diff --git a/colossalai/engine/gradient_handler/__init__.py b/colossalai/engine/gradient_handler/__init__.py index 6177da69b..2dea768ba 100644 --- a/colossalai/engine/gradient_handler/__init__.py +++ b/colossalai/engine/gradient_handler/__init__.py @@ -1,10 +1,9 @@ from ._base_gradient_handler import BaseGradientHandler from ._data_parallel_gradient_handler import DataParallelGradientHandler -from ._zero_gradient_handler import ZeROGradientHandler -from ._sequence_parallel_gradient_handler import SequenceParallelGradientHandler -from ._pipeline_parallel_gradient_handler import PipelineSharedModuleGradientHandler from ._moe_gradient_handler import MoeGradientHandler +from ._pipeline_parallel_gradient_handler import PipelineSharedModuleGradientHandler from ._sequence_parallel_gradient_handler import SequenceParallelGradientHandler +from ._zero_gradient_handler import ZeROGradientHandler __all__ = [ 'BaseGradientHandler', 'DataParallelGradientHandler', 'ZeROGradientHandler', 'PipelineSharedModuleGradientHandler',