[Core] Remove the legacy Ray compiled-graph executor (#61712)

Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Woosuk Kwon
2026-10-14 18:13:27 -07:00
committed by GitHub
co-authored by Claude Opus 5.5
parent bfdbfa1632
commit e57bb4ace7
16 changed files with 14 additions and 1257 deletions
+1 -5
View File
@@ -182,19 +182,15 @@ class MockSubscriber:
@pytest.fixture
def enable_ray_v2_backend():
"""Set env vars for the Ray V2 executor backend and shut down Ray
"""Run the engine in-process for Ray executor tests and shut down Ray
between tests."""
import ray
saved = {
"VLLM_USE_RAY_V2_EXECUTOR_BACKEND": os.environ.get(
"VLLM_USE_RAY_V2_EXECUTOR_BACKEND"
),
"VLLM_ENABLE_V1_MULTIPROCESSING": os.environ.get(
"VLLM_ENABLE_V1_MULTIPROCESSING"
),
}
os.environ["VLLM_USE_RAY_V2_EXECUTOR_BACKEND"] = "1"
os.environ["VLLM_ENABLE_V1_MULTIPROCESSING"] = "0"
if ray.is_initialized():
ray.shutdown()
+1 -10
View File
@@ -294,16 +294,7 @@ def _compare_tp(
if max_num_seqs:
common_args.extend(["--max-num-seqs", f"{max_num_seqs}"])
if distributed_backend == "ray":
# Test Ray Compiled Graph for all the tests
pp_env = {
"VLLM_USE_RAY_COMPILED_DAG_NCCL_CHANNEL": "1",
}
elif distributed_backend == "mp":
pp_env = None
else:
pp_env = None
pp_env = None
tp_env = None
pp_args = [
@@ -38,7 +38,6 @@ def ray_init():
class _AsyncLLMActor:
def start(self, pg, bundle_indices=None, ray_runtime_env=None):
os.environ["VLLM_USE_RAY_V2_EXECUTOR_BACKEND"] = "1"
# Needed so collective_rpc can pickle _get_env_var over the
# AsyncLLM -> EngineCore ZMQ boundary.
os.environ["VLLM_ALLOW_INSECURE_SERIALIZATION"] = "1"