From d42ee34bd50fb1abba159bd02d89a72c750c25dd Mon Sep 17 00:00:00 2001
From: Story Crater Bot <19826264+Riotpiaole@users.noreply.github.com>
Date: Wed, 19 Aug 2026 18:14:22 -0700
Subject: [PATCH] reasoning: halve max-num-seqs to 2 for Kimi swap's first boot

New model's weight footprint (35B total MoE at on-the-fly bnb-4bit) leaves
less confirmed KV-cache headroom on the 32GB card than the old one had --
reducing concurrent-sequence worst case until real memory use is verified.
---
 k8s/apps/llm-serving/reasoning.yaml | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/k8s/apps/llm-serving/reasoning.yaml b/k8s/apps/llm-serving/reasoning.yaml
index e500955..859929e 100644
--- a/k8s/apps/llm-serving/reasoning.yaml
+++ b/k8s/apps/llm-serving/reasoning.yaml
@@ -40,7 +40,13 @@ spec:
       - --tensor-parallel-size=1
       - --max-model-len=16384
       - --gpu-memory-utilization=0.90
-      - --max-num-seqs=4
+      # Halved from 4 -- weights alone (35B total, MoE, all experts loaded
+      # regardless of ~3B active) at on-the-fly bnb 4bit are ~19-21GB on a
+      # 32GB card at 0.90 util (~29.5GB budget), leaving only ~8-10GB for KV
+      # cache + activations. Old DeepSeek-32B was in the same weight
+      # ballpark but MoE bnb overhead here is untested. Starting at 2 to
+      # confirm real headroom via nvidia-smi/vLLM logs before raising back.
+      - --max-num-seqs=2
       - --enable-chunked-prefill
       - --enable-prefix-caching
       # qwen3 parser handles this family's <think> reasoning blocks (best