[BugFix] Fix pipeline parallel (#24621)

Signed-off-by: Nick Hill <nhill@redhat.com>
2025-09-10 23:05:30 -07:00
parent 29799ddacc
commit e2d8c27f68
3 changed files with 6 additions and 2 deletions
--- a/vllm/v1/worker/gpu_model_runner.py
+++ b/vllm/v1/worker/gpu_model_runner.py
@@ -2070,7 +2070,6 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
            sampler_output = self._sample(logits, spec_decode_metadata)

        with record_function_or_nullcontext("Bookkeep"):
-            assert isinstance(hidden_states, torch.Tensor)
            (
                num_nans_in_logits,
                logprobs_lists,