[V1] Logits processors extensibility (#19912)

Signed-off-by: Andrew Feldman <afeldman@redhat.com>
Signed-off-by: Andrew Feldman <afeld2012@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: Nick Hill <nhill@redhat.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
Co-authored-by: Andrew Feldman <afeld2012@gmail.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

vllm/config/__init__.py

@@ -62,6 +62,7 @@ if TYPE_CHECKING:
         QuantizationConfig)
     from vllm.model_executor.model_loader import LoadFormats
     from vllm.model_executor.model_loader.tensorizer import TensorizerConfig
+    from vllm.v1.sample.logits_processor import LogitsProcessor
 
     HfOverrides = Union[dict, Callable[[type], type]]
 else:
@@ -72,6 +73,7 @@ else:
     BaseModelLoader = Any
     LoadFormats = Any
     TensorizerConfig = Any
+    LogitsProcessor = Any
     HfOverrides = Union[dict[str, Any], Callable[[type], type]]
 
     me_quant = LazyLoader("model_executor", globals(),
@@ -465,6 +467,9 @@ class ModelConfig:
     - "transformers" will use the Transformers model implementation."""
     override_attention_dtype: Optional[str] = None
     """Override dtype for attention"""
+    logits_processors: Optional[list[Union[str, type[LogitsProcessor]]]] = None
+    """One or more logits processors' fully-qualified class names or class
+    definitions"""
 
     def compute_hash(self) -> str:
         """