[Refactor] Clean up log once `scope="local"` (#40540)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

vllm/v1/worker/dp_utils.py

@@ -29,7 +29,6 @@ def _get_device_and_group(parallel_config: ParallelConfig):
     if parallel_config.disable_nccl_for_dp_synchronization:
         logger.info_once(
             "Using CPU all reduce to synchronize DP padding between ranks.",
-            scope="local",
         )
         device = "cpu"
         group = get_dp_group().cpu_group

vllm/v1/worker/gpu/eplb_utils.py

@@ -92,9 +92,7 @@ class EPLBController:
         if not is_mixture_of_experts(model):
             return False
 
-        logger.info_once(
-            "EPLB is enabled for model %s.", model_config.model, scope="local"
-        )
+        logger.info_once("EPLB is enabled for model %s.", model_config.model)
         assert self.state is not None
         self.state.add_model(model, model_config)
         self._has_registered_models = True

vllm/v1/worker/gpu_model_runner.py

@@ -4855,7 +4855,6 @@ class GPUModelRunner(
             "Model loading took %s GiB memory and %.6f seconds",
             format_gib(self.model_memory_usage),
             time_after_load - time_before_load,
-            scope="local",
         )
         if not load_dummy_weights:
             prepare_communication_buffer_for_model(self.model)
@@ -4989,7 +4988,7 @@ class GPUModelRunner(
             )
 
         # begin loading weights
-        logger.info_once("Reloading weights inplace...", scope="local")
+        logger.info_once("Reloading weights inplace...")
         if is_checkpoint_format:
             # load weights from checkpoint/ original model format
             initialize_layerwise_reload(model)
@@ -5001,7 +5000,6 @@ class GPUModelRunner(
             logger.warning_once(
                 "Reloading with `is_checkpoint_format=True` requires that "
                 "weights be in kernel format and already sharded",
-                scope="local",
             )
             loaded_weights = set()
             for name, loaded_weight in weights_iterator:
@@ -5015,7 +5013,6 @@ class GPUModelRunner(
         logger.info_once(
             "Reloading and processing weights took %.2f seconds",
             diff_seconds,
-            scope="local",
         )
         if self.model_config.quantization is None and loaded_weights is not None:
             weights_not_loaded = weights_to_load - loaded_weights
@@ -5802,7 +5799,6 @@ class GPUModelRunner(
                             encoder_budget,
                             max_mm_items_per_batch,
                             dummy_modality,
-                            scope="local",
                         )
 
                         # Create dummy batch of multimodal inputs.
@@ -6099,7 +6095,6 @@ class GPUModelRunner(
             "Graph capturing finished in %.0f secs, took %.2f GiB",
             elapsed_time,
             cuda_graph_size / (1 << 30),
-            scope="local",
         )
         return cuda_graph_size
 

vllm/v1/worker/gpu_worker.py

@@ -269,7 +269,7 @@ class Worker(WorkerBase):
             )
 
             if self.use_v2_model_runner:
-                logger.info_once("Using V2 Model Runner", scope="local")
+                logger.info_once("Using V2 Model Runner")
 
             # Set random seed.
             set_random_seed(self.model_config.seed)
@@ -440,7 +440,6 @@ class Worker(WorkerBase):
         logger.info_once(
             "Available KV cache memory: %s GiB",
             format_gib(self.available_kv_cache_memory_bytes),
-            scope="local",
         )
 
         if cudagraph_memory_estimate > 0: