[Misc] Add deprecation warning for beam search (#6402)

Author: WoosukKwon

vllm/envs.py

@@ -41,6 +41,7 @@
     NVCC_THREADS: Optional[str] = None
     VLLM_USE_PRECOMPILED: bool = False
     VLLM_INSTALL_PUNICA_KERNELS: bool = False
+    VLLM_NO_DEPRECATION_WARNING: bool = False
     CMAKE_BUILD_TYPE: Optional[str] = None
     VERBOSE: bool = False
 
@@ -251,6 +252,10 @@
     lambda: os.getenv("VLLM_XLA_CACHE_PATH", "~/.vllm/xla_cache/"),
     "VLLM_FUSED_MOE_CHUNK_SIZE":
     lambda: int(os.getenv("VLLM_FUSED_MOE_CHUNK_SIZE", "65536")),
+
+    # If set, vllm will skip the deprecation warnings.
+    "VLLM_NO_DEPRECATION_WARNING":
+    lambda: bool(int(os.getenv("VLLM_NO_DEPRECATION_WARNING", "0"))),
 }
 
 # end-env-vars-definition

vllm/sampling_params.py

@@ -8,6 +8,11 @@
 from pydantic import Field
 from typing_extensions import Annotated
 
+import vllm.envs as envs
+from vllm.logger import init_logger
+
+logger = init_logger(__name__)
+
 _SAMPLING_EPS = 1e-5
 
 
@@ -184,6 +189,13 @@ def __init__(
 
         self._verify_args()
         if self.use_beam_search:
+            if not envs.VLLM_NO_DEPRECATION_WARNING:
+                logger.warning(
+                    "[IMPORTANT] We plan to discontinue the support for beam "
+                    "search in the next major release. Please refer to "
+                    "https://github.com/vllm-project/vllm/issues/6226 for "
+                    "more information. Set VLLM_NO_DEPRECATION_WARNING=1 to "
+                    "suppress this warning.")
             self._verify_beam_search()
         else:
             self._verify_non_beam_search()