[perf]fix current stream (#11870)
Signed-off-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
@@ -357,10 +357,7 @@ class GroupCoordinator:
|
||||
return out
|
||||
pynccl_comm = self.pynccl_comm
|
||||
assert pynccl_comm is not None
|
||||
# TODO: pynccl should not use `stream=`
|
||||
# it can just always use the current stream.
|
||||
out = pynccl_comm.all_reduce(input_,
|
||||
stream=torch.cuda.current_stream())
|
||||
out = pynccl_comm.all_reduce(input_)
|
||||
if out is None:
|
||||
# fall back to the default all-reduce using PyTorch.
|
||||
# this usually happens during testing.
|
||||
|
||||
Reference in New Issue
Block a user