def test_create_plan_from_proto_correctly(self):
from caffe2.python.net_builder import ops
with Node('trainer'), Task(name='my_task', num_instances=2) as task:
with ops.task_init():
globl = ops.Const(0)
with ops.task_instance_init():
local = ops.Const(0)
with ops.loop(100):
ops.Copy(globl, local)
with ops.task_instance_exit():
ops.Add([globl, local], [globl])
with ops.task_exit():
ops.Mul([globl, globl], [globl])
plan = core.Plan(task.get_step())
test_plan = core.Plan.create_from_proto(plan.Proto())
self.assertEqual(len(plan.Steps()), 1)
self.assertEqual(len(test_plan.Steps()), 1)
self.assertEqual(plan.Steps()[0].Name(), test_plan.Steps()[0].Name())
self.assertEqual(len(plan.Nets()), len(test_plan.Nets()))
for idx in range(0, len(plan.Nets())):
# When we create Net for test_plan, we will end up with new Net
# name with postfix.
net_1 = plan.Nets()[idx]
net_2 = test_plan.Nets()[idx]
trim_size = len(net_1.Name())
self.assertEqual(net_1.Name(), net_2.Name()[:trim_size])
def test_create_plan_from_proto_correctly(self):
from caffe2.python.net_builder import ops
with Node('trainer'), Task(name='my_task', num_instances=2) as task:
with ops.task_init():
globl = ops.Const(0)
with ops.task_instance_init():
local = ops.Const(0)
with ops.loop(100):
ops.Copy(globl, local)
with ops.task_instance_exit():
ops.Add([globl, local], [globl])
with ops.task_exit():
ops.Mul([globl, globl], [globl])
plan = core.Plan(task.get_step())
test_plan = core.Plan.create_from_proto(plan.Proto())
self.assertEqual(len(plan.Steps()), 1)
self.assertEqual(len(test_plan.Steps()), 1)
self.assertEqual(len(plan.Proto().network), 9)
self.assertEqual(len(test_plan.Proto().network), 9)
self.assertEqual(len(plan.Proto().execution_step), 1)
self.assertEqual(len(test_plan.Proto().execution_step), 1)
self.assertEqual(plan.Steps()[0].Name(), test_plan.Steps()[0].Name())
self.assertEqual(len(plan.Nets()), len(test_plan.Nets()))
for idx in range(0, len(plan.Nets())):
# When we create Net for test_plan, we will end up with new Net
# name with postfix.
net_1 = plan.Nets()[idx]
net_2 = test_plan.Nets()[idx]
trim_size = len(net_1.Name())
self.assertEqual(net_1.Name(), net_2.Name()[:trim_size])
def example_task():
with Task():
with ops.task_init():
one = ops.Const(1)
two = ops.Add([one, one])
with ops.task_init():
three = ops.Const(3)
accum = ops.Add([two, three])
# here, accum should be 5
with ops.task_exit():
# here, accum should be 6, since this executes after lines below
seven_1 = ops.Add([accum, one])
six = ops.Add([accum, one])
ops.Add([accum, one], [accum])
seven_2 = ops.Add([accum, one])
o6 = final_output(six)
o7_1 = final_output(seven_1)
o7_2 = final_output(seven_2)
with Task(num_instances=2):
with ops.task_init():
one = ops.Const(1)
with ops.task_instance_init():
local = ops.Const(2)
ops.Add([one, local], [one])
ops.LogInfo('ble')
return o6, o7_1, o7_2
def _runtime_threads_task(name, group, final_outputs, reader, num_threads,
output, capacity):
node_name = str(Node.current())
profiler_name = "{0}/{1}/{2}/{3}/{4}".format(
node_name,
"pipe",
name,
processor_name(input) if input else "NoInput",
processor_name(output) if output else "NoOutput")
with Task(name=name, group=group, outputs=final_outputs,
num_instances=num_threads) as task:
global_exit_net = core.Net('pipe:exit')
global_init_net = core.Net('pipe:init')
reader.setup_ex(global_init_net, global_exit_net)
init_net = core.Net('pipe:instance:init')
exit_net = core.Net('pipe:instance:exit')
read_nets, status, rec = reader.read_record_ex(init_net, exit_net)
init_net.ConstantFill(
[], [status],
shape=[],
value=False,
dtype=core.DataType.BOOL
)
if rec is not None:
out_queue, writer = _init_output(
output, capacity, global_init_net, global_exit_net)
write_nets, _ = writer.write_record_ex(
rec, init_net, exit_net, status)
else:
out_queue = None
write_nets = []
with ops.task_init():
ops.net(global_init_net)
with ops.task_instance_init():
ops.net(init_net)
timer_start_net = core.Net('timer_start')
timer = timer_start_net.TimerBegin([], counter_name=profiler_name)
timer_end_net = core.Net('timer_end')
timer_end_net.TimerEnd(timer, [])
ops.net(core.execution_step(
'body',
[timer_start_net] + list(read_nets) + list(write_nets) +
[timer_end_net],
should_stop_blob=status))
ops.net(timer_end_net)
with ops.task_instance_exit():
ops.net(exit_net)
with ops.task_exit():
ops.net(global_exit_net)
return out_queue, task
def test_multi_instance(self):
NUM_INSTANCES = 10
NUM_ITERS = 15
with TaskGroup() as tg:
with Task(num_instances=NUM_INSTANCES):
with ops.task_init():
counter1 = ops.CreateCounter([], ['global_counter'])
counter2 = ops.CreateCounter([], ['global_counter2'])
counter3 = ops.CreateCounter([], ['global_counter3'])
# both task_counter and local_counter should be thread local
with ops.task_instance_init():
task_counter = ops.CreateCounter([], ['task_counter'])
local_counter = ops.CreateCounter([], ['local_counter'])
with ops.loop(NUM_ITERS):
ops.CountUp(counter1)
ops.CountUp(task_counter)
ops.CountUp(local_counter)
