 0, 'optimal': 0, 'n_detect_unreachable': 0, 'feasible': 0, 'distance': 0}
    
        for gen, gold, _obs, _ends, _n in zip(generations, golds, obstacles, ends, n):
    
            try:
                gen_path = ast.literal_eval(gen)
                for i in range(len(gen_path)):
                    gen_path[i] = (int(gen_path[i][0]), int(gen_path[i][1])) 
                metrics['compliance'] += 1
            except:
                matches = re.findall(pattern, gen)
                gen_path = [(int(x), int(y)) for x, y in matches]
            
            gold_path = [(int(p[0]), int(p[1])) for p in gold]
            _obs = [(int(p[0]), int(p[1])) for p in _obs]
            _ends = [(int(p[0]), int(p[1])) for p in _ends]
    
            m = self._evaluate_path(gen_path, gold_path, _ends, _obs, _n)
    
            if m['is_unreachable']:
                total_unreach += 1
                metrics['n_detect_unreachable'] += m['detect_unreachable']
    
            else:
                for k in ['correct', 'optimal', 'feasible']:
                    metrics[k] += int(m[k])
        
                if m['feasible'] and not m['correct']:
                    metrics['distance'] += m['distance']
                                
                total_reach += 1
    
        diff = metrics['feasible']-metrics['correct']
    
        r_metrics = {}
        r_metrics.update({ 
            'compliance_ratio': metrics['compliance']/(total_reach+total_unreach), 
            'success_ratio': metrics['correct']/total_reach, 
            'optimal_ratio': metrics['optimal']/total_reach, 
            'feasible_ratio': metrics['feasible']/total_reach, 
            'distance': metrics['distance']/diff if diff > 0 else metrics['distance'],
            'unreachable_acc': metrics['n_detect_unreachable']/total_unreach if total_unreach > 0 else -1,
        })
    
        return r_metrics