-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtasks.py
364 lines (235 loc) · 12.7 KB
/
tasks.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
import logging
import os
import traceback
from pickle import PicklingError
import billiard
from billiard.exceptions import SoftTimeLimitExceeded, WorkerLostError
from billiard.pool import MaybeEncodingError
from celery import Celery
from celery.exceptions import CeleryError
from celery.signals import after_setup_logger
import projectconfig
from relnet.environment.graph_edge_env import GraphEdgeEnv
from relnet.utils.config_utils import get_logger_instance, date_format, logging_format
from relnet.utils.general_utils import create_generator_instance
app_settings = projectconfig.get_project_config()
app = Celery('tasks',
backend='amqp://',
broker=app_settings.CELERY_BROKER_URL)
app.conf.update(accept_content=['application/x-python-serialize', 'application/json'],
task_serializer=app_settings.CELERY_TASK_SERIALIZER,
result_serializer=app_settings.CELERY_RESULT_SERIALIZER,
task_acks_late=app_settings.CELERY_TASK_ACKS_LATE,
worker_prefetch_multiplier=app_settings.CELERYD_PREFETCH_MULTIPLIER,
worker_max_tasks_per_child=app_settings.WORKER_MAX_TASKS_PER_CHILD,
worker_concurrency=app_settings.get_number_worker_threads(),
broker_heartbeat=app_settings.BROKER_HEARTBEAT,
broker_pool_limit=app_settings.BROKER_POOL_LIMIT,
timezone='Europe/London',
enable_utc=False)
@after_setup_logger.connect
def setup_loggers(logger, *args, **kwargs):
formatter = logging.Formatter(fmt=logging_format, datefmt=date_format)
fh = logging.FileHandler(f"/tmp/celery-{os.getenv('HOSTNAME')}.log")
fh.setLevel(logging.INFO)
fh.setFormatter(formatter)
logger.addHandler(fh)
class DistributedTaskException(Exception):
pass
ExpectedErrors = (FileNotFoundError,
ValueError,
RuntimeError,
SystemError,
ConnectionResetError,
WorkerLostError,
CeleryError,
billiard.pool.MaybeEncodingError,
PicklingError)
@app.task(bind=True,
rate_limit="60/m",
retry_kwargs={'max_retries': 15},
autoretry_for=(DistributedTaskException,),
retry_backoff=True,
retry_jitter=True)
def optimize_hyperparams_task(self,
agent,
objective_function,
network_generator,
experiment_conditions,
file_paths,
hyperparams,
hyperparams_id,
is_only_hyp_comb,
model_seed,
model_identifier_prefix,
train_kwargs=None,
eval_make_action_kwargs=None,
additional_opts=None):
gen_params = experiment_conditions.gen_params
network_generator_instance = create_generator_instance(network_generator, file_paths)
models_dir = file_paths.models_dir
timings_dir = file_paths.timings_dir
obj_fun_kwargs = {"random_seed": experiment_conditions.obj_fun_seed,
"num_mc_sims_multiplier": experiment_conditions.num_mc_sims_multiplier}
env = GraphEdgeEnv(objective_function(), obj_fun_kwargs, experiment_conditions.possible_edge_percentage,
conn_radius_modifier=network_generator.conn_radius_modifiers[experiment_conditions.restriction_mechanism],
restriction_mechanism=experiment_conditions.restriction_mechanism)
agent_instance = agent(env)
run_options = {}
run_options["random_seed"] = model_seed
run_options["models_path"] = models_dir
run_options["timings_path"] = timings_dir
run_options["log_progress"] = True
log_filename = str(file_paths.construct_log_filepath())
run_options["log_filename"] = log_filename
run_options["model_identifier_prefix"] = model_identifier_prefix
run_options["restore_model"] = False
run_options["use_geometric_features"] = experiment_conditions.use_geometric_features
run_options["log_timings"] = False
run_options.update((additional_opts or {}))
agent_instance.setup(run_options, hyperparams)
if is_only_hyp_comb:
average_reward = 1.0
if file_paths.hyperopt_results_dir is not None:
hyperopt_result_file = f"{file_paths.hyperopt_results_dir.absolute()}/" + \
file_paths.construct_best_validation_file_name(model_identifier_prefix)
hyperopt_result_out = open(hyperopt_result_file, 'w')
hyperopt_result_out.write('%.6f\n' % (average_reward))
hyperopt_result_out.close()
else:
try:
validation_graphs = network_generator_instance.generate_many(gen_params, experiment_conditions.validation_seeds)
if agent.is_trainable:
train_graphs = network_generator_instance.generate_many(gen_params, experiment_conditions.train_seeds)
max_steps = experiment_conditions.agent_budgets[objective_function.name][agent.algorithm_name]
agent_train_kwargs = (train_kwargs or {})
try:
agent_instance.train(train_graphs, validation_graphs, max_steps, **agent_train_kwargs)
except TypeError:
print(f"the graph at fault is from gen {network_generator.name} with seed {experiment_conditions.train_seeds[0]}!")
