[Core] Add MultiprocessingGPUExecutor (#4539)

Co-authored-by: SAHIL SUNEJA <suneja@us.ibm.com>

.buildkite/test-pipeline.yaml

@@ -34,10 +34,14 @@ steps:
   mirror_hardwares: [amd]
   commands:
   - pytest -v -s distributed/test_pynccl_library.py
-  - TEST_DIST_MODEL=facebook/opt-125m pytest -v -s distributed/test_basic_distributed_correctness.py
-  - TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf pytest -v -s distributed/test_basic_distributed_correctness.py
-  - TEST_DIST_MODEL=facebook/opt-125m pytest -v -s distributed/test_chunked_prefill_distributed.py
-  - TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf pytest -v -s distributed/test_chunked_prefill_distributed.py
+  - TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_basic_distributed_correctness.py
+  - TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_basic_distributed_correctness.py
+  - TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_chunked_prefill_distributed.py
+  - TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_chunked_prefill_distributed.py
+  - TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_basic_distributed_correctness.py
+  - TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_basic_distributed_correctness.py
+  - TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_chunked_prefill_distributed.py
+  - TEST_DIST_MODEL=meta-llama/Llama-2-7b-hf DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_chunked_prefill_distributed.py
 
 - label: Distributed Tests (Multiple Groups)
   working_dir: "/vllm-workspace/tests"

tests/distributed/test_basic_distributed_correctness.py

@@ -20,6 +20,7 @@ import torch
 MODELS = [
     os.environ["TEST_DIST_MODEL"],
 ]
+DISTRIBUTED_EXECUTOR_BACKEND = "DISTRIBUTED_EXECUTOR_BACKEND"
 VLLM_ATTENTION_BACKEND = "VLLM_ATTENTION_BACKEND"
 
 
@@ -36,19 +37,21 @@ def test_models(
     dtype: str,
     max_tokens: int,
 ) -> None:
-    enforce_eager = False
+    distributed_executor_backend = os.getenv(DISTRIBUTED_EXECUTOR_BACKEND)
+
     backend_by_env_var = os.getenv(VLLM_ATTENTION_BACKEND)
-    if backend_by_env_var == "FLASHINFER":
-        enforce_eager = True
+    enforce_eager = backend_by_env_var == "FLASHINFER"
 
     hf_model = hf_runner(model, dtype=dtype)
     hf_outputs = hf_model.generate_greedy(example_prompts, max_tokens)
     del hf_model
 
-    vllm_model = vllm_runner(model,
+    vllm_model = vllm_runner(
+        model,
         dtype=dtype,
         tensor_parallel_size=2,
-                             enforce_eager=enforce_eager)
+        enforce_eager=enforce_eager,
+        distributed_executor_backend=distributed_executor_backend)
     vllm_outputs = vllm_model.generate_greedy(example_prompts, max_tokens)
     del vllm_model
 

tests/distributed/test_chunked_prefill_distributed.py

@@ -19,6 +19,7 @@ import torch
 MODELS = [
     os.environ["TEST_DIST_MODEL"],
 ]
+DISTRIBUTED_EXECUTOR_BACKEND = "DISTRIBUTED_EXECUTOR_BACKEND"
 
 
 @pytest.mark.skipif(torch.cuda.device_count() < 2,
@@ -36,6 +37,8 @@ def test_models(
     max_tokens: int,
     chunked_prefill_token_size: int,
 ) -> None:
+    distributed_executor_backend = os.getenv(DISTRIBUTED_EXECUTOR_BACKEND)
+
     # Add a chunked prefill config.
     max_num_seqs = min(chunked_prefill_token_size, 256)
     assert chunked_prefill_token_size != -1
@@ -53,6 +56,7 @@ def test_models(
         max_num_seqs=max_num_seqs,
         enable_chunked_prefill=enable_chunked_prefill,
         max_num_batched_tokens=max_num_batched_tokens,
+        distributed_executor_backend=distributed_executor_backend,
     )
     vllm_outputs = vllm_model.generate_greedy(example_prompts, max_tokens)
     del vllm_model

