[Bugfix] mamba-block-size is set for vision language model (#27773)

Signed-off-by: Chen Zhang <zhangch99@outlook.com>

vllm/config/cache.py

@@ -5,7 +5,7 @@ import hashlib
 from dataclasses import field
 from typing import TYPE_CHECKING, Any, Literal
 
-from pydantic import Field, SkipValidation, field_validator, model_validator
+from pydantic import Field, SkipValidation, field_validator
 from pydantic.dataclasses import dataclass
 
 from vllm.config.utils import config
@@ -185,11 +185,3 @@ class CacheConfig:
             raise ValueError("Too large swap space. " + msg)
         elif cpu_memory_usage > 0.4 * total_cpu_memory:
             logger.warning("Possibly too large swap space. %s", msg)
-
-    @model_validator(mode="after")
-    def validate_mamba_block_size(self) -> "CacheConfig":
-        if self.mamba_block_size is not None and not self.enable_prefix_caching:
-            raise ValueError(
-                "--mamba-block-size can only be set with --enable-prefix-caching"
-            )
-        return self

vllm/config/vllm.py

@@ -17,7 +17,7 @@ from pathlib import Path
 from typing import TYPE_CHECKING, Any, TypeVar
 
 import torch
-from pydantic import ConfigDict, Field
+from pydantic import ConfigDict, Field, model_validator
 from pydantic.dataclasses import dataclass
 
 import vllm.envs as envs
@@ -943,6 +943,20 @@ class VllmConfig:
             f"compilation_config={self.compilation_config!r}"
         )
 
+    @model_validator(mode="after")
+    def validate_mamba_block_size(self) -> "VllmConfig":
+        if self.model_config is None:
+            return self
+        mamba_block_size_is_set = (
+            self.cache_config.mamba_block_size is not None
+            and self.cache_config.mamba_block_size != self.model_config.max_model_len
+        )
+        if mamba_block_size_is_set and not self.cache_config.enable_prefix_caching:
+            raise ValueError(
+                "--mamba-block-size can only be set with --enable-prefix-caching"
+            )
+        return self
+
 
 _current_vllm_config: VllmConfig | None = None
 _current_prefix: str | None = None