TobiTob commited on
Commit
a35e4bd
1 Parent(s): 743e3bb

Delete generate_sequences.py

Browse files
Files changed (1) hide show
  1. generate_sequences.py +0 -201
generate_sequences.py DELETED
@@ -1,201 +0,0 @@
1
- import os
2
- import numpy as np
3
- import pickle
4
- import time
5
- from agents.orderenforcingwrapper import OrderEnforcingAgent
6
- from citylearn.citylearn import CityLearnEnv
7
-
8
- """
9
- This file is used to generate offline data for a decision transformer.
10
- Data is saved as pickle file.
11
- Data structure:
12
- list(
13
- dict(
14
- "observations": nparray(nparray(np.float32)),
15
- "next_observations": nparray(nparray(np.float32)),
16
- "actions": nparray(nparray(np.float32)),
17
- "rewards": nparray(np.oat32),
18
- "terminals": nparray(np.bool_)
19
- )
20
- )
21
- """
22
-
23
-
24
- class Constants:
25
- file_to_save = "non.pkl"
26
- sequence_length = 720
27
- episodes = 1
28
- state_dim = 28 # size of state space
29
- action_dim = 1 # size of action space
30
- schema_path = './data/citylearn_challenge_2022_phase_1/schema.json'
31
-
32
-
33
- def action_space_to_dict(aspace):
34
- """ Only for box space """
35
- return {"high": aspace.high,
36
- "low": aspace.low,
37
- "shape": aspace.shape,
38
- "dtype": str(aspace.dtype)
39
- }
40
-
41
-
42
- def env_reset(env):
43
- observations = env.reset()
44
- action_space = env.action_space
45
- observation_space = env.observation_space
46
- building_info = env.get_building_information()
47
- building_info = list(building_info.values())
48
- action_space_dicts = [action_space_to_dict(asp) for asp in action_space]
49
- observation_space_dicts = [action_space_to_dict(osp) for osp in observation_space]
50
- obs_dict = {"action_space": action_space_dicts,
51
- "observation_space": observation_space_dicts,
52
- "building_info": building_info,
53
- "observation": observations}
54
- return obs_dict
55
-
56
-
57
- def generate_data():
58
- print("========================= Start Data Collection ========================")
59
-
60
- env = CityLearnEnv(schema=Constants.schema_path)
61
- agent = OrderEnforcingAgent()
62
-
63
- dataset = []
64
- observation_data = []
65
- next_observation_data = []
66
- action_data = []
67
- reward_data = []
68
- done_data = []
69
-
70
- obs_dict = env_reset(env)
71
- observations = obs_dict["observation"]
72
-
73
- agent_time_elapsed = 0
74
-
75
- step_start = time.perf_counter()
76
- actions = agent.register_reset(obs_dict)
77
- agent_time_elapsed += time.perf_counter() - step_start
78
-
79
- episodes_completed = 0
80
- sequences_completed = 0
81
- current_step_total = 0
82
- current_step_in_sequence = 0
83
- interrupted = False
84
- episode_metrics = []
85
-
86
- try:
87
- while True:
88
- current_step_in_sequence += 1
89
- current_step_total += 1
90
- next_observations, reward, done, info = env.step(actions)
91
- # ACTION [-1,1] attempts to decrease or increase the electricity stored in the battery by an amount
92
- # equivalent to action times its maximum capacity
93
- # Save environment interactions:
94
- observation_data.append(observations)
95
- next_observation_data.append(next_observations)
96
- action_data.append(actions)
97
- reward_data.append(reward)
98
- done_data.append(False) # always False
99
-
100
- observations = next_observations # observations of next time step
101
-
102
- if current_step_in_sequence >= Constants.sequence_length: # Sequence completed
103
- current_step_in_sequence = 0
104
- sequences_completed += 1
105
-
106
- for bi in range(len(env.buildings)):
107
- obs_building_i = np.zeros((Constants.sequence_length, Constants.state_dim), dtype=np.float32)
108
- n_obs_building_i = np.zeros((Constants.sequence_length, Constants.state_dim), dtype=np.float32)
109
