Differences with torch.distributed.barrier

View Source On Gitee

torch.distributed.barrier

torch.distributed.barrier(
    group=None,
    async_op=False,
    device_ids=None
)

For more information, see torch.distributed.barrier

mindspore.communication.comm_func.barrier

mindspore.communication.comm_func.barrier(group=GlobalComm.WORLD_COMM_GROUP)

For more information, see mindspore.communication.comm_func.barrier

Differences

API function of MindSpore is not consistent with that of PyTorch.

PyTorch: The inputs contains the communication group group, the async op flag async_op. And the working device lists device_ids. The return is a async work handle if async_op=True, otherwise is None.

MindSpore:The inputs contains the communication group group. The async op flag async_op and the working device lists device_ids are not supported.

Categories

Subcategories

PyTorch

MindSpore

Difference

Parameters

Parameter 1

group

group

No difference

Parameter 2

async_op

-

PyTorch: the async op flag. MindSpore: does not have this parameter.

Parameter 3

device_ids

-

PyTorch: the working device lists. MindSpore: does not have this parameter.