tests/lora/test_mixtral.py

@@ -38,8 +38,7 @@ def test_mixtral_lora(mixtral_lora_files, tp_size):
                    enable_lora=True,
                    max_num_seqs=16,
                    max_loras=4,
-                   tensor_parallel_size=tp_size,
-                   worker_use_ray=True)
+                   tensor_parallel_size=tp_size)
 
     expected_lora_output = [
         "give_opinion(name[SpellForce 3], release_year[2017], developer[Grimlore Games], rating[poor])",  # noqa: E501

vllm/config.py

@@ -521,9 +521,7 @@ class ParallelConfig:
     Args:
         pipeline_parallel_size: Number of pipeline parallel groups.
         tensor_parallel_size: Number of tensor parallel groups.
-        worker_use_ray: Whether to use Ray for model workers. Will be set to
-            True if either pipeline_parallel_size or tensor_parallel_size is
-            greater than 1.
+        worker_use_ray: Deprecated, use distributed_executor_backend instead.
         max_parallel_loading_workers: Maximum number of multiple batches
             when load model sequentially. To avoid RAM OOM when using tensor
             parallel and large models.
@@ -533,22 +531,27 @@ class ParallelConfig:
             If None, will use synchronous tokenization.
         ray_workers_use_nsight: Whether to profile Ray workers with nsight, see
             https://docs.ray.io/en/latest/ray-observability/user-guides/profiling.html#profiling-nsight-profiler.
+        distributed_executor_backend: Backend to use for distributed model
+            workers, either "ray" or "mp" (multiprocessing). If either
+            pipeline_parallel_size or tensor_parallel_size is greater than 1,
+            will default to "ray" if Ray is installed or "mp" otherwise.
     """
 
     def __init__(
         self,
         pipeline_parallel_size: int,
         tensor_parallel_size: int,
-        worker_use_ray: bool,
+        worker_use_ray: Optional[bool] = None,
         max_parallel_loading_workers: Optional[int] = None,
         disable_custom_all_reduce: bool = False,
         tokenizer_pool_config: Optional[TokenizerPoolConfig] = None,
         ray_workers_use_nsight: bool = False,
         placement_group: Optional["PlacementGroup"] = None,
+        distributed_executor_backend: Optional[str] = None,
     ) -> None:
         self.pipeline_parallel_size = pipeline_parallel_size
         self.tensor_parallel_size = tensor_parallel_size
-        self.worker_use_ray = worker_use_ray
+        self.distributed_executor_backend = distributed_executor_backend
         self.max_parallel_loading_workers = max_parallel_loading_workers
         self.disable_custom_all_reduce = disable_custom_all_reduce
         self.tokenizer_pool_config = tokenizer_pool_config
@@ -556,14 +559,29 @@ class ParallelConfig:
         self.placement_group = placement_group
 
         self.world_size = pipeline_parallel_size * self.tensor_parallel_size
-        if self.world_size > 1:
-            self.worker_use_ray = True
+        if worker_use_ray:
+            if self.distributed_executor_backend is None:
+                self.distributed_executor_backend = "ray"
+            elif self.distributed_executor_backend != "ray":
+                raise ValueError(f"worker-use-ray can't be used with "
+                                 f"distributed executor backend "
+                                 f"'{self.distributed_executor_backend}'.")
+
+        if self.distributed_executor_backend is None and self.world_size > 1:
+            from vllm.executor import ray_utils
+            ray_found = ray_utils.ray is not None
+            self.distributed_executor_backend = "ray" if ray_found else "mp"
+
         self._verify_args()
 
     def _verify_args(self) -> None:
         if self.pipeline_parallel_size > 1:
             raise NotImplementedError(
                 "Pipeline parallelism is not supported yet.")
+        if self.distributed_executor_backend not in ("ray", "mp", None):
+            raise ValueError(
+                "Unrecognized distributed executor backend. Supported values "
+                "are 'ray' or 'mp'.")
         if not self.disable_custom_all_reduce and self.world_size > 1:
             if is_hip():
                 self.disable_custom_all_reduce = True
@@ -575,7 +593,8 @@ class ParallelConfig:
                 logger.info(
                     "Disabled the custom all-reduce kernel because it is not "
                     "supported with pipeline parallelism.")
-        if self.ray_workers_use_nsight and not self.worker_use_ray:
+        if self.ray_workers_use_nsight and (
+                not self.distributed_executor_backend == "ray"):
             raise ValueError("Unable to use nsight profiling unless workers "
                              "run with Ray.")
 