agent_eval_kwargs = (eval_make_action_kwargs or {})
average_reward = agent_instance.eval(validation_graphs, make_action_kwargs=agent_eval_kwargs)
if file_paths.hyperopt_results_dir is not None:
hyperopt_result_file = f"{file_paths.hyperopt_results_dir.absolute()}/" + \
file_paths.construct_best_validation_file_name(model_identifier_prefix)
hyperopt_result_out = open(hyperopt_result_file, 'w')
hyperopt_result_out.write('%.6f\n' % (average_reward))
hyperopt_result_out.close()
agent_instance.finalize()
except ExpectedErrors as error:
raise DistributedTaskException() from error
return hyperparams, objective_function.name, network_generator_instance.name, average_reward
@app.task(bind=True,
rate_limit="1000/m",
soft_time_limit=24 * 60 * 60,
time_limit=25 * 60 * 60,
autoretry_for=(DistributedTaskException,),
retry_kwargs={'max_retries': 15},
retry_backoff=True,
retry_jitter=True)
def evaluate_for_network_seed_task(self,
agent,
objective_function,
network_generator,
best_hyperparams,
best_hyperparams_id,
experiment_conditions,
file_paths,
net_seed,
model_seeds,
graph_id=None,
eval_make_action_kwargs=None,
additional_opts=None):
try:
log_filename = str(file_paths.construct_log_filepath())
logger = get_logger_instance(log_filename)
local_results = []
gen_params = experiment_conditions.gen_params
network_generator_instance = create_generator_instance(network_generator, file_paths)
models_dir = file_paths.models_dir
timings_dir = file_paths.timings_dir
obj_fun_kwargs = {"random_seed": experiment_conditions.obj_fun_seed,
"num_mc_sims_multiplier": experiment_conditions.num_mc_sims_multiplier}
env = GraphEdgeEnv(objective_function(), obj_fun_kwargs, experiment_conditions.possible_edge_percentage,
conn_radius_modifier=network_generator.conn_radius_modifiers[
experiment_conditions.restriction_mechanism],
restriction_mechanism=experiment_conditions.restriction_mechanism)
for model_seed in model_seeds:
try:
agent_instance = agent(env)
run_options = {}
run_options['random_seed'] = model_seed
run_options["restore_model"] = True
model_identifier_prefix = file_paths.construct_model_identifier_prefix(agent.algorithm_name,
objective_function.name,
network_generator_instance.name,
model_seed,
best_hyperparams_id,
graph_id=graph_id)
run_options["model_identifier_prefix"] = model_identifier_prefix
run_options["models_path"] = models_dir
run_options["timings_path"] = timings_dir
run_options["log_progress"] = True
run_options["log_filename"] = log_filename
run_options["use_geometric_features"] = experiment_conditions.use_geometric_features
run_options["log_timings"] = True
run_options.update((additional_opts or {}))
agent_instance.setup(run_options, best_hyperparams)
result_row = {}
result_row['network_generator'] = network_generator_instance.name
if graph_id is not None:
result_row['graph_id'] = graph_id
result_row['objective_function'] = objective_function.name
result_row['network_seed'] = net_seed
result_row['algorithm'] = agent.algorithm_name
result_row['agent_seed'] = model_seed
result_row['network_size'] = gen_params['n']
test_g_list = [network_generator_instance.generate(gen_params, net_seed)]
agent_eval_kwargs = (eval_make_action_kwargs or {})
result_row['cummulative_reward'] = agent_instance.eval(test_g_list, make_action_kwargs=agent_eval_kwargs)
local_results.append(result_row)
agent_instance.finalize()
except ExpectedErrors as error:
logger.warn("faced the following exception:")
logger.warn(traceback.format_exc(limit=100))
raise DistributedTaskException() from error
return local_results
except SoftTimeLimitExceeded:
logger.warn(f"Task with id {self.request.id} went over the time limit. aborting...")
return []
@app.task(bind=True,
soft_time_limit=24 * 60 * 60,
time_limit=25 * 60 * 60,
autoretry_for=(DistributedTaskException,),
retry_kwargs={'max_retries': 5},
retry_backoff=True,
retry_jitter=True)
def compute_property_for_network_seed(self,
objective_function,
network_generator,
experiment_conditions,
file_paths,
net_seed):
try:
log_filename = str(file_paths.construct_log_filepath())
logger = get_logger_instance(log_filename)
gen_params = experiment_conditions.gen_params
obj_fun_kwargs = {"random_seed": experiment_conditions.obj_fun_seed,
"num_mc_sims": experiment_conditions.num_mc_sims}
gen_instance = network_generator()
try:
graph_instance = gen_instance.generate(gen_params, net_seed)
result_row = {}
result_row['network_generator'] = network_generator.name
result_row['objective_function'] = objective_function.name
result_row['network_seed'] = net_seed
result_row['network_size'] = gen_params['n']
result_row['value'] = objective_function.compute(graph_instance, **obj_fun_kwargs)
except ExpectedErrors as error:
logger.warn("faced the following exception:")
logger.warn(traceback.format_exc(limit=100))
raise DistributedTaskException() from error
return result_row
except SoftTimeLimitExceeded:
logger.warn(f"Task with id {self.request.id} went over the time limit. aborting...")
return []