[Model Runner V2] Support pooling models (#35120)

Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
2026-02-27 18:03:01 -08:00
parent 405f28d38d
commit 86ac7bcf84
6 changed files with 209 additions and 14 deletions
--- a/vllm/v1/worker/gpu_worker.py
+++ b/vllm/v1/worker/gpu_worker.py
@@ -700,6 +700,12 @@ class Worker(WorkerBase):
            output = self.model_runner.execute_model(
                scheduler_output, intermediate_tensors
            )
+            if (
+                self.use_v2_model_runner
+                and self.model_runner.is_pooling_model
+                and output is None
+            ):
+                output = self.model_runner.pool()  # type: ignore
            if isinstance(
                output, ModelRunnerOutput | AsyncModelRunnerOutput | NoneType
            ):