From 9dfa16fae9680f2024e355ada7e6f75646362d0a Mon Sep 17 00:00:00 2001
From: Frank <frank@dankware>
Date: Sun, 9 Aug 2026 12:08:26 -0700
Subject: [PATCH] Disable speculative decoding - OOM on stable vLLM too

---
 scripts/vllm-awq-qwen36-dflash-ampere.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/vllm-awq-qwen36-dflash-ampere.sh b/scripts/vllm-awq-qwen36-dflash-ampere.sh
index c80dea8..9d00d39 100644
--- a/scripts/vllm-awq-qwen36-dflash-ampere.sh
+++ b/scripts/vllm-awq-qwen36-dflash-ampere.sh
@@ -22,7 +22,7 @@ exec docker run --rm --name vllm-awq-qwen36-dflash-ampere \
     -v ~/.cache/huggingface:/root/.cache/huggingface \
     vllm/vllm-openai:nightly "${MODEL}" \
     --served-model-name "frank-default-small" \
-    --speculative-config '{"method": "dflash", "model": "z-lab/Qwen3.6-35B-A3B-DFlash", "num_speculative_tokens": 12}' \
+    # --speculative-config '{"method": "dflash", "model": "z-lab/Qwen3.6-35B-A3B-DFlash", "num_speculative_tokens": 12}' \
     --max-num-seqs 2 \
     --max-model-len 32768 \
     --tensor-parallel-size 1 \