[Observability] add model initializing duration log

vllm/model_executor/model_loader/base_loader.py

@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import time
 from abc import ABC, abstractmethod
 
 import torch
@@ -70,11 +71,17 @@ def load_model(
         target_device = torch.device(load_device)
         with set_default_torch_dtype(model_config.dtype):
             with target_device:
+                time_before_load = time.perf_counter()
                 model = self.create_model(
                     vllm_config=vllm_config,
                     model_config=model_config,
                     prefix=prefix,
                 )
+                time_after_load = time.perf_counter()
+                logger.info_once(
+                    "Initializing model took %.6f seconds",
+                    time_after_load - time_before_load,
+                )
 
             logger.debug("Loading weights on %s ...", load_device)
             self.load_weights(model, model_config)