improve logging & fix vllm version

python/pyproject.toml

@@ -20,7 +20,7 @@ dependencies = [
 
 [project.optional-dependencies]
 srt = ["aiohttp", "fastapi", "psutil", "rpyc", "torch", "uvloop", "uvicorn",
-       "zmq", "vllm>=0.4.2", "interegular", "pydantic", "pillow", "packaging", "huggingface_hub", "hf_transfer", "outlines>=0.0.34"]
+       "zmq", "vllm==0.4.2", "interegular", "pydantic", "pillow", "packaging", "huggingface_hub", "hf_transfer", "outlines>=0.0.34"]
 openai = ["openai>=1.0", "numpy", "tiktoken"]
 anthropic = ["anthropic>=0.20.0", "numpy"]
 all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]"]

python/sglang/srt/managers/router/model_runner.py

@@ -235,6 +235,7 @@ class ModelRunner:
         }
 
         # Init torch distributed
+        logger.info(f"[rank={self.tp_rank}] Set cuda device.")
         torch.cuda.set_device(self.tp_rank)
         logger.info(f"[rank={self.tp_rank}] Init torch begin. Avail mem={get_available_gpu_memory(self.tp_rank):.2f} GB")
         torch.distributed.init_process_group(