[Bugfix] Share resolved KV cache layout across config copies

tests/config/test_config_utils.py

@@ -8,7 +8,8 @@
 
 from vllm.config.cache import CacheConfig
 from vllm.config.scheduler import SchedulerConfig
-from vllm.config.utils import get_hash_factors, hash_factors, normalize_value
+from vllm.config.utils import get_hash_factors, hash_factors, normalize_value, replace
+from vllm.v1.kv_cache_layout import KVCacheLayout
 
 # Helpers
 
@@ -217,6 +218,17 @@ def test_cache_config_hash_ignores_kv_cache_sizing_knobs():
     assert CacheConfig(gpu_memory_utilization=0.5).compute_hash() == base_hash
 
 
+def test_cache_config_replace_shares_resolved_kv_cache_layout():
+    target = CacheConfig(cache_dtype="auto")
+    draft = replace(target, cache_dtype="fp8")
+
+    target.kv_cache_layout = "LBHNC"
+
+    assert target.get_resolved_kv_cache_layout() is KVCacheLayout.LBHNC
+    assert draft.get_resolved_kv_cache_layout() is KVCacheLayout.LBHNC
+    assert target.metrics_info()["kv_cache_layout"] == "LBHNC"
+
+
 def test_scheduler_config_hash_includes_max_num_seqs():
     """Per-request workspace sizes must invalidate compiled graphs."""
     base_hash = SchedulerConfig(

vllm/config/cache.py

@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from collections.abc import Callable
-from dataclasses import field
+from dataclasses import dataclass, field
 from functools import cache
 from typing import Any, ClassVar, Literal
 
@@ -73,6 +73,13 @@ def _get_prefix_cache_retention_interval() -> int | None:
 KVOffloadingBackend = Literal["native", "lmcache"]
 
 
+@dataclass
+class ResolvedKVCacheLayout:
+    """Shared state for the resolved physical KV cache layout."""
+
+    value: str | None = None
+
+
 @config
 class CacheConfig:
     """Configuration for the KV cache."""
@@ -86,17 +93,29 @@ class CacheConfig:
     """Whether block_size was explicitly provided. Derived automatically."""
     user_specified_mamba_block_size: bool = field(default=False, init=False)
     """Whether mamba_block_size was explicitly provided. Derived automatically."""
-    kv_cache_layout: str | None = field(default=None, init=False)
-    """Resolved physical KV cache layout name (a ``KVCacheLayout`` member).
+    _resolved_kv_cache_layout: ResolvedKVCacheLayout = field(
+        default_factory=ResolvedKVCacheLayout,
+        repr=False,
+    )
+    """Shared state for the resolved physical KV cache layout name.
 
-    ``None`` means the layout has not been resolved yet. The engine core
+    A ``None`` value means the layout has not been resolved yet. The engine core
     resolves it once (``resolve_kv_cache_layout``) before memory profiling, and
     every worker process adopts the resolved name — via the
     ``set_kv_cache_layout`` RPC, or ``KVCacheConfig.kv_cache_layout`` for
     workers spawned after resolution — before the KV cache is allocated. Once
     set the value is final and read with ``get_resolved_kv_cache_layout``,
     which raises on ``None``. Tests and standalone tools may pre-set a value,
     which resolution then honors as-is."""
+
+    @property
+    def kv_cache_layout(self) -> str | None:
+        return self._resolved_kv_cache_layout.value
+
+    @kv_cache_layout.setter
+    def kv_cache_layout(self, value: str | None) -> None:
+        self._resolved_kv_cache_layout.value = value
+
     prefix_match_unit: int | None = Field(default=None, gt=0)
     """The finest token boundary (in tokens) a prefix-cache hit can land on.
 
@@ -296,7 +315,10 @@ def compute_hash(self) -> str:
     def metrics_info(self):
         # convert cache_config to dict(key: str, value: str) for prometheus
         # metrics info
-        return {key: str(value) for key, value in self.__dict__.items()}
+        info = {key: str(value) for key, value in self.__dict__.items()}
+        info.pop("_resolved_kv_cache_layout")
+        info["kv_cache_layout"] = str(self.kv_cache_layout)
+        return info
 
     _block_size_resolved: bool = field(default=False, init=False)
     """Guard against pydantic re-running _apply_block_size_default."""