[Executor] CUDA Graph support padding batch (#2844)

* cuda graph support padding batch

* Integrate the startup parameters for the graph optimization backend and provide support for user - defined capture sizes.

* Do not insert max_num_seqs when the user specifies a capture list

* Support set graph optimization config from YAML file

* update cuda graph ci

* fix ci bug

* fix ci bug

fastdeploy/model_executor/forward_meta.py

@@ -17,11 +17,11 @@
 import logging
 from dataclasses import dataclass
 from enum import IntEnum, auto
-from typing import TYPE_CHECKING, Optional
-from fastdeploy.model_executor.layers.attention import AttentionBackend
+from typing import Optional
 
 import paddle
-    
+
+from fastdeploy.model_executor.layers.attention import AttentionBackend
 
 logger = logging.getLogger(__name__)
 
@@ -64,8 +64,6 @@ class ForwardMeta():
 
     # Use cuda graph in this step or not. Used to avoid run cuda graph when in dummy run or prefill stage.
     step_use_cudagraph: bool = False
-    # Batch type flag
-    is_decode_batch: bool = False
 
     # Attention backend object
     attn_backend: AttentionBackend = None