From 21a4a16b4bf2d87a45e86f2636341a1d2c0dcf5d Mon Sep 17 00:00:00 2001
From: abing <wangbingjia.wbj@ant-intl.com>
Date: Thu, 17 Sep 2026 19:43:41 -0700
Subject: [PATCH] update code

(cherry picked from commit 1453150a0bdf2cce425e9e2da83ef16b8e8f2859)
---
 .../sglang/srt/arg_groups/deepseek_v4_hook.py  | 18 +++++++++++++++---
 .../layers/attention/deepseek_v4_backend.py    |  6 +++++-
 .../runner/prefill_cuda_graph_runner.py        |  6 +++---
 3 files changed, 23 insertions(+), 7 deletions(-)

diff --git a/python/sglang/srt/arg_groups/deepseek_v4_hook.py b/python/sglang/srt/arg_groups/deepseek_v4_hook.py
index 6f901263e..2d5c38ad9 100644
--- a/python/sglang/srt/arg_groups/deepseek_v4_hook.py
+++ b/python/sglang/srt/arg_groups/deepseek_v4_hook.py
@@ -263,9 +263,21 @@ def validate_deepseek_v41_features(server_args: ServerArgs) -> None:
     from sglang.srt.model_executor.cuda_graph_config import Backend, Phase, with_phase
 
     prefill_graph = cfg.cuda_graph_config.prefill
-    if prefill_graph.backend != Backend.DISABLED and prefill_graph.max_seq_len is None:
-        # The captured low-ratio indexer scores a static context width; 16k
-        # keeps it inside the candidate window at under 1 ms per layer.
+    cp_breakable_prefill = (
+        cfg.enable_prefill_cp
+        and cfg.cp_strategy == "interleave"
+        and cfg.tp_size > 1
+        and prefill_graph.backend == Backend.BREAKABLE
+    )
+    if (
+        prefill_graph.backend != Backend.DISABLED
+        and prefill_graph.max_seq_len is None
+        and not cp_breakable_prefill
+    ):
+        # The non-CP captured low-ratio indexer scores a static context width.
+        # CP BCG runs these sources eagerly with live prefix metadata, so this
+        # default would only force long-prefix CP batches back to eager.
+        # Explicit max_seq_len values still constrain both paths.
         declare_resolution(
             server_args,
             "validate_deepseek_v41_features",
diff --git a/python/sglang/srt/layers/attention/deepseek_v4_backend.py b/python/sglang/srt/layers/attention/deepseek_v4_backend.py
index 86bb2e580..d11eb406d 100644
--- a/python/sglang/srt/layers/attention/deepseek_v4_backend.py
+++ b/python/sglang/srt/layers/attention/deepseek_v4_backend.py
@@ -1580,8 +1580,12 @@ class DeepseekV4AttnBackend(
 
     @property
     def low_ratio_prefill_graph(self) -> bool:
+        """Whether ratio-1/2 sources use captured projections and indexer metadata."""
         return (
-            bool(self.low_ratios) and _has_dense_fp4_indexer() and _is_sm100_or_newer()
+            bool(self.low_ratios)
+            and _has_dense_fp4_indexer()
+            and _is_sm100_or_newer()
+            and get_parallel().attn_cp_size == 1
         )
 
     def can_run_prefill_cuda_graph(self, forward_batch: ForwardBatch) -> bool:
diff --git a/python/sglang/srt/model_executor/runner/prefill_cuda_graph_runner.py b/python/sglang/srt/model_executor/runner/prefill_cuda_graph_runner.py
index 17bfceb8b..48695bc62 100644
--- a/python/sglang/srt/model_executor/runner/prefill_cuda_graph_runner.py
+++ b/python/sglang/srt/model_executor/runner/prefill_cuda_graph_runner.py
@@ -1336,9 +1336,9 @@ class PrefillCudaGraphRunner(BaseCudaGraphRunner):
             batch_max_context_len=batch_max_context_len,
         ):
             return False
-        if getattr(self, "enable_cp_bcg_capture", False) and is_cp_active(
-            forward_batch
-        ):
+        if getattr(self, "enable_cp_bcg_capture", False):
+            if not is_cp_active(forward_batch):
+                return False
             assert self.prefill_cp_bcg_input is not None
             if (
                 self.prefill_cp_bcg_input.select_replay_bucket_for_batch(