def run_sessions(self):
logger.info('Running sessions')
if util.get_lab_mode() in (
'train', 'eval') and self.spec['meta']['max_session'] > 1:
# when training a single spec over multiple sessions
session_datas = self.parallelize_sessions()
else:
session_datas = []
for _s in range(self.spec['meta']['max_session']):
self.info_space.tick('session')
session = self.SessionClass(deepcopy(self.spec),
deepcopy(self.info_space))
session_data = session.run()
session_datas.append(session_data)
if analysis.is_unfit(session_data, session):
break
return session_datas
def run_sessions(self):
logger.info('Running sessions')
info_spaces = []
for _s in range(self.spec['meta']['max_session']):
self.info_space.tick('session')
info_spaces.append(deepcopy(self.info_space))
if util.get_lab_mode() == 'train' and len(info_spaces) > 1:
# when training a single spec over multiple sessions
session_datas = util.parallelize_fn(
self.init_session_and_run, info_spaces,
ps.get(self.spec['meta'], 'resources.num_cpus', util.NUM_CPUS))
else:
session_datas = []
for info_space in info_spaces:
session_data = self.init_session_and_run(info_space)
session_datas.append(session_data)
if analysis.is_unfit(session_data):
break
return session_datas
def run(self):
num_cpus = ps.get(self.spec['meta'], 'resources.num_cpus',
util.NUM_CPUS)
info_spaces = []
for _s in range(self.spec['meta']['max_session']):
self.info_space.tick('session')
info_spaces.append(deepcopy(self.info_space))
if util.get_lab_mode() == 'train' and len(info_spaces) > 1:
session_datas = util.parallelize_fn(self.init_session_and_run,
info_spaces, num_cpus)
else: # dont parallelize when debugging to allow render
session_datas = []
for info_space in info_spaces:
session_data = self.init_session_and_run(info_space)
session_datas.append(session_data)
if analysis.is_unfit(session_data):
break
self.session_data_dict = {
data.index[0]: data
for data in session_datas
}
self.data = analysis.analyze_trial(self)
self.close()
return self.data