[Bugfix] Fix make_request_output TypeError on multimodal generation-stage outputs (#4579)

Signed-off-by: QianCyrus <happyloser233@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

tests/engine/test_output_processor.py

@@ -4,6 +4,7 @@
 
 from unittest.mock import MagicMock
 
+import numpy as np
 import pytest
 import torch
 from vllm.outputs import PoolingRequestOutput
@@ -250,6 +251,25 @@ def test_no_detokenizer_make_request_output():
     assert AUDIO in result.outputs[0].multimodal_output
 
 
+def test_no_detokenizer_make_request_output_with_routed_experts():
+    """make_request_output accepts the routed_experts arg that the multimodal
+    output channel (_process_mm_only_outputs) passes positionally, and attaches
+    it to the completion output.
+
+    Regression: the call site passes 6 positional args
+    (..., kv_transfer_params, routed_experts) but the override previously took
+    only 5, raising TypeError on every generation-stage output.
+    """
+    s = _make_no_detok_state(RequestOutputKind.CUMULATIVE)
+    s.add_multimodal_tensor(torch.randn(10), mm_type=AUDIO)
+    routed_experts = np.zeros((2, 3), dtype=np.int32)
+    # Mirror the exact call shape of _process_mm_only_outputs.
+    result = s.make_request_output([], None, FinishReason.STOP, None, None, routed_experts)
+    assert result is not None
+    assert not isinstance(result, PoolingRequestOutput)
+    assert result.outputs[0].routed_experts is routed_experts
+
+
 def test_no_detokenizer_stream_interval_skipped():
     """stream_interval > 1 does not crash when detokenizer is None."""
     kwargs = {**_NO_DETOK_STATE_KWARGS, "stream_interval": 5}

vllm_omni/engine/output_processor.py

@@ -250,6 +250,7 @@ class OmniRequestState(RequestState):
         finish_reason: FinishReason | None,
         stop_reason: int | str | None,
         kv_transfer_params: dict[str, Any] | None = None,
+        routed_experts: Any = None,
     ) -> OmniRequestOutput | PoolingRequestOutput | None:
         """Create a request output from generation results.
 
@@ -263,6 +264,8 @@ class OmniRequestState(RequestState):
             finish_reason: Optional finish reason indicating why generation stopped
             stop_reason: Optional stop reason (token ID or stop string)
             kv_transfer_params: Optional KV cache transfer parameters
+            routed_experts: Optional MoE routed-expert ids for this step,
+                attached to the completion output for generation stages
 
         Returns:
             OmniRequestOutput or PoolingRequestOutput if output should be
@@ -310,7 +313,7 @@ class OmniRequestState(RequestState):
 
         external_req_id = self.external_req_id
 
-        output = self._new_completion_output(new_token_ids, finish_reason, stop_reason)
+        output = self._new_completion_output(new_token_ids, finish_reason, stop_reason, routed_experts)
 
         if self.parent_req is None:
             outputs = [output]
@@ -332,6 +335,7 @@ class OmniRequestState(RequestState):
         token_ids: list[int],
         finish_reason: FinishReason | None,
         stop_reason: int | str | None,
+        routed_experts: Any = None,
     ) -> MultimodalCompletionOutput | CompletionOutput:
         """Create a completion output with multimodal data attached.
 
@@ -352,6 +356,7 @@ class OmniRequestState(RequestState):
                 cumulative_logprob=None,
                 finish_reason=str(finish_reason) if finished else None,
                 stop_reason=stop_reason if finished else None,
+                routed_experts=routed_experts,
             )
         else:
             base_output = super()._new_completion_output(token_ids, finish_reason, stop_reason)