diff --git a/vllm/patches/vllm_for_multi_arc.patch b/vllm/patches/vllm_for_multi_arc.patch
index 9e305f1b..89976a90 100644
--- a/vllm/patches/vllm_for_multi_arc.patch
+++ b/vllm/patches/vllm_for_multi_arc.patch
@@ -11851,16 +11851,17 @@ index 000000000..b01b55f70
 +                if "mlp.experts" in name:
 +                    continue
 +
-+                name = name.replace(weight_name, param_name)
++                name_mapped = name.replace(weight_name, param_name)
 +                # Skip loading extra bias for GPTQ models.
-+                if name.endswith(".bias") and name not in params_dict:
++                if name_mapped.endswith(".bias") and name_mapped not in params_dict:
 +                    continue
 +                # Skip layers on other devices.
 +                if is_pp_missing_parameter(name, self):
 +                    continue
 +                # name = apply_attn_prefix(name, params_dict)
-+                if name not in params_dict:
++                if name_mapped not in params_dict:
 +                    continue
++                name = name_mapped
 +                param = params_dict[name]
 +
 +                # Use weight_loader_v2 for tuple shard_id (GDN projections).
@@ -12604,15 +12605,16 @@ index 000000000..547c1885a
 +                if "mlp.experts" in name:
 +                    continue
 +
-+                name = name.replace(weight_name, param_name)
++                name_mapped = name.replace(weight_name, param_name)
 +                # Skip loading extra bias for GPTQ models.
-+                if name.endswith(".bias") and name not in params_dict:
++                if name_mapped.endswith(".bias") and name_mapped not in params_dict:
 +                    continue
 +                # Skip layers on other devices.
-+                if is_pp_missing_parameter(name, self):
++                if is_pp_missing_parameter(name_mapped, self):
 +                    continue
 +                if name not in params_dict:
 +                    continue
++                name = name_mapped
 +                param = params_dict[name]
 +                weight_loader = param.weight_loader
 +                weight_loader(param, loaded_weight, shard_id)