@@ -887,7 +906,8 @@ class SpeculativeConfig:
             pipeline_parallel_size=target_parallel_config.
             pipeline_parallel_size,
             tensor_parallel_size=target_parallel_config.tensor_parallel_size,
-            worker_use_ray=target_parallel_config.worker_use_ray,
+            distributed_executor_backend=target_parallel_config.
+            distributed_executor_backend,
             max_parallel_loading_workers=target_parallel_config.
             max_parallel_loading_workers,
             disable_custom_all_reduce=target_parallel_config.

vllm/engine/arg_utils.py

@@ -34,6 +34,7 @@ class EngineArgs:
     seed: int = 0
     max_model_len: Optional[int] = None
     worker_use_ray: bool = False
+    distributed_executor_backend: Optional[str] = None
     pipeline_parallel_size: int = 1
     tensor_parallel_size: int = 1
     max_parallel_loading_workers: Optional[int] = None
@@ -221,10 +222,17 @@ class EngineArgs:
             ' Can be overridden per request via guided_decoding_backend'
             ' parameter.')
         # Parallel arguments
-        parser.add_argument('--worker-use-ray',
+        parser.add_argument(
+            '--distributed-executor-backend',
+            choices=['ray', 'mp'],
+            default=EngineArgs.distributed_executor_backend,
+            help='Backend to use for distributed serving. When more than 1 GPU '
+            'is used, will be automatically set to "ray" if installed '
+            'or "mp" (multiprocessing) otherwise.')
+        parser.add_argument(
+            '--worker-use-ray',
             action='store_true',
-                            help='Use Ray for distributed serving, will be '
-                            'automatically set when using more than 1 GPU.')
+            help='Deprecated, use --distributed-executor-backend=ray.')
         parser.add_argument('--pipeline-parallel-size',
                             '-pp',
                             type=int,

vllm/engine/async_llm_engine.py

@@ -348,27 +348,31 @@ class AsyncLLMEngine:
         """Creates an async LLM engine from the engine arguments."""
         # Create the engine configs.
         engine_config = engine_args.create_engine_config()
+        distributed_executor_backend = (
+            engine_config.parallel_config.distributed_executor_backend)
 
         if engine_config.device_config.device_type == "neuron":
             from vllm.executor.neuron_executor import NeuronExecutorAsync
             executor_class = NeuronExecutorAsync
         elif engine_config.device_config.device_type == "cpu":
-            assert not engine_config.parallel_config.worker_use_ray, (
-                "Ray is not supported with the CPU backend.")
+            assert distributed_executor_backend is None, (
+                "Distributed execution is not supported with the CPU backend.")
             from vllm.executor.cpu_executor import CPUExecutorAsync
             executor_class = CPUExecutorAsync
-        elif engine_config.parallel_config.worker_use_ray:
+        elif distributed_executor_backend == "ray":
             initialize_ray_cluster(engine_config.parallel_config)
             from vllm.executor.ray_gpu_executor import RayGPUExecutorAsync
             executor_class = RayGPUExecutorAsync
+        elif distributed_executor_backend == "mp":
+            from vllm.executor.multiproc_gpu_executor import (
+                MultiprocessingGPUExecutorAsync)
+            executor_class = MultiprocessingGPUExecutorAsync
         else:
-            assert engine_config.parallel_config.world_size == 1, (
-                "Ray is required if parallel_config.world_size > 1.")
             from vllm.executor.gpu_executor import GPUExecutorAsync
             executor_class = GPUExecutorAsync
         # Create the async LLM engine.
         engine = cls(
-            engine_config.parallel_config.worker_use_ray,
+            distributed_executor_backend == "ray",
             engine_args.engine_use_ray,
             **engine_config.to_dict(),
             executor_class=executor_class,

vllm/engine/llm_engine.py

@@ -274,6 +274,8 @@ class LLMEngine:
         """Creates an LLM engine from the engine arguments."""
         # Create the engine configs.
         engine_config = engine_args.create_engine_config()
+        distributed_executor_backend = (
+            engine_config.parallel_config.distributed_executor_backend)
 
