From 54da74e83e428fd26fdc5858a30e5c951e9db0df Mon Sep 17 00:00:00 2001
From: Liangsheng Yin <hnyls2002@gmail.com>
Date: Wed, 2 Sep 2026 21:09:04 -0700
Subject: [PATCH] [Fix] Broadcast PP dynamic-chunk profiling failures so every
 rank disables together (#37675)

---
 python/sglang/srt/managers/scheduler.py       |  1 +
 .../dynamic_chunk_sizer.py                    | 57 +++++++++++--------
 2 files changed, 34 insertions(+), 24 deletions(-)

diff --git a/python/sglang/srt/managers/scheduler.py b/python/sglang/srt/managers/scheduler.py
index 4d2755acf..b002d9d87 100644
--- a/python/sglang/srt/managers/scheduler.py
+++ b/python/sglang/srt/managers/scheduler.py
@@ -1265,6 +1265,7 @@ class Scheduler(
             page_size=self.page_size,
             device=self.device,
             pp_group=self.pp_group,
+            world_group=self.world_group,
             pp_rank=self.ps.pp_rank,
         )
         if sizer.profile_and_fit():
diff --git a/python/sglang/srt/managers/scheduler_components/dynamic_chunk_sizer.py b/python/sglang/srt/managers/scheduler_components/dynamic_chunk_sizer.py
index 0dc3ab208..ca1f868d4 100644
--- a/python/sglang/srt/managers/scheduler_components/dynamic_chunk_sizer.py
+++ b/python/sglang/srt/managers/scheduler_components/dynamic_chunk_sizer.py
@@ -8,10 +8,8 @@ from typing import TYPE_CHECKING, List, Optional, Tuple
 
 import numpy as np
 import torch
-import torch.distributed
 from tqdm import tqdm
 
-from sglang.srt.distributed.communication_op import attn_cp_tp_broadcast_pyobj
 from sglang.srt.environ import envs
 from sglang.srt.layers.dp_attention import (
     get_attention_dp_rank,
@@ -23,6 +21,7 @@ from sglang.srt.managers.schedule_batch import Req, ScheduleBatch
 from sglang.srt.mem_cache.common import release_kv_cache
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch, PPProxyTensors
 from sglang.srt.sampling.sampling_params import SamplingParams
+from sglang.srt.utils import broadcast_pyobj
 from sglang.srt.utils.common import get_device_module
 
 if TYPE_CHECKING:
@@ -54,6 +53,7 @@ class DynamicChunkSizer:
         page_size: int,
         device: str,
         pp_group: GroupCoordinator,
+        world_group: GroupCoordinator,
         pp_rank: int,
     ):
         self.model_runner = model_runner
@@ -67,41 +67,50 @@ class DynamicChunkSizer:
         self.page_size = page_size
         self.device = device
         self.pp_group = pp_group
+        self.world_group = world_group
         self.pp_rank = pp_rank
         self.predictor = ChunkSizePredictor()
 
     def profile_and_fit(self) -> bool:
         """PP0 profiles synthetic prefills and every rank fits the same samples;
         returns whether the predictor is ready."""
+        samples: Optional[Tuple[List[int], List[float]]] = None
+
+        if self.pp_group.is_first_rank:
+            try:
+                samples = self._profile_prefill_latency()
+            except Exception as e:
+                logger.warning(
+                    f"[PP Dynamic Chunk] Failed to profile prefill latency: {e!r}. "
+                    "Dynamic chunking will be disabled."
+                )
+
+        # The samples are global, so one broadcast from global rank 0 (a PP0 rank)
+        # reaches every stage and attention rank; a failure travels as None.
+        samples = broadcast_pyobj(
+            [samples], self.world_group.rank, self.world_group.cpu_group, src=0
+        )[0]
+
+        if samples is None:
+            return False
+
+        seq_lens, latencies = samples
+        # Quadratic model: f(l) = al^2 + bl + c
         try:
-            seq_lens: List[int] = []
-            latencies: List[float] = []
-
-            if self.pp_group.is_first_rank:
-                seq_lens, latencies = self._profile_prefill_latency()
-
-                seq_lens, latencies = attn_cp_tp_broadcast_pyobj([seq_lens, latencies])
-
-            # Broadcast data to all ranks
-            if torch.distributed.is_available() and torch.distributed.is_initialized():
-                data_to_sync = [seq_lens, latencies]
-                self.pp_group.broadcast_object_list(data_to_sync, src=0)
-                seq_lens, latencies = data_to_sync
-
-            # Quadratic model: f(l) = al^2 + bl + c
             self.predictor.fit(seq_lens, latencies)
-            self.predictor.set_target_latency(self.chunked_prefill_size)
-            self.predictor.is_ready = True
-            logger.info(
-                f"[PP Dynamic Chunk] [PP{self.pp_rank}] Predictor ready (quadratic). "
-                f"Target latency: {self.predictor.target_latency:.2f}ms"
-            )
         except Exception as e:
+            # Every rank fits the same samples, so this fails on all of them alike.
             logger.warning(
-                f"[PP Dynamic Chunk] Failed to profile prefill latency: {e!r}. "
+                f"[PP Dynamic Chunk] Failed to fit the chunk-size predictor: {e!r}. "
                 "Dynamic chunking will be disabled."
             )
             return False
+        self.predictor.set_target_latency(self.chunked_prefill_size)
+        self.predictor.is_ready = True
+        logger.info(
+            f"[PP Dynamic Chunk] [PP{self.pp_rank}] Predictor ready (quadratic). "
+            f"Target latency: {self.predictor.target_latency:.2f}ms"
+        )
         return True
 
     def predict(self, history_len: int) -> Optional[int]: