bugfix for the order of dummy run pad and sync

Author: LiuYi-Up

vllm_ascend/worker/model_runner_v1.py

@@ -2062,9 +2062,13 @@ def _dummy_run(
         if self.is_kv_producer and not self.is_kv_consumer:
             with_prefill = True
 
+        has_lora = True if self.lora_config and self.compilation_config.cudagraph_specialize_lora else False
+        _ag_mode, batch_descriptor = \
+            self.cudagraph_dispatcher.dispatch(num_tokens=num_tokens, uniform_decode=uniform_decode, has_lora=has_lora)
+
         # Padding for DP
         (num_tokens, num_tokens_across_dp,
-         with_prefill) = self._sync_metadata_across_dp(num_tokens,
+         with_prefill) = self._sync_metadata_across_dp(batch_descriptor.num_tokens,
                                                        with_prefill)
 
         # If cudagraph_mode.decode_mode() == FULL and
@@ -2112,9 +2116,9 @@ def _dummy_run(
         if not is_profile and self.dynamic_eplb:
             self.eplb_updator.forward_before()
 
-        has_lora = True if self.lora_config and self.compilation_config.cudagraph_specialize_lora else False
-        _ag_mode, batch_descriptor = \
-            self.cudagraph_dispatcher.dispatch(num_tokens=num_tokens, uniform_decode=uniform_decode, has_lora=has_lora)
+        if num_tokens != batch_descriptor.num_tokens:
+            _ag_mode, batch_descriptor = self.cudagraph_dispatcher.dispatch(
+                num_tokens=num_tokens, uniform_decode=uniform_decode, has_lora=has_lora)
 
         num_tokens_padded = batch_descriptor.num_tokens
         num_reqs_padded = (batch_descriptor.num_reqs if