more tests passing

Signed-off-by: Lucas Wilkinson <[email protected]>

Author: LucasWilkinson

tests/v1/core/test_scheduler.py

@@ -818,9 +818,6 @@ def test_schedule_spec_decoding_stats(spec_tokens, output_tokens, expected):
     scheduler_stats = engine_core_outputs[0].scheduler_stats \
          if engine_core_outputs else None
 
-
-    print("engine_core_outputs", engine_core_outputs)
-
     if expected[0] == 0:
         assert scheduler_stats.spec_decoding_stats is None
     else:

vllm/v1/core/sched/scheduler.py

@@ -668,8 +668,7 @@ def _try_schedule_encoder_inputs(
                 num_new_tokens = start_pos - num_computed_tokens
                 break
 
-            if (not self.encoder_cache_manager.can_allocate(request, i)
-                    or num_encoder_tokens > encoder_budget):
+            if num_encoder_tokens > encoder_budget:
                 # The encoder cache is full or the encoder budget is exhausted.
                 # NOTE(woosuk): We assume that the encoder input tokens should
                 # be processed altogether, as the encoder usually uses

vllm/v1/request.py

@@ -137,15 +137,15 @@ def __getattribute__(self, name: str) -> Any:
         return object.__getattribute__(self, name)
 
     def append_output_token_ids(
-            self,
-            token_ids: Union[int, list[int]],
-        ) -> None:
-            if isinstance(token_ids, int):
-                self._output_token_ids.append(token_ids)
-                self._all_token_ids.append(token_ids)
-            else:
-                self._output_token_ids.extend(token_ids)
-                self._all_token_ids.extend(token_ids)
+        self,
+        token_ids: Union[int, list[int]],
+    ) -> None:
+        if isinstance(token_ids, int):
+            self._output_token_ids.append(token_ids)
+            self._all_token_ids.append(token_ids)
+        else:
+            self._output_token_ids.extend(token_ids)
+            self._all_token_ids.extend(token_ids)
 
     @property
     def num_tokens(self) -> int: