defcp_aware_sdpa(q:Any,k:Any,v:Any,**kw:Any)->Any:"""CP-aware SDPA replacement: all-gather K/V plus the D-04 mask."""fired["hit"]=Truereturn_cp_sdpa_call(orig_sdpa,cp_mesh,q,k,v,kw)F.scaled_dot_product_attention=cp_aware_sdpatry:out=original_forward(hidden_states,*args,**kwargs)