[V1] [Hybrid] Enable Full CUDA graph by default for hybrid models in V1 (#22594)

Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>

vllm/model_executor/models/config.py

@@ -4,6 +4,7 @@ from copy import deepcopy
 from typing import TYPE_CHECKING
 
 import vllm.envs as envs
+from vllm.config.compilation import CUDAGraphMode
 from vllm.logger import init_logger
 from vllm.model_executor.models import ModelRegistry
 from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE, cdiv
@@ -275,6 +276,42 @@ class GptOssForCausalLMConfig(VerifyAndUpdateConfig):
                     "%d for performance.", 1024)
 
 
+class MambaModelConfig(VerifyAndUpdateConfig):
+
+    @classmethod
+    def verify_and_update_config(cls, vllm_config: "VllmConfig") -> None:
+        """
+        Enable FULL_AND_PIECEWISE cuda graph mode by default (required
+        to get good performance for mamba layers in V1).
+
+        Args:
+            vllm_config: vLLM Config
+        """
+
+        if not envs.VLLM_USE_V1:
+            return
+
+        model_config = vllm_config.model_config
+        compilation_config = vllm_config.compilation_config
+
+        model_cls, _ = ModelRegistry.resolve_model_cls(
+            model_config.architecture,
+            model_config=model_config,
+        )
+
+        # TODO(tdoublep): remove as full cuda graph support is added
+        FCG_NOT_SUPPORTED_MODELS = [
+            "Lfm2ForCausalLM", "MiniMaxText01ForCausalLM"
+        ]
+
+        if (model_config.architecture not in FCG_NOT_SUPPORTED_MODELS
+                and compilation_config.cudagraph_mode is None):
+            logger.info(
+                "Hybrid or mamba-based model detected: setting cudagraph mode "
+                "to FULL_AND_PIECEWISE in order to optimize performance.")
+            compilation_config.cudagraph_mode = CUDAGraphMode.FULL_AND_PIECEWISE
+
+
 class HybridAttentionMambaModelConfig(VerifyAndUpdateConfig):
 
     @classmethod
@@ -293,6 +330,9 @@ class HybridAttentionMambaModelConfig(VerifyAndUpdateConfig):
         if not envs.VLLM_USE_V1:
             return
 
+        # Enable FULL_AND_PIECEWISE by default
+        MambaModelConfig.verify_and_update_config(vllm_config)
+
         cache_config = vllm_config.cache_config
         model_config = vllm_config.model_config
         parallel_config = vllm_config.parallel_config
@@ -374,4 +414,6 @@ MODELS_CONFIG_MAP: dict[str, type[VerifyAndUpdateConfig]] = {
     "JambaForSequenceClassification": JambaForSequenceClassificationConfig,
     "GraniteMoeHybridForCausalLM": GraniteMoeHybridModelConfig,
     "GptOssForCausalLM": GptOssForCausalLMConfig,
+    "MambaForCausalLM": MambaModelConfig,
+    "Mamba2ForCausalLM": MambaModelConfig,
 }