mirror of https://github.com/InternLM/InternLM
fix moe zero mode bugs
parent
bbb5651582
commit
c801336732
|
@ -150,7 +150,7 @@ class HybridZeroOptimizer(BaseOptimizer):
|
|||
# if zero is used, expert dp group will use ParallelMode.EXPERT_DATA as the real zero mode
|
||||
zero_mode = (
|
||||
ParallelMode.ZERO1
|
||||
if param_group["dp_mode"] == gpc.get_world_size(ParallelMode.ZERO1) == 1 or ParallelMode.DATA
|
||||
if gpc.get_world_size(ParallelMode.ZERO1) == 1 or param_group["dp_mode"] == ParallelMode.DATA
|
||||
else ParallelMode.EXPERT_DATA
|
||||
)
|
||||
self._zero_local_rank.append(gpc.get_local_rank(zero_mode))
|
||||
|
|
Loading…
Reference in New Issue