-
Notifications
You must be signed in to change notification settings - Fork 116
/
test_GymConverter.py
354 lines (296 loc) · 15.5 KB
/
test_GymConverter.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
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
# Copyright (c) 2019-2020, RTE (https://www.rte-france.com)
# See AUTHORS.txt
# This Source Code Form is subject to the terms of the Mozilla Public License, version 2.0.
# If a copy of the Mozilla Public License, version 2.0 was not distributed with this file,
# you can obtain one at http://mozilla.org/MPL/2.0/.
# SPDX-License-Identifier: MPL-2.0
# This file is part of Grid2Op, Grid2Op a testbed platform to model sequential decision making in power systems.
# TODO test the json part but... https://github.com/openai/gym-http-api/issues/62 or https://github.com/openai/gym/issues/1841
import tempfile
import json
from grid2op.dtypes import dt_float, dt_bool, dt_int
from grid2op.tests.helper_path_test import *
from grid2op.MakeEnv import make
from grid2op.Converter import IdToAct, ToVect
from grid2op.gym_compat import GymActionSpace, GymObservationSpace
from grid2op.gym_compat import GymEnv
from grid2op.gym_compat import ContinuousToDiscreteConverter
import pdb
import warnings
warnings.simplefilter("error")
class BaseTestGymConverter:
def __init__(self):
self.tol = 1e-6
def _aux_test_json(self, space, obj=None):
if obj is None:
obj = space.sample()
obj_json = space.to_jsonable([obj])
# test save to json
with tempfile.TemporaryFile(mode="w") as f:
json.dump(obj_json, fp=f)
# test read from json
obj2 = space.from_jsonable(obj_json)[0]
# test they are equal
for k, v in obj2.items():
assert k in obj
tmp = obj[k]
if isinstance(tmp, (int, float, dt_float, dt_int, dt_bool)):
assert np.all(np.abs(float(obj[k]) - float(obj2[k])) <= self.tol)
elif len(tmp) == 1:
assert np.all(np.abs(float(obj[k]) - float(obj2[k])) <= self.tol)
else:
assert np.all(np.abs(obj[k].astype(dt_float) - obj2[k].astype(dt_float)) <= self.tol)
for k, v in obj.items():
assert k in obj2 # make sure every keys of obj are in obj2
class TestWithoutConverterWCCI(unittest.TestCase, BaseTestGymConverter):
def setUp(self) -> None:
BaseTestGymConverter.__init__(self)
def get_env_name(self):
return "l2rpn_wcci_2020"
def test_creation(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make(self.get_env_name(), test=True) as env:
# test i can create
obs_space = GymObservationSpace(env)
act_space = GymActionSpace(env)
def test_json(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make(self.get_env_name(), test=True) as env:
# test i can create
obs_space = GymObservationSpace(env)
act_space = GymActionSpace(env)
obs_space.seed(0)
act_space.seed(0)
self._aux_test_json(obs_space)
self._aux_test_json(act_space)
def test_to_from_gym_obs(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make(self.get_env_name(), test=True) as env:
obs_space = GymObservationSpace(env)
obs = env.reset()
gym_obs = obs_space.to_gym(obs)
self._aux_test_json(obs_space, gym_obs)
assert obs_space.contains(gym_obs)
obs2 = obs_space.from_gym(gym_obs)
# TODO there is not reason that these 2 are equal: reset, will erase everything
# TODO whereas creating the observation
# assert obs == obs2
obs_diff, attr_diff = obs.where_different(obs2)
for el in attr_diff:
assert el in obs.attr_list_json, f"{el} should be equal in obs and obs2"
for i in range(10):
obs, *_ = env.step(env.action_space())
gym_obs = obs_space.to_gym(obs)
self._aux_test_json(obs_space, gym_obs)
assert obs_space.contains(gym_obs), "gym space does not contain the observation for ts {}".format(i)
obs2 = obs_space.from_gym(gym_obs)
# TODO there is not reason that these 2 are equal: reset, will erase everything
# TODO whereas creating the observation
# assert obs == obs2, "obs and converted obs are not equal for ts {}".format(i)
obs_diff, attr_diff = obs.where_different(obs2)
for el in attr_diff:
assert el in obs.attr_list_json, f"{el} should be equal in obs and obs2 for ts {i}"
def test_to_from_gym_act(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make(self.get_env_name(), test=True) as env:
act_space = GymActionSpace(env)
act = env.action_space()
gym_act = act_space.to_gym(act)
self._aux_test_json(act_space, gym_act)
assert act_space.contains(gym_act)
act2 = act_space.from_gym(gym_act)
assert act == act2
act_space.seed(0)
for i in range(10):
gym_act = act_space.sample()
act = act_space.from_gym(gym_act)
self._aux_test_json(act_space, gym_act)
gym_act2 = act_space.to_gym(act)
act2 = act_space.from_gym(gym_act2)
assert act == act2
class BaseTestConverter(BaseTestGymConverter):
def init_converter(self, env):
raise NotImplementedError()
def test_creation(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
# test i can create
converter = self.init_converter(env)
act_space = GymActionSpace(env=env, converter=converter)
act_space.sample()
def test_json(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
# test i can create
converter = self.init_converter(env)
act_space = GymActionSpace(env=env, converter=converter)
act_space.seed(0)
self._aux_test_json(act_space)
def test_to_from_gym_act(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
converter = self.init_converter(env)
act_space = GymActionSpace(env=env, converter=converter)
act_space.seed(0)
converter.seed(0)
gym_act = act_space.sample()
act = act_space.from_gym(gym_act)
self._aux_test_json(act_space, gym_act)
gym_act2 = act_space.to_gym(act)
