[Bugfix] Fix in_profile_run in mtp_proposer dummy_run (#5165)

### What this PR does / why we need it?
This PR aims to fix failure of `enable_force_load_balance` caused by
missing `in_profile_run` in `dummy_run` of mtp_proposer.

### Does this PR introduce _any_ user-facing change?
N/A

### How was this patch tested?
by ci

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: Zetong Li <slippersss@126.com>

vllm_ascend/spec_decode/eagle_proposer.py

@@ -143,7 +143,8 @@ class EagleProposer(Proposer):
                   num_tokens_across_dp: Optional[torch.Tensor] = None,
                   aclgraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
                   batch_descriptor=None,
-                  dummy_compute_logits=lambda hidden_states: None):
+                  dummy_compute_logits=lambda hidden_states: None,
+                  is_profile=False):
         # update global cos, sin
         update_cos_sin(self.positions[:num_tokens])
 

vllm_ascend/spec_decode/mtp_proposer.py

@@ -229,7 +229,8 @@ class MtpProposer(Proposer):
                   num_tokens_across_dp=None,
                   aclgraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
                   batch_descriptor=None,
-                  dummy_compute_logits=lambda hidden_states: None) -> None:
+                  dummy_compute_logits=lambda hidden_states: None,
+                  is_profile=False) -> None:
 
         (
             num_tokens,
@@ -299,7 +300,8 @@ class MtpProposer(Proposer):
                     num_actual_tokens=0,
                     aclgraph_runtime_mode=aclgraph_runtime_mode,
                     batch_descriptor=batch_descriptor,
-                    is_mtp_model=True):
+                    is_mtp_model=True,
+                    in_profile_run=is_profile):
                 if self.enable_shared_expert_dp:
                     positions = positions.unsqueeze(-1)
                     positions = torch.ops.vllm.maybe_pad_and_reduce(positions)

vllm_ascend/spec_decode/ngram_proposer.py

@@ -27,7 +27,8 @@ class NgramProposer(VllmNgramProposer, Proposer):
                   num_tokens_across_dp=None,
                   aclgraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
                   batch_descriptor=None,
-                  dummy_compute_logits=lambda hidden_states: None):
+                  dummy_compute_logits=lambda hidden_states: None,
+                  is_profile=False):
         pass
 
     def generate_token_ids(self,

vllm_ascend/spec_decode/suffix_proposer.py

@@ -27,7 +27,8 @@ class SuffixDecodingProposer(VllmSuffixDecodingProposer, Proposer):
                   num_tokens_across_dp=None,
                   aclgraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
                   batch_descriptor=None,
-                  dummy_compute_logits=lambda hidden_states: None):
+                  dummy_compute_logits=lambda hidden_states: None,
+                  is_profile=False):
         pass
 
     def generate_token_ids(self,

vllm_ascend/worker/model_runner_v1.py

@@ -2164,7 +2164,8 @@ class NPUModelRunner(GPUModelRunner):
                     aclgraph_runtime_mode=aclgraph_runtime_mode,
                     batch_descriptor=batch_descriptor,
                     dummy_compute_logits=dummy_drafter_compute_logits,
-                    in_graph_capturing=not force_attention)
+                    in_graph_capturing=not force_attention,
+                    is_profile=is_profile)
             if is_profile and self.dynamic_eplb:
                 self.model.clear_all_moe_loads()
             if not is_profile and self.dynamic_eplb: