Enable DFlash speculative decoding (12 tokens), increase CPU offload to 20GB

2026-08-09 11:54:40 -07:00
parent 8c13c7bbaa
commit 578c3e8f71
1 changed files with 2 additions and 2 deletions
@@ -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": 4}' \
+    --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 \
@@ -33,6 +33,6 @@ exec docker run --rm --name vllm-awq-qwen36-dflash-ampere \
    --reasoning-parser qwen3 \
    --trust-remote-code \
    --gpu-memory-utilization 0.95 \
-    --cpu-offload-gb 13 \
+    --cpu-offload-gb 20 \
    --cpu-offload-params "experts" \
    --default-chat-template-kwargs '{"enable_thinking": true, "preserve_thinking": true}'