From 5a05d4528cefbe35ca341f2bce873476c5f540f3 Mon Sep 17 00:00:00 2001
From: project6-dev <dev@project6.ai>
Date: Thu, 13 Aug 2026 02:35:50 +0000
Subject: [PATCH] fix(stability): prevent OOM crash + disable garbled
 topk_softmax kernel
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Sub 655 analysis: 634/881 connection errors (server crash during replay).
Root cause: max-model-len=256000 + gpu-memory-utilization=0.95 + max-num-seqs=2
caused OOM on long-context requests (128K+ tokens).

Changes:
- max-model-len: 256000 → 131072 (enough for replay, prevents OOM)
- gpu-memory-utilization: 0.95 → 0.90 (safety margin)
- max-num-seqs: 2 → 1 (avoid concurrent long-context OOM)
- max-num-batched-tokens: 4096 → 8192 (match proven config)
- BI100_MOE_COREX_TOPK_SOFTMAX=0 (CUB kernel causes garbled output on
  BI-V100; PyTorch topk+softmax path is correct and fast enough)

Expected impact: server stays alive through entire replay+opencompass run.
Sub 655 successful requests had output_tps_avg=11.5 — the TPS is fine,
we just need the server to not crash.
---
 computility-run.yaml | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/computility-run.yaml b/computility-run.yaml
index 9fd7ac4d..c6a03f5b 100644
--- a/computility-run.yaml
+++ b/computility-run.yaml
@@ -8,18 +8,18 @@ command:
     - --served-model-name
     - llm
     - --max-model-len
-    - '256000'
+    - '131072'
     - --gpu-memory-utilization
-    - '0.95'
+    - '0.90'
     - --trust-remote-code
     - -tp
     - '4'
     - --max-num-seqs
-    - '2'
+    - '1'
     - --disable-log-requests
     - --disable-frontend-multiprocessing
     - --max-num-batched-tokens
-    - '4096'
+    - '8192'
     - --enable-chunked-prefill
     - --max-seq-len-to-capture
     - '32768'
@@ -45,3 +45,5 @@ env:
       value: admission64
     - name: BI100_GDN_RESTORE_MODE
       value: hybrid64
+    - name: BI100_MOE_COREX_TOPK_SOFTMAX
+      value: '0'