[V1] log GPU blocks num for MultiprocExecutor (vllm-project#11656)

Signed-off-by: Bowen Wang <[email protected]>

vllm/v1/executor/multiproc_executor.py

@@ -95,6 +95,7 @@ def initialize(self, num_gpu_blocks: int) -> None:
         Initialize the KV caches and begin the model execution loop of the
         underlying workers.
         """
+        logger.info("# GPU blocks: %d", num_gpu_blocks)
         self.collective_rpc("initialize_cache", args=(num_gpu_blocks, ))
         self.collective_rpc("compile_or_warm_up_model")