Function Differences with torch.distributed.get_rank
torch.distributed.get_rank
torch.distributed.get_rank(group=None)
For more information, see torch.distributed.get_rank.
mindspore.communication.get_rank
mindspore.communication.get_rank(group=GlobalComm.WORLD_COMM_GROUP)
For more information, see mindspore.communication.get_rank.
Differences
PyTorch: The input of this interface is the communication group group.
The output is the rank of the process that calls this interface, and returns -1 if the process is not in group.
MindSpore: The input of this interface is the communication group group. The output is the rank of the process that calls this interface. Since get_rank should be called after init, the process that calls this interface should be in group. If get_rank is called before init, it will raise error.
Class |
Sub-class |
PyTorch |
MindSpore |
Difference |
|---|---|---|---|---|
Parameter |
Parameter 1 |
group |
group |
No difference |
