From 7cc4d5ed0203fa4c60d99ee26b8000ec270a4c11 Mon Sep 17 00:00:00 2001 From: TianyeDong Date: Sun, 5 Jul 2026 12:38:25 -0400 Subject: [PATCH] refactor(rlix-validation): remove deprecated no-op assert_partial_overlap_standalone_safe The function was a documented no-op with no callers. Drop it and its __all__ entry. Co-Authored-By: Claude Opus 4.8 --- miles/utils/rlix_validation.py | 19 ------------------- 1 file changed, 19 deletions(-) diff --git a/miles/utils/rlix_validation.py b/miles/utils/rlix_validation.py index 3d582e67c7..2a232e421c 100644 --- a/miles/utils/rlix_validation.py +++ b/miles/utils/rlix_validation.py @@ -405,29 +405,10 @@ def assert_rlix_topology(args: Any, sglang_config: Any | None = None) -> None: ) -def assert_partial_overlap_standalone_safe(args: Any) -> None: - """Deprecated no-op kept for callsite stability. - - Originally intended to refuse partial-overlap topologies under the - standalone entry. Removed because the standalone ``create_placement_groups`` - path offsets the rollout pool by ``actor_num_nodes * actor_num_gpus_per_node``, - so zero-based ``train ⊂ infer`` is also true for valid disjoint - standalone configs (e.g. 4 actor GPUs + ``--rollout-num-gpus 8``). - Auto-classifying intent from args alone is unsafe; the env-var guard - (``RLIX_CONTROL_PLANE=rlix``) is the only reliable signal at the standalone - entry. RLix-mode partial-overlap correctness is enforced by - :func:`assert_rlix_topology` (C1) inside the RLix entry driver. - """ - # Intentional no-op — see docstring. - del args - return None - - __all__ = [ "is_rlix_mode", "train_devices_subset_of_infer", "async_generation_enabled", "single_updateable_model_and_server", "assert_rlix_topology", - "assert_partial_overlap_standalone_safe", ]