[hotfix] Add layer norm gradients all-reduce for sequence parallel.#4915

littsk merged 4 commits intohpcaitech:hotfix/sequence_parallelhpcaitech/ColossalAI:hotfix/sequence_parallelfrom

littsk:hotfix/add_grad_all_reduce_for_sequence_parallellittsk/ColossalAI:hotfix/add_grad_all_reduce_for_sequence_parallelCopy head branch name to clipboard