## Motivation As tensor parallelism is implemented by shardformer, ColoTensor is not needed to be regard as distributed tensor anymore. ## Overview ColoParam should be remained, but all components related to distributed training can be removed.
Motivation
As tensor parallelism is implemented by shardformer, ColoTensor is not needed to be regard as distributed tensor anymore.
Overview
ColoParam should be remained, but all components related to distributed training can be removed.