# create 4 trajectories
trajectories = project.new_trajectory(pdb_file, 100, 4)
tasks = map(engine.run, trajectories)
map(project.tasks.add, tasks)
# now start adaptive loop
def strategy_trajectory(loops, num):
for loop in range(loops):
trajectories = project.new_ml_trajectory(20, number=num)
tasks = map(engine.run, trajectories)
map(project.tasks.add, tasks)
yield [t.is_done for t in tasks]
ev1 = ExecutionPlan(strategy_trajectory(100, 10))
project.add_event(ev1)
def strategy_model(steps):
while ev1:
num = len(project.trajectories)
task = modeller.execute(list(project.trajectories))
project.tasks.add(task)
yield task.is_done
cond = project.on_ntraj(num + steps)
yield lambda: cond() or not ev1
ev2 = ExecutionPlan(strategy_model(100))
project.add_event(ev2)
try:
while project._events:
events = [
ExecutionPlan(strategy_trajectory(scheduler, 100, 10))
for scheduler in gpu_scheduler
]
map(project.add_event, events)
def strategy_model(scheduler, steps):
while any(events):
num = len(project.trajectories)
task = scheduler(modeller.execute(list(project.trajectories)))
yield task.is_done
cond = project.on_ntraj(num + steps)
yield lambda: cond() or not any(events)
ev3 = ExecutionPlan(strategy_model(pyemma_scheduler, steps))
project.add_event(ev3)
print
CURSOR_UP_ONE = '\x1b[1A'
ERASE_LINE = '\x1b[2K'
# try:
# while project._events:
# sys.stdout.write('# of trajectories : %8d / # of models : %8d \n' % (
# len(project.trajectories),
# len(project.models)
# ))
# sys.stdout.flush()
# time.sleep(1.0)
# sys.stdout.write(CURSOR_UP_ONE + ERASE_LINE)