[Bugfix] Fix Qwen3-TTS prefill probe on MRv2 runner (#8065)

Signed-off-by: Sy03 <1370724210@qq.com>

mkdocs.yml

@@ -139,15 +139,16 @@ plugins:
             show_source: false
           inventories:
           - https://docs.python.org/3/objects.inv
-          - https://typing-extensions.readthedocs.io/en/latest/objects.inv
+          # Read the Docs rate-limits these external inventories in strict builds.
+          # - https://typing-extensions.readthedocs.io/en/latest/objects.inv
           # Disabled because Read the Docs strict builds fail if this
           # external inventory intermittently resets the connection.
           # - https://docs.aiohttp.org/en/stable/objects.inv
-          - https://pillow.readthedocs.io/en/stable/objects.inv
+          # - https://pillow.readthedocs.io/en/stable/objects.inv
           - https://numpy.org/doc/stable/objects.inv
           # Temporarily disabled due to decompression errors
           # - https://pytorch.org/docs/stable/objects.inv
-          - https://psutil.readthedocs.io/stable/objects.inv
+          # - https://psutil.readthedocs.io/stable/objects.inv
 
 markdown_extensions:
   - attr_list

tests/e2e/features/prefill_phase/worker_extension.py

@@ -8,11 +8,12 @@ from typing import TYPE_CHECKING
 import torch
 
 if TYPE_CHECKING:
-    from vllm_omni.worker.gpu_model_runner import OmniGPUModelRunner
+    from vllm_omni.worker.gpu_model_runner import OmniGPUModelRunner as LegacyRunner
+    from vllm_omni.worker_v2.omni_ar_model_runner import OmniARModelRunner
 
 
 class PhaseContractWorkerExtension:
-    model_runner: "OmniGPUModelRunner"
+    model_runner: "LegacyRunner | OmniARModelRunner"
 
     def start_phase_contract_probe(self):
         runner = self.model_runner
@@ -20,10 +21,16 @@ class PhaseContractWorkerExtension:
         self.phase_contract_codes: list[list[list[int]]] = []
         tail_embeddings = {}
         original_preprocess = runner.model.preprocess
-        original_mtp = runner._talker_mtp_forward
+        model_state = getattr(runner, "model_state", None)
+        is_v2 = model_state is not None
+        original_mtp = model_state._run_batched_mtp if model_state is not None else runner._talker_mtp_forward
 
         def preprocess(input_ids, input_embeds, **info):
-            rid = info["request_id"]
+            rid = info["req_id"] if is_v2 else info["request_id"]
+            if is_v2:
+                prompt_len = info["_omni_prompt_len"]
+                computed = info["_omni_num_computed_tokens"]
+            else:
                 index = runner.input_batch.req_id_to_index[rid]
                 prompt_len = len(runner.requests[rid].prompt_token_ids)
                 computed = int(runner.input_batch.num_computed_tokens_cpu[index])
@@ -34,11 +41,14 @@ class PhaseContractWorkerExtension:
             span = input_ids.numel()
             self.phase_contract_events.append((prompt_len, computed, span, info["_omni_is_prefill"]))
             if computed == prompt_len - 1 and span == 1:
+                if is_v2:
+                    tail_embeddings[rid] = result[1].detach().clone()
+                else:
                     offset = int(runner.query_start_loc.cpu[index])
                     tail_embeddings[offset] = result[1].detach().clone()
             return result
 
-        def mtp(decode_req_ids, inputs_embeds, start_offsets=None):
+        def legacy_mtp(decode_req_ids, inputs_embeds, start_offsets=None):
             for rid in decode_req_ids:
                 index = runner.input_batch.req_id_to_index[rid]
                 assert int(runner.input_batch.num_computed_tokens_cpu[index]) >= len(
@@ -51,8 +61,38 @@ class PhaseContractWorkerExtension:
             for rid in decode_req_ids:
                 self.phase_contract_codes.append(runner.model_intermediate_buffer[rid]["codes"]["audio"].tolist())
 
+        def v2_mtp(mtp_batches, input_ids, embeds, input_batch, gpu_keys, *args, **kwargs):
+            assert model_state is not None
+            for batch_index, _offset, _mtp_inputs in mtp_batches:
+                req_index = int(input_batch.idx_mapping_np[batch_index])
+                computed = model_state._get_input_batch_num_computed(input_batch, req_index, batch_index)
+                prompt_len = model_state._get_req_state_value(runner.req_states.prompt_len, req_index)
+                assert computed is not None and prompt_len is not None and computed >= prompt_len, (
+                    "prefill row was routed through MTP"
+                )
+            original_mtp(mtp_batches, input_ids, embeds, input_batch, gpu_keys, *args, **kwargs)
+            for batch_index, _offset, _mtp_inputs in mtp_batches:
+                req_index = int(input_batch.idx_mapping_np[batch_index])
+                codes = model_state.intermediate_buffer.buffers[req_index]["codes"]["audio"]
+                self.phase_contract_codes.append(codes.tolist())
+
         runner.model.preprocess = preprocess
-        setattr(runner, "_talker_mtp_forward", mtp)
+        if model_state is not None:
+            original_run_preprocess = model_state.run_preprocess
+
+            def run_preprocess(input_batch, model_inputs, *args, **kwargs):
+                original_run_preprocess(input_batch, model_inputs, *args, **kwargs)
+                embeds = model_inputs["inputs_embeds"]
+                for rid, expected in tail_embeddings.items():
+                    batch_index = input_batch.req_ids.index(rid)
+                    offset = int(input_batch.query_start_loc_np[batch_index])
+                    torch.testing.assert_close(embeds[offset : offset + 1], expected, rtol=0, atol=0)
+                tail_embeddings.clear()
+
+            model_state.run_preprocess = run_preprocess
+            model_state._run_batched_mtp = v2_mtp
+        else:
+            runner._talker_mtp_forward = legacy_mtp
         return True
 
     def get_phase_contract_probe(self):