         # Initialize the cluster and specify the executor class.
         if engine_config.device_config.device_type == "neuron":
@@ -282,13 +284,15 @@ class LLMEngine:
         elif engine_config.device_config.device_type == "cpu":
             from vllm.executor.cpu_executor import CPUExecutor
             executor_class = CPUExecutor
-        elif engine_config.parallel_config.worker_use_ray:
+        elif distributed_executor_backend == "ray":
             initialize_ray_cluster(engine_config.parallel_config)
             from vllm.executor.ray_gpu_executor import RayGPUExecutor
             executor_class = RayGPUExecutor
+        elif distributed_executor_backend == "mp":
+            from vllm.executor.multiproc_gpu_executor import (
+                MultiprocessingGPUExecutor)
+            executor_class = MultiprocessingGPUExecutor
         else:
-            assert engine_config.parallel_config.world_size == 1, (
-                "Ray is required if parallel_config.world_size > 1.")
             from vllm.executor.gpu_executor import GPUExecutor
             executor_class = GPUExecutor
 

vllm/executor/multiproc_gpu_executor.py

+import asyncio
+import os
+from functools import partial
+from typing import Any, Dict, Optional, Tuple
+
+from vllm.executor.distributed_gpu_executor import (  # yapf: disable
+    DistributedGPUExecutor, DistributedGPUExecutorAsync)
+from vllm.executor.multiproc_worker_utils import (ProcessWorkerWrapper,
+                                                  ResultHandler, WorkerMonitor)
+from vllm.logger import init_logger
+from vllm.utils import (get_distributed_init_method, get_ip, get_open_port,
+                        get_vllm_instance_id, make_async)
+
+logger = init_logger(__name__)
+
+
+class MultiprocessingGPUExecutor(DistributedGPUExecutor):
+    """Python multiprocessing-based multi-GPU executor"""
+
+    def _init_executor(self) -> None:
+        assert (
+            not self.speculative_config
+        ), "Speculative decoding not yet supported for MultiProcGPU backend."
+
+        # Create the parallel GPU workers.
+        world_size = self.parallel_config.tensor_parallel_size
+
+        # Set CUDA_VISIBLE_DEVICES for the driver, inherited by workers
+        if "CUDA_VISIBLE_DEVICES" not in os.environ:
+            os.environ["CUDA_VISIBLE_DEVICES"] = (",".join(
+                map(str, range(world_size))))
+
+        # Ensure that VLLM_INSTANCE_ID is set, to be inherited by workers
+        os.environ["VLLM_INSTANCE_ID"] = get_vllm_instance_id()
+
+        from torch.cuda import device_count
+        assert world_size <= device_count(), (
+            "please set tensor_parallel_size to less than max local gpu count")
+
+        distributed_init_method = get_distributed_init_method(
+            get_ip(), get_open_port())
+
+        if world_size == 1:
+            self.workers = []
+        else:
+            result_handler = ResultHandler()
+            self.workers = [
+                ProcessWorkerWrapper(
+                    result_handler,
+                    partial(
+                        self._create_worker,
+                        rank=rank,
+                        local_rank=rank,
+                        distributed_init_method=distributed_init_method,
+                    )) for rank in range(1, world_size)
+            ]
+
+            self.worker_monitor = WorkerMonitor(self.workers, result_handler)
+            result_handler.start()
+            self.worker_monitor.start()
+
+        self.driver_worker = self._create_worker(
+            distributed_init_method=distributed_init_method)
+        self._run_workers("init_device")
+        self._run_workers("load_model",
+                          max_concurrent_workers=self.parallel_config.
+                          max_parallel_loading_workers)
+
+    def shutdown(self):
+        if (worker_monitor := getattr(self, "worker_monitor",
+                                      None)) is not None:
+            worker_monitor.close()
+
+    def _run_workers(
+        self,
+        method: str,
+        *args,
+        driver_args: Optional[Tuple[Any, ...]] = None,
+        driver_kwargs: Optional[Dict[str, Any]] = None,
+        max_concurrent_workers: Optional[int] = None,
+        **kwargs,
+    ) -> Any:
+        """Runs the given method on all workers."""
+
+        if max_concurrent_workers:
+            raise NotImplementedError(
+                "max_concurrent_workers is not supported yet.")
+
+        # Start the workers first.
+        worker_outputs = [
+            worker.execute_method(method, *args, **kwargs)
+            for worker in self.workers
+        ]
+
+        if driver_args is None:
+            driver_args = args
+        if driver_kwargs is None:
+            driver_kwargs = kwargs
+
+        # Start the driver worker after all the ray workers.
+        driver_worker_method = getattr(self.driver_worker, method)
+        driver_worker_output = driver_worker_method(*driver_args,
+                                                    **driver_kwargs)
+
+        # Get the results of the workers.
+        return [driver_worker_output
+                ] + [output.get() for output in worker_outputs]
+
+    def check_health(self) -> None:
+        """Raises an error if engine is unhealthy."""
+        if not self.worker_monitor.is_alive():
+            raise RuntimeError("Worker processes are not running")
+
+
+class MultiprocessingGPUExecutorAsync(MultiprocessingGPUExecutor,
+                                      DistributedGPUExecutorAsync):
+
+    async def _run_workers_async(
+        self,
+        method: str,
+        *args,
+        driver_args: Optional[Tuple[Any, ...]] = None,
+        driver_kwargs: Optional[Dict[str, Any]] = None,
+        **kwargs,
+    ) -> Any:
+        """Runs the given method on all workers."""
+        if driver_args is None:
+            driver_args = args
+        if driver_kwargs is None:
+            driver_kwargs = kwargs
+
+        driver_executor = make_async(getattr(self.driver_worker, method))
+
+        # Run all the workers asynchronously.
+        coros = [driver_executor(*driver_args, **driver_kwargs)] + [
+            worker.execute_method_async(method, *args, **kwargs)
+            for worker in self.workers
+        ]
+
+        return await asyncio.gather(*coros)

