-
Notifications
You must be signed in to change notification settings - Fork 67
/
Copy pathtest_ablation.py
255 lines (210 loc) · 10.5 KB
/
test_ablation.py
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
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
# License: BSD 3 clause
"""
Tests related to ablation experiments.
:author: Michael Heilman (mheilman@ets.org)
:author: Nitin Madnani (nmadnani@ets.org)
:author: Dan Blanchard (dblanchard@ets.org)
:author: Aoife Cahill (acahill@ets.org)
"""
import csv
import json
import unittest
from skll.experiments import run_configuration
from skll.utils.testing import (
config_dir,
create_jsonlines_feature_files,
fill_in_config_paths,
output_dir,
remove_jsonlines_feature_files,
test_dir,
train_dir,
unlink,
)
class TestAblation(unittest.TestCase):
"""Test class for ablation tests."""
@classmethod
def setUpClass(cls):
"""Create necessary directories for testing."""
for dir_path in [train_dir, test_dir, output_dir]:
dir_path.mkdir(exist_ok=True)
# create jsonlines feature files
create_jsonlines_feature_files(train_dir)
@classmethod
def tearDownClass(cls):
"""Clean up after tests."""
for output_file in output_dir.glob("ablation_cv_*"):
unlink(output_file)
config_files = [
"test_ablation.cfg",
"test_ablation_all_combos.cfg",
"test_ablation_feature_hasher.cfg",
"test_ablation_feature_hasher_all_combos.cfg",
"test_ablation_sampler.cfg",
"test_ablation_sampler_all_combos.cfg",
"test_ablation_feature_hasher_sampler.cfg",
"test_ablation_feature_hasher_sampler_all_combos.cfg",
]
for cf in config_files:
unlink(config_dir / cf)
remove_jsonlines_feature_files(train_dir)
def check_ablation_rows(self, reader):
"""
Ensure that all ablated faetures and featureset values are correct.
Parameters
----------
reader : object
Reader object that iterates over the output summary file.
Returns
-------
int
The total number of rows in the summary file.
"""
row_num = 0
for row_num, row in enumerate(reader, 1):
if row["ablated_features"]:
fs_str, ablated_str = row["featureset_name"].split("_minus_")
actual_ablated = json.loads(row["ablated_features"])
else:
fs_str, ablated_str = row["featureset_name"].split("_all")
actual_ablated = []
expected_fs = set(fs_str.split("+"))
expected_ablated = ablated_str.split("+") if ablated_str else []
expected_fs = sorted(expected_fs - set(expected_ablated))
actual_fs = json.loads(row["featureset"])
self.assertEqual(expected_ablated, actual_ablated)
self.assertEqual(expected_fs, actual_fs)
return row_num
def test_ablation_cv(self):
"""Test ablation + cross-validation."""
config_template_path = config_dir / "test_ablation.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=1, local=True)
# read in the summary file and make sure it has
# 7 ablated featuresets * (10 folds + 1 average line) * 2 learners = 154
# lines
with open(output_dir / "ablation_cv_plain_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 154)
# make sure there are 7 ablated featuresets * 2 learners = 12 results files
num_result_files = len(list(output_dir.glob("ablation_cv_plain*.results")))
self.assertEqual(num_result_files, 14)
def test_ablation_cv_all_combos(self):
"""Test ablation all-combos + cross-validation."""
config_template_path = config_dir / "test_ablation_all_combos.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=None, local=True)
# read in the summary file and make sure it has
# 10 ablated featuresets * (10 folds + 1 average line) * 2 learners = 220
# lines
with open(output_dir / "ablation_cv_plain_all_combos_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 220)
# make sure there are 10 ablated featuresets * 2 learners = 20 results
# files
num_result_files = len(list(output_dir.glob("ablation_cv_plain_all_combos*results")))
self.assertEqual(num_result_files, 20)
def test_ablation_cv_feature_hasher(self):
"""Test ablation + cross-validation + feature hashing."""
