-
Notifications
You must be signed in to change notification settings - Fork 0
/
learning_infer.py
173 lines (135 loc) · 6.48 KB
/
learning_infer.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
import configparser
import importlib
import json
import logging
import os
import pickle
import numpy as np
from learning_advisor.gym_db.common import EnvironmentType
from learning_advisor.stable_baselines.common.vec_env import DummyVecEnv, SubprocVecEnv, VecNormalize
from learning_advisor.learning_utils.workload import Query, Workload
BUDGET = 500
VERY_HIGH_BUDGET = 1000000
def pre_infer_obj(exp_load, model_load, env_load, db_conf=None):
with open(exp_load, "rb") as rf:
swirl_exp = pickle.load(rf)
if "max_indexes" not in swirl_exp.exp_config.keys():
swirl_exp.exp_config["max_indexes"] = 5
if db_conf is not None:
swirl_exp.schema.db_config = db_conf
swirl_model = swirl_exp.model_type.load(model_load)
swirl_model.training = False
ParallelEnv = SubprocVecEnv if swirl_exp.exp_config["parallel_environments"] > 1 else DummyVecEnv
training_env = ParallelEnv([swirl_exp.make_env(env_id,
environment_type=EnvironmentType.TRAINING,
workloads_in=None,
db_config=swirl_exp.schema.db_config)
# for env_id in range(1)])
for env_id in range(swirl_exp.exp_config["parallel_environments"])])
swirl_model.set_env(VecNormalize.load(env_load, training_env))
swirl_model.env.training = False
return swirl_exp, swirl_model
def get_swirl_res(swirl_exp, query_text, swirl_model):
if query_text is None:
eval_workload = swirl_exp.workload_generator.wl_testing[0]
elif isinstance(query_text, list):
eval_workload = list()
for qid, sql in enumerate(query_text):
query = Query(qid, sql, frequency=1)
# assign column value to `query` object.
swirl_exp.workload_generator._store_indexable_columns(query)
workload = Workload([query], description="")
workload.budget = BUDGET
eval_workload.append(workload)
elif query_text.endswith(".pickle"):
with open(query_text, "rb") as rf:
eval_workload = pickle.load(rf)[0]
n_eval_episodes = len(eval_workload)
evaluation_env = swirl_exp.DummyVecEnv(
[swirl_exp.make_env(0, EnvironmentType.TESTING,
workloads_in=eval_workload,
db_config=swirl_exp.schema.db_config)])
evaluation_env = swirl_exp.VecNormalize(
evaluation_env, norm_obs=True, norm_reward=False,
gamma=swirl_exp.exp_config["rl_algorithm"]["gamma"], training=False
)
training_env = swirl_model.get_vec_normalize_env()
swirl_exp.sync_envs_normalization(training_env, evaluation_env)
logging.disable(logging.WARNING)
swirl_exp.evaluate_policy(swirl_model, evaluation_env, n_eval_episodes)
logging.disable(logging.INFO)
performances = evaluation_env.get_attr("episode_performances")[0]
return performances
def get_drlindex_res(drlindex_exp, query_text, drlindex_model):
eval_workload = list()
for qid, sql in enumerate(query_text):
query = Query(qid, sql, frequency=1)
# assign column value to `query` object.
drlindex_exp.workload_generator._store_indexable_columns(query)
workload = Workload([query], description="")
workload.budget = VERY_HIGH_BUDGET
eval_workload.append(workload)
n_eval_episodes = len(eval_workload)
evaluation_env = drlindex_exp.DummyVecEnv(
[drlindex_exp.make_env(0, EnvironmentType.TESTING,
workloads_in=eval_workload,
db_config=drlindex_exp.schema.db_config)])
evaluation_env = drlindex_exp.VecNormalize(
evaluation_env, norm_obs=True, norm_reward=False,
gamma=drlindex_exp.exp_config["rl_algorithm"]["gamma"], training=False
)
training_env = drlindex_model.get_vec_normalize_env()
drlindex_exp.sync_envs_normalization(training_env, evaluation_env)
logging.disable(logging.WARNING)
drlindex_exp.evaluate_policy(drlindex_model, evaluation_env, n_eval_episodes)
logging.disable(logging.INFO)
performances = evaluation_env.get_attr("episode_performances")[0]
return performances
def get_dqn_res(dqn_exp, query_text, dqn_model):
eval_workload = list()
for qid, sql in enumerate(query_text):
query = Query(qid, sql, frequency=1)
# assign column value to `query` object.
dqn_exp.workload_generator._store_indexable_columns(query)
workload = Workload([query], description="")
workload.budget = VERY_HIGH_BUDGET
eval_workload.append(workload)
n_eval_episodes = len(eval_workload)
evaluation_env = dqn_exp.DummyVecEnv(
[dqn_exp.make_env(0, EnvironmentType.TESTING,
workloads_in=eval_workload,
db_config=dqn_exp.schema.db_config)])
evaluation_env = dqn_exp.VecNormalize(
evaluation_env, norm_obs=True, norm_reward=False,
gamma=dqn_exp.exp_config["rl_algorithm"]["gamma"], training=False
)
training_env = dqn_model.get_vec_normalize_env()
dqn_exp.sync_envs_normalization(training_env, evaluation_env)
logging.disable(logging.WARNING)
dqn_exp.evaluate_policy(dqn_model, evaluation_env, n_eval_episodes)
logging.disable(logging.INFO)
performances = evaluation_env.get_attr("episode_performances")[0]
return performances
def get_eval_env(swirl_exp, swirl_model, qtext_list, budget):
if qtext_list.endswith(".pickle"):
with open(qtext_list, "rb") as rf:
eval_workload = pickle.load(rf)[0]
else:
eval_workload = list()
for qid, sql in enumerate(qtext_list):
query = Query(qid, sql, frequency=1)
swirl_exp.workload_generator._store_indexable_columns(query)
workload = Workload([query], description="")
workload.budget = budget
eval_workload.append(workload)
evaluation_env = swirl_exp.DummyVecEnv(
[swirl_exp.make_env(0, EnvironmentType.TESTING,
workloads_in=eval_workload,
db_config=swirl_exp.schema.db_config)])
evaluation_env = swirl_exp.VecNormalize(
evaluation_env, norm_obs=True, norm_reward=False,
gamma=swirl_exp.exp_config["rl_algorithm"]["gamma"], training=False
)
training_env = swirl_model.get_vec_normalize_env()
swirl_exp.sync_envs_normalization(training_env, evaluation_env)
return evaluation_env