Function Differences with torch.distributed.all_gather

View Source On Gitee

torch.distributed.all_gather

torch.distributed.all_gather(
    tensor_list,
    tensor,
    group=None,
    async_op=False
)

For more information, see torch.distributed.all_gather.

mindspore.ops.AllGather

class mindspore.ops.AllGather(group=GlobalComm.WORLD_COMM_GROUP)(input_x)

For more information, see mindspore.ops.AllGather.

Differences

PyTorch: The inputs are the tensor broadcasted by the current process tensor, the communication group group and the async op flag async_op. The output is tensor_list after AllGather op, whose type is list[Tensor] and the length is the number of devices in the communication group. The return is a async work handle if async_op=True, otherwise is None.

MindSpore: The input of this interface is input_x that is a tensor. The output is a tensor, whose first dimension equals the number of devices in the communication group. This interface currently does not support the configuration of async_op.

Class

Sub-class

PyTorch

MindSpore

Difference

Param

Param 1

tensor_list

-

PyTorch: the output after AllGather. MindSpore does not have this parameter

Param 2

tensor

-

PyTorch: the tensor broadcasted by the current process. MindSpore does not have this parameter

Param 3

group

group

No dfference

Param 4

async_op

-

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

Input

Single input

-

input_x

PyTorch: not applied. MindSpore: the input tensor of AllGather.