config_template_path = config_dir / "test_ablation_feature_hasher.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=1, local=True)
# read in the summary file and make sure it has
# 7 ablated featuresets * (10 folds + 1 average line) * 2 learners = 154
# lines
with open(output_dir / "ablation_cv_feature_hasher_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 154)
# make sure there are 7 ablated featuresets * 2 learners = 14 results files
num_result_files = len(list(output_dir.glob("ablation_cv_feature_hasher_*.results")))
self.assertEqual(num_result_files, 14)
def test_ablation_cv_feature_hasher_all_combos(self):
"""Test ablation all-combos + cross-validation + feature hashing."""
config_template_path = config_dir / "test_ablation_feature_hasher_all_combos.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=None, local=True)
# read in the summary file and make sure it has
# 10 ablated featuresets
# * (10 folds + 1 average line)
# * 2 learners
# = 220 lines in total
with open(output_dir / "ablation_cv_feature_hasher_all_combos_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 220)
# make sure there are 10 ablated featuresets * 2 learners = 20 results
# files
num_result_files = len(
list(output_dir.glob("ablation_cv_feature_hasher_all_combos*.results"))
)
self.assertEqual(num_result_files, 20)
def test_ablation_cv_sampler(self):
"""Test ablation + cross-validation + samplers."""
config_template_path = config_dir / "test_ablation_sampler.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=1, local=True)
# read in the summary file and make sure it has
# 7 ablated featuresets * (10 folds + 1 average line) * 2 learners = 154
# lines
with open(output_dir / "ablation_cv_sampler_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 154)
# make sure there are 6 ablated featuresets * 2 learners = 12 results files
num_result_files = len(list(output_dir.glob("ablation_cv_sampler_f*.results")))
self.assertEqual(num_result_files, 14)
def test_ablation_cv_all_combos_sampler(self):
"""Test ablation all-combos + cross-validation + samplers."""
config_template_path = config_dir / "test_ablation_sampler_all_combos.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=None, local=True)
# read in the summary file and make sure it has
# 10 ablated featuresets * (10 folds + 1 average line) * 2 learners = 220
# lines
with open(output_dir / "ablation_cv_sampler_all_combos_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 220)
# make sure there are 10 ablated featuresets * 2 learners = 20 results
# files
num_result_files = len(list(output_dir.glob("ablation_cv_sampler_all_combos*.results")))
self.assertEqual(num_result_files, 20)
def test_ablation_cv_feature_hasher_sampler(self):
"""Test ablation + cross-validation + feature hashing + samplers."""
config_template_path = config_dir / "test_ablation_feature_hasher_sampler.template.cfg"
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=1, local=True)
# read in the summary file and make sure it has
# 7 ablated featuresets * (10 folds + 1 average line) * 2 learners = 154
# lines
with open(output_dir / "ablation_cv_feature_hasher_sampler_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 154)
# make sure there are 7 ablated featuresets * 2 learners = 14 results files
num_result_files = len(
list(output_dir.glob("ablation_cv_feature_hasher_sampler_f*.results"))
)
self.assertEqual(num_result_files, 14)
def test_ablation_cv_feature_hasher_all_combos_sampler(self):
"""Test ablation all-combos + cross-validation + feature hashing + samplers."""
config_template_path = (
config_dir / "test_ablation_feature_hasher_sampler_all_combos.template.cfg"
)
config_path = fill_in_config_paths(config_template_path)
run_configuration(config_path, quiet=True, ablation=None, local=True)
# read in the summary file and make sure it has
# 10 ablated featuresets * (10 folds + 1 average line) * 2 learners = 220
# lines
with open(output_dir / "ablation_cv_feature_hasher_sampler_all_combos_summary.tsv") as f:
reader = csv.DictReader(f, dialect=csv.excel_tab)
num_rows = self.check_ablation_rows(reader)
self.assertEqual(num_rows, 220)
# make sure there are 10 ablated featuresets * 2 learners = 20 results
# files
num_result_files = len(
list(output_dir.glob("ablation_cv_feature_hasher_sampler_all_combos*.results"))
)
self.assertEqual(num_result_files, 20)