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
|
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# 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.
# ==============================================================================
"""Tests for class Evaluator."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tempfile
from tensorflow.contrib.eager.python import evaluator
from tensorflow.contrib.eager.python import metrics
from tensorflow.contrib.summary import summary_test_util
from tensorflow.python.data.ops import dataset_ops
from tensorflow.python.eager import context
from tensorflow.python.eager import test
from tensorflow.python.framework import ops
from tensorflow.python.ops import variables
from tensorflow.python.training import training_util
class IdentityModel(object):
def eval_data(self, d):
return d
class PrefixLModel(object):
def eval_data(self, d):
return {"l_" + key: d[key] for key in d}
class SimpleEvaluator(evaluator.Evaluator):
def __init__(self, model):
super(SimpleEvaluator, self).__init__(model)
self.mean = self.track_metric(metrics.Mean("mean"))
def call(self, eval_data):
self.mean(eval_data)
class DelegatingEvaluator(evaluator.Evaluator):
def __init__(self, model):
super(DelegatingEvaluator, self).__init__(model)
self.sub = self.track_evaluator("inner", SimpleEvaluator(model))
self.mean = self.track_metric(metrics.Mean("outer-mean"))
def call(self, eval_data):
# Keys here come from PrefixLModel, which adds "l_".
self.mean(eval_data["l_outer"])
self.sub.call(eval_data["l_inner"])
# pylint: disable=not-callable
class EvaluatorTest(test.TestCase):
def testSimple(self):
e = SimpleEvaluator(IdentityModel())
e(3.0)
e([5.0, 7.0, 9.0])
results = e.all_metric_results()
self.assertEqual(set(["mean"]), set(results.keys()))
self.assertEqual(6.0, results["mean"].numpy())
def testWriteSummaries(self):
e = SimpleEvaluator(IdentityModel())
e(3.0)
e([5.0, 7.0, 9.0])
training_util.get_or_create_global_step()
logdir = tempfile.mkdtemp()
e.all_metric_results(logdir)
events = summary_test_util.events_from_logdir(logdir)
self.assertEqual(len(events), 2)
self.assertEqual(events[1].summary.value[0].simple_value, 6.0)
def testComposition(self):
e = DelegatingEvaluator(PrefixLModel())
e({"inner": 2.0, "outer": 100.0})
e({"inner": 4.0, "outer": 1000.0})
results = e.all_metric_results()
self.assertEqual(set(["inner/mean", "outer-mean"]), set(results.keys()))
self.assertEqual(3.0, results["inner/mean"].numpy())
self.assertEqual(550.0, results["outer-mean"].numpy())
def testMetricVariables(self):
e = DelegatingEvaluator(PrefixLModel())
e({"inner": 2.0, "outer": 100.0})
prefix_count = {}
for v in e.metric_variables:
p = v.name.split("/")[0]
prefix_count[p] = prefix_count.get(p, 0) + 1
self.assertEqual({"outer_mean": 2, "mean": 2}, prefix_count)
def testDatasetEager(self):
e = SimpleEvaluator(IdentityModel())
ds = dataset_ops.Dataset.from_tensor_slices([3.0, 5.0, 7.0, 9.0])
results = e.evaluate_on_dataset(ds)
self.assertEqual(set(["mean"]), set(results.keys()))
self.assertEqual(6.0, results["mean"].numpy())
def testDatasetGraph(self):
with context.graph_mode(), ops.Graph().as_default(), self.test_session():
e = SimpleEvaluator(IdentityModel())
ds = dataset_ops.Dataset.from_tensor_slices([3.0, 5.0, 7.0, 9.0])
init_op, call_op, results_op = e.evaluate_on_dataset(ds)
results = e.run_evaluation(init_op, call_op, results_op)
self.assertEqual(set(["mean"]), set(results.keys()))
self.assertEqual(6.0, results["mean"])
def testWriteSummariesGraph(self):
with context.graph_mode(), ops.Graph().as_default(), self.test_session():
e = SimpleEvaluator(IdentityModel())
ds = dataset_ops.Dataset.from_tensor_slices([3.0, 5.0, 7.0, 9.0])
training_util.get_or_create_global_step()
logdir = tempfile.mkdtemp()
init_op, call_op, results_op = e.evaluate_on_dataset(
ds, summary_logdir=logdir)
variables.global_variables_initializer().run()
e.run_evaluation(init_op, call_op, results_op)
events = summary_test_util.events_from_logdir(logdir)
self.assertEqual(len(events), 2)
self.assertEqual(events[1].summary.value[0].simple_value, 6.0)
def testModelProperty(self):
m = IdentityModel()
e = SimpleEvaluator(m)
self.assertIs(m, e.model)
def testMetricsProperty(self):
e = DelegatingEvaluator(PrefixLModel())
names = set([(p, m.name) for p, m in e.metrics])
self.assertEqual(set([("", "outer-mean"), ("inner/", "mean")]), names)
def testSharedMetric(self):
class MetricArgEvaluator(evaluator.Evaluator):
def __init__(self, model, m):
super(MetricArgEvaluator, self).__init__(model)
self.m = self.track_metric(m)
metric = metrics.Mean("mean")
model = IdentityModel()
e = MetricArgEvaluator(model, metric)
with self.assertRaisesRegexp(ValueError, "already added"):
MetricArgEvaluator(model, metric)
del e
def testMetricTrackedTwice(self):
class MetricTwiceEvaluator(evaluator.Evaluator):
def __init__(self, model):
super(MetricTwiceEvaluator, self).__init__(model)
self.m = self.track_metric(metrics.Mean("mean"))
self.track_metric(self.m) # okay to track same metric again
MetricTwiceEvaluator(IdentityModel())
class SparseSoftmaxEvaluatorTest(test.TestCase):
def testSimple(self):
e = evaluator.SparseSoftmaxEvaluator(IdentityModel())
e({e.loss_key: 1.0, e.label_key: 5, e.predicted_class_key: 5})
e({e.loss_key: [0.0, 3.0, 4.0],
e.label_key: [1, 2, 3],
e.predicted_class_key: [1, 1, 3]})
results = e.all_metric_results()
self.assertEqual(set(["Avg Loss", "Accuracy"]), set(results.keys()))
self.assertEqual(2.0, results["Avg Loss"].numpy())
self.assertEqual(0.75, results["Accuracy"].numpy())
if __name__ == "__main__":
test.main()
|
