fix fullgraph

Signed-off-by: MengqingCao <cmq0113@163.com>

Author: MengqingCao

vllm_ascend/worker/model_runner_v1.py

@@ -2298,6 +2298,12 @@ def _generate_dummy_run_hidden_states(self, with_prefill,
                                    positions=positions,
                                    intermediate_tensors=intermediate_tensors,
                                    inputs_embeds=inputs_embeds)
+        forward_context = get_forward_context()
+        assert forward_context is not None
+        if forward_context.cudagraph_runtime_mode == CUDAGraphMode.FULL:
+            update_attn_params(self.update_stream, forward_context,
+                               positions.shape[0])
+
         if self.drafter and self.drafter.name == SpecDcodeType.EAGLE3:
             hidden_states, _ = hidden_states
         else:

vllm_ascend/worker/worker_v1.py

@@ -333,7 +333,9 @@ def pin_lora(self, lora_id: int) -> bool:
         return self.model_runner.pin_lora(lora_id)
 
     def execute_dummy_batch(self) -> None:
-        self.model_runner._dummy_run(1, uniform_decode=True)
+        self.model_runner._dummy_run(1,
+                                     uniform_decode=True,
+                                     force_attention=True)
 
     def _init_worker_distributed_environment(self) -> None:
         """Initialize the distributed environment."""