vllm/executor/ray_gpu_executor.py

@@ -31,7 +31,7 @@ class RayGPUExecutor(DistributedGPUExecutor):
         assert (not self.speculative_config
                 ), "Speculative decoding not yet supported for RayGPU backend."
 
-        assert self.parallel_config.worker_use_ray
+        assert self.parallel_config.distributed_executor_backend == "ray"
         placement_group = self.parallel_config.placement_group
 
         # Disable Ray usage stats collection.
@@ -264,7 +264,7 @@ class RayGPUExecutor(DistributedGPUExecutor):
                              f"required, but found {current_version}")
 
         from ray.dag import InputNode, MultiOutputNode
-        assert self.parallel_config.worker_use_ray
+        assert self.parallel_config.distributed_executor_backend == "ray"
 
         # Right now, compiled DAG requires at least 1 arg. We send
         # a dummy value for now. It will be fixed soon.

vllm/executor/ray_utils.py

@@ -44,7 +44,7 @@ try:
 
 except ImportError as e:
     logger.warning(
-        "Failed to import Ray with %r. For distributed inference, "
+        "Failed to import Ray with %r. For multi-node inference, "
         "please install Ray with `pip install ray`.", e)
     ray = None  # type: ignore
     RayWorkerWrapper = None  # type: ignore
@@ -67,7 +67,7 @@ def initialize_ray_cluster(
     """
     if ray is None:
         raise ImportError(
-            "Ray is not installed. Please install Ray to use distributed "
+            "Ray is not installed. Please install Ray to use multi-node "
             "serving.")
 
     # Connect to a ray cluster.