Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

dont wait for send op under dygraph pp #46209

Merged
merged 2 commits into from
Sep 20, 2022
Merged

dont wait for send op under dygraph pp #46209

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6b4e744
bug fix for allgather partial wait
FeixLiu Sep 19, 2022
0acf4db
dont wait send
FeixLiu Sep 19, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
57 changes: 27 additions & 30 deletions python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -329,23 +329,21 @@ def _p2p_helper(tensor_send_next, tensor_send_prev, recv_prev, recv_next):
for d in tensor_send_prev:
if _in_legacy_dygraph():
paddle.distributed.wait(d, use_calc_stream=True)
tasks.append(
send_partial(d,
dst=0,
nranks=mp_degree,
rank_id=mp_rank,
group=_hcg.send_prev_group,
use_calc_stream=False))
else:
if _in_legacy_dygraph():
paddle.distributed.wait(tensor_send_prev, use_calc_stream=True)
tasks.append(
send_partial(tensor_send_prev,
send_partial(d,
dst=0,
nranks=mp_degree,
rank_id=mp_rank,
group=_hcg.send_prev_group,
use_calc_stream=False))
use_calc_stream=False)
else:
if _in_legacy_dygraph():
paddle.distributed.wait(tensor_send_prev, use_calc_stream=True)
send_partial(tensor_send_prev,
dst=0,
nranks=mp_degree,
rank_id=mp_rank,
group=_hcg.send_prev_group,
use_calc_stream=False)

if tensor_recv_prev is not None:
if isinstance(tensor_recv_prev, tuple):
Expand All @@ -371,23 +369,21 @@ def _p2p_helper(tensor_send_next, tensor_send_prev, recv_prev, recv_next):
for d in tensor_send_next:
if _in_legacy_dygraph():
paddle.distributed.wait(d, use_calc_stream=True)
tasks.append(
send_partial(d,
dst=1,
nranks=mp_degree,
rank_id=mp_rank,
group=_hcg.send_next_group,
use_calc_stream=False))
else:
if _in_legacy_dygraph():
paddle.distributed.wait(tensor_send_next, use_calc_stream=True)
tasks.append(
send_partial(tensor_send_next,
send_partial(d,
dst=1,
nranks=mp_degree,
rank_id=mp_rank,
group=_hcg.send_next_group,
use_calc_stream=False))
use_calc_stream=False)
else:
if _in_legacy_dygraph():
paddle.distributed.wait(tensor_send_next, use_calc_stream=True)
send_partial(tensor_send_next,
dst=1,
nranks=mp_degree,
rank_id=mp_rank,
group=_hcg.send_next_group,
use_calc_stream=False)

if tensor_recv_next is not None:
if isinstance(tensor_recv_next, tuple):
Expand Down Expand Up @@ -438,10 +434,11 @@ def _p2p_helper(tensor_send_next, tensor_send_prev, recv_prev, recv_next):
group=mp_group,
use_calc_stream=True))

for task in tasks:
# wait partial all gather tasks
if task is not None:
task.wait()
if in_dygraph_mode():
for task in tasks:
# wait partial all gather tasks
if task is not None:
task.wait()

return tensor_recv_prev, tensor_recv_next

Expand Down