[1/N] Elastic EP Milestone 2 (#34861)

Signed-off-by: Yongji Wu <wuyongji317@gmail.com>
Signed-off-by: Itay Alroy <ialroy@nvidia.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Signed-off-by: Ron Tourgeman <rtourgeman@nvidia.com>
Co-authored-by: Yongji Wu <wuyongji317@gmail.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Ron Tourgeman <rtourgeman@nvidia.com>

vllm/v1/executor/multiproc_executor.py

@@ -38,6 +38,7 @@ from vllm.distributed.parallel_state import (
     get_pcp_group,
     get_pp_group,
     get_tp_group,
+    model_parallel_is_initialized,
 )
 from vllm.envs import enable_envs_cache
 from vllm.logger import init_logger
@@ -580,17 +581,20 @@ class WorkerProc:
             )
             self.async_output_copy_thread.start()
 
-        # Initialize device
-        self.worker.init_device()
-
-        # Set process title and log prefix
         self.setup_proc_title_and_log_prefix(
             enable_ep=vllm_config.parallel_config.enable_expert_parallel
         )
 
         # Load model
         self._init_message_queues(input_shm_handle, vllm_config)
-        self.worker.load_model()
+        is_eep_new_worker = envs.VLLM_ELASTIC_EP_SCALE_UP_LAUNCH
+        if not is_eep_new_worker:
+            self.worker.init_device()
+            # Update process title now that parallel groups are initialized
+            self.setup_proc_title_and_log_prefix(
+                enable_ep=vllm_config.parallel_config.enable_expert_parallel
+            )
+            self.worker.load_model()
 
         # Enable environment variable cache (e.g. assume no more
         # environment variable overrides after this point)
@@ -885,6 +889,13 @@ class WorkerProc:
 
     @staticmethod
     def setup_proc_title_and_log_prefix(enable_ep: bool) -> None:
+        # Check if parallel groups are initialized first
+        if not model_parallel_is_initialized():
+            # Parallel groups not yet initialized, use default process name
+            set_process_title(name="Worker")
+            decorate_logs("Worker")
+            return
+
         dp_size = get_dp_group().world_size
         dp_rank = get_dp_group().rank_in_group
         pp_size = get_pp_group().world_size