act2 = act_space.from_gym(gym_act2)
g2op_act = converter.convert_act(act)
g2op_act2 = converter.convert_act(act2)
assert g2op_act == g2op_act2
act_space.seed(0)
for i in range(10):
gym_act = act_space.sample()
act = act_space.from_gym(gym_act)
self._aux_test_json(act_space, gym_act)
gym_act2 = act_space.to_gym(act)
act2 = act_space.from_gym(gym_act2)
g2op_act = converter.convert_act(act)
g2op_act2 = converter.convert_act(act2)
assert g2op_act == g2op_act2
class TestIdToAct(unittest.TestCase, BaseTestConverter):
def init_converter(self, env):
return IdToAct(env.action_space)
def setUp(self) -> None:
BaseTestGymConverter.__init__(self)
class TestToVect(unittest.TestCase, BaseTestConverter):
def init_converter(self, env):
return ToVect(env.action_space)
def setUp(self) -> None:
BaseTestGymConverter.__init__(self)
class TestDropAttr(unittest.TestCase):
"""test the method to remove part of the attribute of the action / observation space"""
def test_keep_only_attr(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
env = make("educ_case14_redisp", test=True)
gym_env = GymEnv(env)
attr_kept = sorted(("rho", "line_status", "actual_dispatch", "target_dispatch"))
ob_space = gym_env.observation_space
ob_space = ob_space.keep_only_attr(attr_kept)
assert np.all(sorted(ob_space.spaces.keys()) == attr_kept)
def test_ignore_attr(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
env = make("educ_case14_redisp", test=True)
gym_env = GymEnv(env)
attr_deleted = sorted(("rho", "line_status", "actual_dispatch", "target_dispatch"))
ob_space = gym_env.observation_space
ob_space = ob_space.ignore_attr(attr_deleted)
for el in attr_deleted:
assert not el in ob_space.spaces
class TestContinuousToDiscrete(unittest.TestCase):
"""test the ContinuousToDiscreteConverter converter"""
def setUp(self) -> None:
self.tol = 1e-4
def test_split_in_3(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
env = make("educ_case14_redisp", test=True)
gym_env = GymEnv(env)
act_space = gym_env.action_space
act_space = act_space.reencode_space("redispatch",
ContinuousToDiscreteConverter(nb_bins=3,
init_space=act_space["redispatch"])
)
# with 3 interval like [-10, 10] (the second generator)
# should be split => 0 -> [-10, -3.33), 1 => [-3.33, 3.33), [3.33, 10.]
# test the "all 0" action (all 0 => encoded to 1, because i have 3 bins)
g2op_object = np.array([0., 0., 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [1, 1, 0, 0, 0, 1])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(res2 == 0.)
# test the all 0 action, but one is not 0 (negative)
g2op_object = np.array([0., -3.2, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [1, 1, 0, 0, 0, 1])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(res2 == 0.)
# test the all 0 action, but one is not 0 (positive)
g2op_object = np.array([0., 3.2, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [1, 1, 0, 0, 0, 1])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(res2 == 0.)
# test one is 2
g2op_object = np.array([0., 3.4, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [1, 2, 0, 0, 0, 1])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., 5.0, 0., 0., 0., 0.]) <= self.tol)
# test one is 0
g2op_object = np.array([0., -3.4, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [1, 0, 0, 0, 0, 1])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., -5.0, 0., 0., 0., 0.]) <= self.tol)
def test_split_in_5(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
env = make("educ_case14_redisp", test=True)
gym_env = GymEnv(env)
act_space = gym_env.action_space
act_space = act_space.reencode_space("redispatch",
ContinuousToDiscreteConverter(nb_bins=5,
init_space=act_space["redispatch"])
)
# with 5
g2op_object = np.array([0., 0., 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 2, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(res2 == 0.)
# test the all 0 action, but one is not 0 (negative)
g2op_object = np.array([0., -1.9, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 2, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(res2 == 0.)
# positive side
g2op_object = np.array([0., 2.1, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 3, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., 3.33333, 0., 0., 0., 0.]) <= self.tol)
g2op_object = np.array([0., 5.9, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 3, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., 3.33333, 0., 0., 0., 0.]) <= self.tol)
g2op_object = np.array([0., 6.1, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 4, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., 6.666666, 0., 0., 0., 0.]) <= self.tol)
# negative side
g2op_object = np.array([0., -2.1, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 1, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., -3.3333, 0., 0., 0., 0.]) <= self.tol)
g2op_object = np.array([0., -5.9, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 1, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., -3.33333, 0., 0., 0., 0.]) <= self.tol)
g2op_object = np.array([0., -6.1, 0., 0., 0., 0.])
res = act_space._keys_encoding["_redispatch"].g2op_to_gym(g2op_object)
assert np.all(res == [2, 0, 0, 0, 0, 2])
res2 = act_space._keys_encoding["_redispatch"].gym_to_g2op(res)
assert np.all(np.abs(res2 - [0., -6.666666, 0., 0., 0., 0.]) <= self.tol)
class TestWithoutConverterStorage(TestWithoutConverterWCCI):
def get_env_name(self):
return "educ_case14_storage"
if __name__ == "__main__":
unittest.main()