# gather sum of squares of local counters to make sure that
# each local counter counted exactly up to NUM_ITERS, and
# that there was no false sharing of counter instances.
with ops.task_instance_exit():
count2 = ops.RetrieveCount(task_counter)
with ops.loop(ops.Mul([count2, count2])):
ops.CountUp(counter2)
# This should have the same effect as the above
count3 = ops.RetrieveCount(local_counter)
with ops.loop(ops.Mul([count3, count3])):
ops.CountUp(counter3)
# The code below will only run once
with ops.task_exit():
total1 = final_output(ops.RetrieveCount(counter1))
total2 = final_output(ops.RetrieveCount(counter2))
total3 = final_output(ops.RetrieveCount(counter3))
with LocalSession() as session:
session.run(tg)
self.assertEquals(total1.fetch(), NUM_INSTANCES * NUM_ITERS)
self.assertEquals(total2.fetch(), NUM_INSTANCES * (NUM_ITERS**2))
self.assertEquals(total3.fetch(), NUM_INSTANCES * (NUM_ITERS**2))
def test_multi_instance(self):
NUM_INSTANCES = 10
NUM_ITERS = 15
with TaskGroup() as tg:
with Task(num_instances=NUM_INSTANCES):
with ops.task_init():
counter1 = ops.CreateCounter([], ['global_counter'])
counter2 = ops.CreateCounter([], ['global_counter2'])
counter3 = ops.CreateCounter([], ['global_counter3'])
# both task_counter and local_counter should be thread local
with ops.task_instance_init():
task_counter = ops.CreateCounter([], ['task_counter'])
local_counter = ops.CreateCounter([], ['local_counter'])
with ops.loop(NUM_ITERS):
ops.CountUp(counter1)
ops.CountUp(task_counter)
ops.CountUp(local_counter)
# gather sum of squares of local counters to make sure that
# each local counter counted exactly up to NUM_ITERS, and
# that there was no false sharing of counter instances.
with ops.task_instance_exit():
count2 = ops.RetrieveCount(task_counter)
with ops.loop(ops.Mul([count2, count2])):
ops.CountUp(counter2)
# This should have the same effect as the above
count3 = ops.RetrieveCount(local_counter)
with ops.loop(ops.Mul([count3, count3])):
ops.CountUp(counter3)
# The code below will only run once
with ops.task_exit():
total1 = final_output(ops.RetrieveCount(counter1))
total2 = final_output(ops.RetrieveCount(counter2))
total3 = final_output(ops.RetrieveCount(counter3))
with LocalSession() as session:
session.run(tg)
self.assertEquals(total1.fetch(), NUM_INSTANCES * NUM_ITERS)
self.assertEquals(total2.fetch(), NUM_INSTANCES * (NUM_ITERS ** 2))
self.assertEquals(total3.fetch(), NUM_INSTANCES * (NUM_ITERS ** 2))
def proc(rec):
# executed once
with ops.task_init():
counter1 = ops.CreateCounter([], ['global_counter'])
counter2 = ops.CreateCounter([], ['global_counter2'])
counter3 = ops.CreateCounter([], ['global_counter3'])
# executed once per thread
with ops.task_instance_init():
task_counter = ops.CreateCounter([], ['task_counter'])
# executed on each iteration
ops.CountUp(counter1)
ops.CountUp(task_counter)
# executed once per thread
with ops.task_instance_exit():
with ops.loop(ops.RetrieveCount(task_counter)):
ops.CountUp(counter2)
ops.CountUp(counter3)
# executed once
with ops.task_exit():
totals[0] = final_output(ops.RetrieveCount(counter1))
totals[1] = final_output(ops.RetrieveCount(counter2))
totals[2] = final_output(ops.RetrieveCount(counter3))
return rec
def proc(rec):
# executed once
with ops.task_init():
counter1 = ops.CreateCounter([], ['global_counter'])
counter2 = ops.CreateCounter([], ['global_counter2'])
counter3 = ops.CreateCounter([], ['global_counter3'])
# executed once per thread
with ops.task_instance_init():
task_counter = ops.CreateCounter([], ['task_counter'])
# executed on each iteration
ops.CountUp(counter1)
ops.CountUp(task_counter)
# executed once per thread
with ops.task_instance_exit():
with ops.loop(ops.RetrieveCount(task_counter)):
ops.CountUp(counter2)
ops.CountUp(counter3)
# executed once
with ops.task_exit():
totals[0] = final_output(ops.RetrieveCount(counter1))
totals[1] = final_output(ops.RetrieveCount(counter2))
totals[2] = final_output(ops.RetrieveCount(counter3))
return rec
