[None][fix] Add missing allow_partial_loading param to CuteDSL and ConfigurableMoE load_weights (#12761)

Signed-off-by: Xianjie <5410381+qiaoxj07@users.noreply.github.com>

Author: qiaoxj07

tensorrt_llm/_torch/modules/fused_moe/configurable_moe.py

@@ -1237,15 +1237,15 @@ def create_weights(self):
         )
         return self.backend.create_weights()
 
-    def load_weights(self, weights: List[Dict]):
+    def load_weights(self, weights: List[Dict], allow_partial_loading: bool = False):
         """
         Load weights - delegated to backend
 
         """
         assert hasattr(self.backend, "load_weights"), (
             f"Backend {self.backend.__class__.__name__} must implement load_weights()"
         )
-        return self.backend.load_weights(weights)
+        return self.backend.load_weights(weights, allow_partial_loading)
 
     def post_load_weights(self):
         """

tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py

@@ -979,8 +979,10 @@ def forward_chunk(
                          enable_alltoall=False)
         return x
 
-    def load_weights(self, weights: Dict[str, torch.Tensor]):
-        super().load_weights(weights)
+    def load_weights(self,
+                     weights: List[Dict],
+                     allow_partial_loading: bool = False):
+        super().load_weights(weights, allow_partial_loading)
         dwdp_handle_collector = getattr(self, "dwdp_handle_collector", None)
         if dwdp_handle_collector is not None:
             dwdp_handle_collector.register_weights(self)