[V1] Scheduler Refactoring [1/N] - Add Scheduler Interface (#15250)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Cody Yu <hao.yu.cody@gmail.com>
Co-authored-by: Nick Hill <nhill@redhat.com>

vllm/v1/worker/gpu_model_runner.py

@@ -45,7 +45,7 @@ from vllm.v1.worker.lora_model_runner_mixin import LoRAModelRunnerMixin
 if TYPE_CHECKING:
     import xgrammar as xgr
 
-    from vllm.v1.core.scheduler_output import SchedulerOutput
+    from vllm.v1.core.sched.output import SchedulerOutput
 else:
     xgr = LazyLoader("xgr", globals(), "xgrammar")
 

vllm/v1/worker/gpu_worker.py

@@ -28,7 +28,7 @@ from vllm.v1.worker.worker_base import WorkerBase
 logger = init_logger(__name__)
 
 if TYPE_CHECKING:
-    from vllm.v1.core.scheduler_output import SchedulerOutput
+    from vllm.v1.core.sched.output import SchedulerOutput
 
 
 class Worker(WorkerBase):

vllm/v1/worker/tpu_model_runner.py

@@ -37,7 +37,7 @@ from vllm.v1.utils import bind_kv_cache
 from vllm.v1.worker.gpu_input_batch import CachedRequestState, InputBatch
 
 if TYPE_CHECKING:
-    from vllm.v1.core.scheduler import SchedulerOutput
+    from vllm.v1.core.sched.output import SchedulerOutput
 
 logger = init_logger(__name__)
 

vllm/v1/worker/tpu_worker.py

@@ -17,7 +17,7 @@ from vllm.distributed import (ensure_model_parallel_initialized,
 from vllm.logger import init_logger
 from vllm.model_executor import set_random_seed
 from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE
-from vllm.v1.core.scheduler import SchedulerOutput
+from vllm.v1.core.sched.output import SchedulerOutput
 from vllm.v1.kv_cache_interface import (FullAttentionSpec, KVCacheConfig,
                                         KVCacheSpec)
 from vllm.v1.outputs import ModelRunnerOutput