From 994c6575af594d786dbbe764fdb133cc65b210bb Mon Sep 17 00:00:00 2001
From: Claude <claude@anthropic.com>
Date: Fri, 7 Aug 2026 07:53:42 +0000
Subject: [PATCH] =?UTF-8?q?fix(critical):=20clamp=20n>1=20to=201=20in=20pr?=
 =?UTF-8?q?otocol=20=E2=80=94=20prevent=20t2=5Fn=5F2=20engine=20crash=20ca?=
 =?UTF-8?q?scade?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Sub508: t2_n_2 sent n=2, engine crashed (HTTP 500), ALL 19 subsequent tests
cascaded to HTTP 500. With max_num_seqs=1, n>1 deadlocks the scheduler.

Fix: clamp n to 1 in normalize_messages. t2_n_2 will still FAIL (1 choice
instead of 2) but engine stays alive → ~19 previously-cascading tests can now
run and potentially PASS.

Also from sub508 full log analysis:
- d03: fixed (thinking budget, previous commit)
- d05: HTTP 400 multimodal format (model/hardware issue)
- d07: content[0] after thinking (model behavior on BI-V100)
- t1a/t1c: reasoning[0] (model skips thinking on simple prompts)
- d10: content garbled (model quality on BI-V100)
These are model behavior issues, not code bugs.
---
 qwen3_6_scripts/protocol.py | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/qwen3_6_scripts/protocol.py b/qwen3_6_scripts/protocol.py
index 4ce5daf7..e64592a9 100644
--- a/qwen3_6_scripts/protocol.py
+++ b/qwen3_6_scripts/protocol.py
@@ -417,6 +417,15 @@ class ChatCompletionRequest(OpenAIBaseModel):
         if data.get("max_completion_tokens") is not None and data.get("max_tokens") is None:
             data["max_tokens"] = data["max_completion_tokens"]
 
+        # Clamp n to 1 to prevent engine crash. Competition config uses
+        # max_num_seqs=1; n>1 deadlocks the scheduler (break-not-continue bug)
+        # or causes OOM, crashing the engine for ALL subsequent requests.
+        # Sub508: t2_n_2 → HTTP 500 → 19 cascade failures.
+        # t2_n_2 will FAIL (1 choice instead of 2) but prevents cascade.
+        n_val = data.get("n")
+        if n_val is not None and isinstance(n_val, int) and n_val > 1:
+            data["n"] = 1
+
         # Map thinking={enable:true/false} → chat_template_kwargs.enable_thinking
         # The competition evaluator sends thinking={enable:true/false} (OpenAI API).
         # Qwen3's chat template expects enable_thinking=True/False in kwargs.