[HiCache]Fix CP support for hybrid model (#22782)

Co-authored-by: hzh0425 <hzh0425@apache.org>

python/sglang/srt/mem_cache/hybrid_cache/hybrid_cache_controller.py

@@ -162,6 +162,8 @@ class HybridCacheController(BaseHiCacheController):
         storage_backend_extra_config: Optional[dict] = None,
         pp_rank: int = 0,
         pp_size: int = 1,
+        attn_cp_rank: int = 0,
+        attn_cp_size: int = 1,
         transfer_layer_num: Optional[int] = None,
         enable_storage_metrics: bool = False,
     ):
@@ -180,6 +182,8 @@ class HybridCacheController(BaseHiCacheController):
             storage_backend_extra_config=storage_backend_extra_config,
             pp_rank=pp_rank,
             pp_size=pp_size,
+            attn_cp_rank=attn_cp_rank,
+            attn_cp_size=attn_cp_size,
             enable_storage_metrics=enable_storage_metrics,
         )
         # Override layer_num: hybrid models transfer all layers (For example, Linear Model (KV + Mamba)),

python/sglang/srt/mem_cache/hybrid_cache/hybrid_pool_assembler.py

@@ -92,6 +92,8 @@ def build_nsa_hybrid_stack(
             storage_backend_extra_config=extra_config,
             pp_rank=radix_cache.pp_rank,
             pp_size=radix_cache.pp_size,
+            attn_cp_rank=params.attn_cp_rank,
+            attn_cp_size=params.attn_cp_size,
             transfer_layer_num=layer_num,
             enable_storage_metrics=enable_storage_metrics,
         )
@@ -190,6 +192,8 @@ def build_mamba_hybrid_stack(
             storage_backend_extra_config=extra_config,
             pp_rank=params.pp_rank,
             pp_size=params.pp_size,
+            attn_cp_rank=params.attn_cp_rank,
+            attn_cp_size=params.attn_cp_size,
             transfer_layer_num=transfer_layer_num,
             enable_storage_metrics=enable_storage_metrics,
         )