[V1] Unify VLLM_ENABLE_V1_MULTIPROCESSING handling in RayExecutor (#11472)

2024-12-24 19:49:46 -08:00
parent 3f3e92e1f2
commit 9832e5572a
3 changed files with 4 additions and 8 deletions
--- a/tests/basic_correctness/test_basic_correctness.py
+++ b/tests/basic_correctness/test_basic_correctness.py
@@ -127,11 +127,6 @@ def test_models_distributed(
    if attention_backend:
        os.environ["VLLM_ATTENTION_BACKEND"] = attention_backend

-    # Import VLLM_USE_V1 dynamically to handle patching
-    from vllm.envs import VLLM_USE_V1
-    if VLLM_USE_V1 and distributed_executor_backend != "mp":
-        os.environ["VLLM_ENABLE_V1_MULTIPROCESSING"] = "0"
-
    dtype = "half"
    max_tokens = 5