[Bugfix] set default set cuda_graph_sizes to min(self.max_num_seqs * 2, 512) (#20628)

Signed-off-by: izhuhaoran <izhuhaoran@qq.com>

Author: izhuhaoran

vllm/config.py

@@ -2147,11 +2147,12 @@ class SchedulerConfig:
     NOTE: This will be replaced by speculative config in the future; it is
     present to enable correctness tests until then."""
 
-    cuda_graph_sizes: list[int] = field(default_factory=lambda: [512])
-    """Cuda graph capture sizes, default is 512.
-    1. if one value is provided, then the capture list would follow the
+    cuda_graph_sizes: list[int] = field(default_factory=list)
+    """Cuda graph capture sizes
+    1. if none provided, then default set to [min(max_num_seqs * 2, 512)]
+    2. if one value is provided, then the capture list would follow the
     pattern: [1, 2, 4] + [i for i in range(8, cuda_graph_sizes + 1, 8)]
-    2. more than one value (e.g. 1 2 128) is provided, then the capture list
+    3. more than one value (e.g. 1 2 128) is provided, then the capture list
     will follow the provided list."""
 
     delay_factor: float = 0.0
@@ -2316,6 +2317,13 @@ def __post_init__(self) -> None:
                 self.max_num_partial_prefills, self.max_long_partial_prefills,
                 self.long_prefill_token_threshold)
 
+        # NOTE: Default set cuda_graph_sizes to [min(max_num_seqs * 2, 512)].
+        # This avoids OOM in tight memory scenarios with small max_num_seqs,
+        # and prevents capture of many large graphs (>512) that would greatly
+        # increase startup time with limited performance benefit.
+        if not self.cuda_graph_sizes:
+            self.cuda_graph_sizes = [min(self.max_num_seqs * 2, 512)]
+
     @model_validator(mode='after')
     def _verify_args(self) -> Self:
         if (self.max_num_batched_tokens < self.max_model_len