v0.4.0
[Inference] Delete duplicated copy_vector (#5716)
add paged-attetionv2: support seq length split across thread block (#5707)
[shardformer] update colo attention to support custom mask (#5510)
[Inference/Refactor] Refactor compilation mechanism and unified multi hw (#5613)
[install]fix setup (#5786)
fix typo under extensions/ (#5330)
[devops] fix extention building (#5427)