[Misc] Add vLLM version getter to utils (#5098)

setup.py

@@ -314,7 +314,7 @@ def find_version(filepath: str) -> str:
 
 
 def get_vllm_version() -> str:
-    version = find_version(get_path("vllm", "__init__.py"))
+    version = find_version(get_path("vllm", "version.py"))
 
     if _is_cuda():
         cuda_version = str(get_nvcc_cuda_version())

vllm/__init__.py

@@ -12,9 +12,10 @@ from vllm.outputs import (CompletionOutput, EmbeddingOutput,
 from vllm.pooling_params import PoolingParams
 from vllm.sampling_params import SamplingParams
 
-__version__ = "0.5.0"
+from .version import __version__
 
 __all__ = [
+    "__version__",
     "LLM",
     "ModelRegistry",
     "PromptStrictInputs",

vllm/engine/llm_engine.py

@@ -6,7 +6,6 @@ from typing import Type, TypeVar, Union
 
 from transformers import GenerationConfig, PreTrainedTokenizer
 
-import vllm
 from vllm.config import (CacheConfig, DecodingConfig, DeviceConfig, LoadConfig,
                          LoRAConfig, ModelConfig, ParallelConfig,
                          SchedulerConfig, SpeculativeConfig,
@@ -38,6 +37,7 @@ from vllm.transformers_utils.tokenizer_group import (BaseTokenizerGroup,
 from vllm.usage.usage_lib import (UsageContext, is_usage_stats_enabled,
                                   usage_message)
 from vllm.utils import Counter
+from vllm.version import __version__ as VLLM_VERSION
 
 logger = init_logger(__name__)
 _LOCAL_LOGGING_INTERVAL_SEC = 5
@@ -169,7 +169,7 @@ class LLMEngine:
             "enforce_eager=%s, kv_cache_dtype=%s, "
             "quantization_param_path=%s, device_config=%s, "
             "decoding_config=%r, seed=%d, served_model_name=%s)",
-            vllm.__version__,
+            VLLM_VERSION,
             model_config.model,
             speculative_config,
             model_config.tokenizer,

vllm/entrypoints/openai/api_server.py

@@ -15,7 +15,6 @@ from fastapi.responses import JSONResponse, Response, StreamingResponse
 from prometheus_client import make_asgi_app
 from starlette.routing import Mount
 
-import vllm
 import vllm.envs as envs
 from vllm.engine.arg_utils import AsyncEngineArgs
 from vllm.engine.async_llm_engine import AsyncLLMEngine
@@ -29,6 +28,7 @@ from vllm.entrypoints.openai.serving_completion import OpenAIServingCompletion
 from vllm.entrypoints.openai.serving_embedding import OpenAIServingEmbedding
 from vllm.logger import init_logger
 from vllm.usage.usage_lib import UsageContext
+from vllm.version import __version__ as VLLM_VERSION
 
 TIMEOUT_KEEP_ALIVE = 5  # seconds
 
@@ -93,7 +93,7 @@ async def show_available_models():
 
 @app.get("/version")
 async def show_version():
-    ver = {"version": vllm.__version__}
+    ver = {"version": VLLM_VERSION}
     return JSONResponse(content=ver)
 
 
@@ -174,7 +174,7 @@ if __name__ == "__main__":
             raise ValueError(f"Invalid middleware {middleware}. "
                              f"Must be a function or a class.")
 
-    logger.info("vLLM API server version %s", vllm.__version__)
+    logger.info("vLLM API server version %s", VLLM_VERSION)
     logger.info("args: %s", args)
 
     if args.served_model_name is not None:

vllm/entrypoints/openai/run_batch.py

@@ -5,7 +5,6 @@ from io import StringIO
 
 import aiohttp
 
-import vllm
 from vllm.engine.arg_utils import AsyncEngineArgs, nullable_str
 from vllm.engine.async_llm_engine import AsyncLLMEngine
 from vllm.entrypoints.openai.protocol import (BatchRequestInput,
@@ -15,6 +14,7 @@ from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
 from vllm.logger import init_logger
 from vllm.usage.usage_lib import UsageContext
 from vllm.utils import random_uuid
+from vllm.version import __version__ as VLLM_VERSION
 
 logger = init_logger(__name__)
 
@@ -135,7 +135,7 @@ async def main(args):
 if __name__ == "__main__":
     args = parse_args()
 
-    logger.info("vLLM API server version %s", vllm.__version__)
+    logger.info("vLLM API server version %s", VLLM_VERSION)
     logger.info("args: %s", args)
 
     asyncio.run(main(args))

vllm/usage/usage_lib.py

@@ -16,6 +16,7 @@ import requests
 import torch
 
 import vllm.envs as envs
+from vllm.version import __version__ as VLLM_VERSION
 
 _config_home = envs.VLLM_CONFIG_ROOT
 _USAGE_STATS_JSON_PATH = os.path.join(_config_home, "vllm/usage_stats.json")
@@ -163,9 +164,8 @@ class UsageMessage:
         ])
 
         # vLLM information
-        import vllm  # delayed import to prevent circular import
         self.context = usage_context.value
-        self.vllm_version = vllm.__version__
+        self.vllm_version = VLLM_VERSION
         self.model_architecture = model_architecture
 
         # Metadata

vllm/version.py

@@ -0,0 +1 @@
+__version__ = "0.5.0"