[[V0 deprecation]]Remove VLLM_USE_V1 env (#28204)

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

vllm/envs.py

@@ -100,7 +100,6 @@ if TYPE_CHECKING:
     VLLM_SKIP_P2P_CHECK: bool = False
     VLLM_DISABLED_KERNELS: list[str] = []
     VLLM_DISABLE_PYNCCL: bool = False
-    VLLM_USE_V1: bool = True
     VLLM_ROCM_USE_AITER: bool = False
     VLLM_ROCM_USE_AITER_PAGED_ATTN: bool = False
     VLLM_ROCM_USE_AITER_LINEAR: bool = True
@@ -884,8 +883,6 @@ environment_variables: dict[str, Callable[[], Any]] = {
     "VLLM_DISABLE_PYNCCL": lambda: (
         os.getenv("VLLM_DISABLE_PYNCCL", "False").lower() in ("true", "1")
     ),
-    # If set, use the V1 code path.
-    "VLLM_USE_V1": lambda: bool(int(os.getenv("VLLM_USE_V1", "1"))),
     # Disable aiter ops unless specifically enabled.
     # Acts as a parent switch to enable the rest of the other operations.
     "VLLM_ROCM_USE_AITER": lambda: (
@@ -1538,16 +1535,6 @@ def is_set(name: str):
     raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
 
 
-def set_vllm_use_v1(use_v1: bool):
-    if is_set("VLLM_USE_V1"):
-        raise ValueError(
-            "Should not call set_vllm_use_v1() if VLLM_USE_V1 is set "
-            "explicitly by the user. Please raise this as a Github "
-            "Issue and explicitly set VLLM_USE_V1=0 or 1."
-        )
-    os.environ["VLLM_USE_V1"] = "1" if use_v1 else "0"
-
-
 def compute_hash() -> str:
     """
     WARNING: Whenever a new key is added to this environment

vllm/usage/usage_lib.py

@@ -42,7 +42,6 @@ _USAGE_ENV_VARS_TO_COLLECT = [
     "VLLM_USE_FLASHINFER_SAMPLER",
     "VLLM_PP_LAYER_PARTITION",
     "VLLM_USE_TRITON_AWQ",
-    "VLLM_USE_V1",
     "VLLM_ENABLE_V1_MULTIPROCESSING",
 ]