From 659fba6a5d967c2809f72a7d51a271bc8a966186 Mon Sep 17 00:00:00 2001
From: Frank <frank@dankware>
Date: Sat, 8 Aug 2026 12:56:28 -0700
Subject: [PATCH] Lower --max-num-batched-tokens to 1536 to free VRAM for
 weights

---
 vllm-qwen36.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/vllm-qwen36.sh b/vllm-qwen36.sh
index d8a947b..1e0216a 100755
--- a/vllm-qwen36.sh
+++ b/vllm-qwen36.sh
@@ -20,7 +20,7 @@ exec docker run --rm --name vllm-qwen36 \
     --gpu-memory-utilization 0.95 \
     --max-num-seqs 4 \
     --max-model-len 32768 \
-    --max-num-batched-tokens 8192 \
+    --max-num-batched-tokens 1536 \
     --trust-remote-code \
     --cpu-offload-gb 40 \
     --cpu-offload-params "experts" \