def libE_local(sim_specs, gen_specs, exit_criteria,
persis_info, alloc_specs, libE_specs, H0):
"Main routine for thread/process launch of libE."
nworkers = libE_specs['nworkers']
check_inputs(libE_specs, alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
exctr = Executor.executor
if exctr is not None:
local_host = [socket.gethostname()]
exctr.add_comm_info(libE_nodes=local_host, serial_setup=True)
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
# Launch worker team and set up logger
wcomms = start_proc_team(nworkers, sim_specs, gen_specs, libE_specs)
if not libE_specs.get('disable_log_files', False):
manager_logging_config()
# Set up cleanup routine to shut down worker team
def cleanup():
"Handler to clean up comms team."
kill_proc_team(wcomms, timeout=libE_specs.get('worker_timeout', 1))
# Run generic manager
return libE_manager(wcomms, sim_specs, gen_specs, exit_criteria,
persis_info, alloc_specs, libE_specs, hist,
on_cleanup=cleanup)
def libE_mpi_manager(mpi_comm, sim_specs, gen_specs, exit_criteria,
persis_info, alloc_specs, libE_specs, H0):
"Manager routine run at rank 0."
from libensemble.comms.mpi import MainMPIComm
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
# Lauch worker team
wcomms = [MainMPIComm(mpi_comm, w) for w in range(1, mpi_comm.Get_size())]
if not libE_specs.get('disable_log_files', False):
manager_logging_config()
# Set up abort handler
def on_abort():
"Shut down MPI on error."
comms_abort(mpi_comm)
# Run generic manager
return libE_manager(wcomms,
sim_specs,
gen_specs,
exit_criteria,
persis_info,
alloc_specs,
libE_specs,
hist,
on_abort=on_abort)
def libE_local(sim_specs, gen_specs, exit_criteria, persis_info, alloc_specs,
libE_specs, H0):
"Main routine for thread/process launch of libE."
nworkers = libE_specs['nworkers']
check_inputs(libE_specs, alloc_specs, sim_specs, gen_specs, exit_criteria,
H0)
jobctl = JobController.controller
if jobctl is not None:
local_host = socket.gethostname()
jobctl.add_comm_info(libE_nodes=local_host, serial_setup=True)
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
# Launch worker team and set up logger
wcomms = start_proc_team(nworkers, sim_specs, gen_specs, libE_specs)
manager_logging_config()
# Set up cleanup routine to shut down worker team
def cleanup():
"Handler to clean up comms team."
kill_proc_team(wcomms, timeout=libE_specs.get('worker_timeout'))
# Run generic manager
return libE_manager(wcomms,
sim_specs,
gen_specs,
exit_criteria,
persis_info,
alloc_specs,
libE_specs,
hist,
on_cleanup=cleanup)
def hist_setup2(sim_max=10, H0_in=[]):
sim_specs, gen_specs, exit_criteria = make_criteria_and_specs_1(
simx=sim_max)
alloc_specs = {
'alloc_f': give_sim_work_first,
'out': [('allocated', bool)]
} # default for libE
H0 = H0_in
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
return hist, sim_specs, gen_specs, exit_criteria, alloc_specs
def test_decide_work_and_resources():
sim_specs, gen_specs, exit_criteria = setup.make_criteria_and_specs_1()
hist = History(al, sim_specs, gen_specs, exit_criteria, H0)
mgr = man.Manager(hist, libE_specs, al, sim_specs, gen_specs, exit_criteria)
W = mgr.W
# Don't give out work when all workers are active
W['active'] = 1
Work, persis_info = al['alloc_f'](W, hist.H, sim_specs, gen_specs, al, {})
assert len(Work) == 0
def libE_tcp_mgr(sim_specs, gen_specs, exit_criteria, persis_info, alloc_specs,
libE_specs, H0):
"Main routine for TCP multiprocessing launch of libE at manager."
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
# Set up a worker launcher
launchf = libE_tcp_worker_launcher(libE_specs)
# Get worker launch parameters and fill in defaults for TCP/IP conn
if 'nworkers' in libE_specs:
workers = None
nworkers = libE_specs['nworkers']
elif 'workers' in libE_specs:
workers = libE_specs['workers']
nworkers = len(workers)
ip = libE_specs.get('ip', None) or get_ip()
port = libE_specs.get('port', 0)
authkey = libE_specs.get('authkey', libE_tcp_authkey())
with ServerQCommManager(port, authkey.encode('utf-8')) as manager:
# Get port if needed because of auto-assignment
if port == 0:
_, port = manager.address
if not libE_specs.get('disable_log_files', False):
manager_logging_config()
logger.info("Launched server at ({}, {})".format(ip, port))
# Launch worker team and set up logger
worker_procs, wcomms =\
libE_tcp_start_team(manager, nworkers, workers,
ip, port, authkey, launchf)
def cleanup():
"Handler to clean up launched team."
for wp in worker_procs:
launcher.cancel(wp, timeout=libE_specs.get('worker_timeout'))
# Run generic manager
return libE_manager(wcomms,
sim_specs,
gen_specs,
exit_criteria,
persis_info,
alloc_specs,
libE_specs,
hist,
on_cleanup=cleanup)
def libE_local(sim_specs, gen_specs, exit_criteria, persis_info, alloc_specs,
libE_specs, H0):
"Main routine for thread/process launch of libE."