- acts_building_i = np.zeros((Constants.sequence_length, Constants.action_dim), dtype=np.float32)
110
- rwds_building_i = np.zeros(Constants.sequence_length, dtype=np.float32)
111
- dones_building_i = np.zeros(Constants.sequence_length, dtype=np.bool_)
112
- for ti in range(Constants.sequence_length):
113
- obs_building_i[ti] = np.array(observation_data[ti][bi])
114
- n_obs_building_i[ti] = np.array(next_observation_data[ti][bi])
115
- acts_building_i[ti] = np.array(action_data[ti][bi])
116
- rwds_building_i[ti] = reward_data[ti][bi]
117
- dones_building_i[ti] = done_data[ti]
118
-
119
- dict_building_i = {
120
- "observations": obs_building_i,
121
- "next_observations": n_obs_building_i,
122
- "actions": acts_building_i,
123
- "rewards": rwds_building_i,
124
- "terminals": dones_building_i
125
- }
126
- dataset.append(dict_building_i)
127
-
128
- observation_data = []
129
- next_observation_data = []
130
- action_data = []
131
- reward_data = []
132
- done_data = []
133
- print("Sequence completed:", sequences_completed)
134
-
135
- if done:
136
- episodes_completed += 1
137
-
138
- metrics_t = env.evaluate()
139
- metrics = {"price_cost": metrics_t[0], "emmision_cost": metrics_t[1], "grid_cost": metrics_t[2]}
140
- if np.any(np.isnan(metrics_t)):
141
- raise ValueError("Episode metrics are nan, please contant organizers")
142
- episode_metrics.append(metrics)
143
- print(f"Episode complete: {episodes_completed} | Latest episode metrics: {metrics}", )
144
-
145
- obs_dict = env_reset(env)
146
- observations = obs_dict["observation"]
147
-
148
- step_start = time.perf_counter()
149
- actions = agent.register_reset(obs_dict)
150
- agent_time_elapsed += time.perf_counter() - step_start
151
- else:
152
- step_start = time.perf_counter()
153
- actions = agent.compute_action(next_observations)
154
- agent_time_elapsed += time.perf_counter() - step_start
155
-
156
- if current_step_total % 1000 == 0:
157
- print(f"Num Steps: {current_step_total}, Num episodes: {episodes_completed}")
158
-
159
- if episodes_completed >= Constants.episodes:
160
- break
161
- except KeyboardInterrupt:
162
- print("========================= Stopping Generation ==========================")
163
- interrupted = True
164
-
165
- if not interrupted:
166
- print("========================= Generation Completed =========================")
167
-
168
- if len(episode_metrics) > 0:
169
- print("Agent Performance:")
170
- print("Average Price Cost:", np.mean([e['price_cost'] for e in episode_metrics]))
171
- print("Average Emission Cost:", np.mean([e['emmision_cost'] for e in episode_metrics]))
172
- print("Average Grid Cost:", np.mean([e['grid_cost'] for e in episode_metrics]))
173
- print(f"Total time taken by agent: {agent_time_elapsed}s")
174
-
175
- print("========================= Writing Data File ============================")
176
-
177
- length = 0
178
- for data in dataset:
179
- if len(data["observations"]) > length:
180
- length = len(data["observations"])
181
-
182
- print("Amount Of Sequences: ", len(dataset))
183
- print("Longest Sequence: ", length)
184
- total_values = (2 * Constants.state_dim + Constants.action_dim + 2) * length * len(dataset)
185
- print("Total values to store: ", total_values)
186
-
187
- # create or overwrite pickle file
188
- with open(Constants.file_to_save, "wb") as f:
189
- pickle.dump(dataset, f)
190
-
191
- print("========================= Writing Completed ============================")
192
- file_size = os.stat(Constants.file_to_save).st_size
193
- if file_size > 1e+6:
194
- string_byte = "(" + str(round(file_size / 1e+6)) + " MB)"
195
- else:
196
- string_byte = "(" + str(round(file_size / 1e+3)) + " kB)"
197
- print("==> Data saved in", Constants.file_to_save, string_byte)
198
-
199
-
200
- if __name__ == '__main__':
201
- generate_data()