enforce v2 block manager and check logit processor

vllm/engine/arg_utils.py

@@ -871,6 +871,9 @@ def create_engine_config(self, ) -> EngineConfig:
             if self.enable_chunked_prefill:
                 raise ValueError("Chunked prefill is not supported with "
                                  "multi-step (--num-scheduler-steps > 1)")
+            if not self.use_v2_block_manager:
+                raise ValueError("BlockSpaceManagerV2 is required for "
+                                 "multi-step (--num-scheduler-steps > 1)")
 
         # make sure num_lookahead_slots is set the higher value depending on
         # if we are using speculative decoding or multi-step

vllm/worker/multi_step_model_runner.py

@@ -497,8 +497,9 @@ def _pythonize_sampler_output(
         next_token_ids = sample_result
         parent_ids = [0]
         seq_outputs: List[SequenceOutput] = []
-        assert len(seq_group.sampling_params.logits_processors) == 0, (
-            "Logits Processors are not supported in multi-step decoding")
+        if seq_group.sampling_params.logits_processors:
+            assert len(seq_group.sampling_params.logits_processors) == 0, (
+                "Logits Processors are not supported in multi-step decoding")
         for parent_id, next_token_id in zip(parent_ids, next_token_ids):
             # TODO(will): support logprobs
             # Hard coded logprob