[Multi Modal] Add an env var for message queue max chunk bytes (#19242)

Signed-off-by: yZhen <yZhen@fb.com>
Co-authored-by: yZhen <yZhen@fb.com>

vllm/v1/executor/multiproc_executor.py

@@ -20,6 +20,7 @@ from typing import Any, Callable, Optional, Union, cast
 
 import cloudpickle
 
+import vllm.envs as envs
 from vllm.config import VllmConfig
 from vllm.distributed import (destroy_distributed_environment,
                               destroy_model_parallel)
@@ -72,7 +73,10 @@ class MultiprocExecutor(Executor):
 
         # Initialize worker and set up message queues for SchedulerOutputs
         # and ModelRunnerOutputs
-        self.rpc_broadcast_mq = MessageQueue(self.world_size, self.world_size)
+        max_chunk_bytes = envs.VLLM_MQ_MAX_CHUNK_BYTES_MB * 1024 * 1024
+        self.rpc_broadcast_mq = MessageQueue(self.world_size,
+                                             self.world_size,
+                                             max_chunk_bytes=max_chunk_bytes)
         scheduler_output_handle = self.rpc_broadcast_mq.export_handle()
 
         # Create workers