nworkers = libE_specs['nworkers']
check_inputs(libE_specs, alloc_specs, sim_specs, gen_specs, exit_criteria,
H0)
exctr = Executor.executor
if exctr is not None:
local_host = [socket.gethostname()]
exctr.add_comm_info(libE_nodes=local_host, serial_setup=True)
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
# On Python 3.8 on macOS, the default start method for new processes was
# switched to 'spawn' by default due to 'fork' potentially causing crashes.
# These crashes haven't yet been observed with libE, but with 'spawn' runs,
# warnings about leaked semaphore objects are displayed instead.
# The next several statements enforce 'fork' on macOS (Python 3.8)
if os.uname().sysname == 'Darwin':
from multiprocessing import set_start_method
set_start_method('fork', force=True)
# Launch worker team and set up logger
wcomms = start_proc_team(nworkers, sim_specs, gen_specs, libE_specs)
if not libE_specs.get('disable_log_files', False):
manager_logging_config()
# Set up cleanup routine to shut down worker team
def cleanup():
"Handler to clean up comms team."
kill_proc_team(wcomms, timeout=libE_specs.get('worker_timeout', 1))
# Run generic manager
return libE_manager(wcomms,
sim_specs,
gen_specs,
exit_criteria,
persis_info,
alloc_specs,
libE_specs,
hist,
on_cleanup=cleanup)
def libE(sim_specs,
gen_specs,
exit_criteria,
persis_info={},
alloc_specs={
'alloc_f': give_sim_work_first,
'out': [('allocated', bool)]
},
libE_specs={
'comm': MPI.COMM_WORLD,
'color': 0
},
H0=[]):
"""
libE(sim_specs, gen_specs, exit_criteria, persis_info={}, alloc_specs={'alloc_f': give_sim_work_first, 'out':[('allocated',bool)]}, libE_specs={'comm': MPI.COMM_WORLD, 'color': 0}, H0 =[])
This is the outer libEnsemble routine. If the rank in libE_specs['comm'] is
0, manager_main is run. Otherwise, worker_main is run.
If an exception is encountered by the manager or workers, the history array
is dumped to file and MPI abort is called.
Parameters
----------
sim_specs: :obj:`dict`
Specifications for the simulation function
:doc:`(example)<data_structures/sim_specs>`
gen_specs: :obj:`dict`
Specifications for the generator function
:doc:`(example)<data_structures/gen_specs>`
exit_criteria: :obj:`dict`
Tell libEnsemble when to stop a run
:doc:`(example)<data_structures/exit_criteria>`
persis_info: :obj:`dict`, optional
Persistent information to be passed between user functions
:doc:`(example)<data_structures/persis_info>`
alloc_specs: :obj:`dict`, optional
Specifications for the allocation function
:doc:`(example)<data_structures/alloc_specs>`
libE_specs: :obj:`dict`, optional
Specifications for libEnsemble
:doc:`(example)<data_structures/libE_specs>`
H0: :obj:`dict`, optional
A previous libEnsemble history to be prepended to the history in the
current libEnsemble run
:doc:`(example)<data_structures/history_array>`
Returns
-------
H: :obj:`dict`
History array storing rows for each point. :doc:`(example)<data_structures/history_array>`
Dictionary containing persistent info
persis_info: :obj:`dict`
Final state of persistent information
:doc:`(example)<data_structures/persis_info>`
exit_flag: :obj:`int`
Flag containing job status: 0 = No errors,
1 = Exception occured and MPI aborted,
2 = Manager timed out and ended simulation
"""
#sys.excepthook = comms_abort(libE_specs['comm'])
H = exit_flag = []
libE_specs = check_inputs(libE_specs, alloc_specs, sim_specs, gen_specs,
exit_criteria, H0)
if libE_specs['comm'].Get_rank() == 0:
CalcInfo.make_statdir()
libE_specs['comm'].Barrier()
if libE_specs['comm'].Get_rank() == 0:
hist = History(alloc_specs, sim_specs, gen_specs, exit_criteria, H0)
try:
persis_info, exit_flag = manager_main(hist, libE_specs,
alloc_specs, sim_specs,
gen_specs, exit_criteria,
persis_info)
except Exception as e:
# Manager exceptions are fatal
eprint(traceback.format_exc())
eprint("\nManager exception raised .. aborting ensemble:\n"
) #datetime
eprint(
"\nDumping ensemble history with {} sims evaluated:\n".format(
hist.sim_count)) #datetime
filename = 'libE_history_at_abort_' + str(hist.sim_count) + '.npy'
np.save(filename, hist.trim_H())
sys.stdout.flush()
sys.stderr.flush()
#sys.excepthook = comms_abort(libE_specs['comm'])
comms_abort(libE_specs['comm'])
#raise
else:
logger.debug("Manager exiting")
print(libE_specs['comm'].Get_size(), exit_criteria)
sys.stdout.flush()
else:
try:
worker_main(libE_specs, sim_specs, gen_specs)
except Exception as e:
eprint(
"\nWorker exception raised on rank {} .. aborting ensemble:\n".
format(libE_specs['comm'].Get_rank()))
eprint(traceback.format_exc())
sys.stdout.flush()
sys.stderr.flush()
#First try to signal manager to dump history
comms_signal_abort_to_man(libE_specs['comm'])
#comms_abort(libE_specs['comm'])
else:
logger.debug("Worker {} exiting".format(
libE_specs['comm'].Get_rank()))
# Create calc summary file
libE_specs['comm'].Barrier()
if libE_specs['comm'].Get_rank() == 0:
CalcInfo.merge_statfiles()
H = hist.trim_H()
return H, persis_info, exit_flag
