Baseline: ThinkingCap 27B FP8 on dankclaw with enforce-eager and validation timeout

scripts/serve-thinkingcap.sh

@@ -21,6 +21,7 @@ exec "${VENV}/bin/vllm" serve "${MODEL}" \
     --host 0.0.0.0 \
     --kv-cache-dtype fp8 \
     --gpu-memory-utilization 0.80 \
+    --enforce-eager \
     --max-model-len 32768 \
     --max-num-batched-tokens 2048 \
     --max-num-seqs 32 \

scripts/validate.sh

@@ -14,7 +14,7 @@ echo ""
 
 START_TIME=$(date +%s%N)
 
-RESPONSE=$(curl -s -w "\n%{http_code}" \
+RESPONSE=$(curl -s --max-time 60 -w "\n%{http_code}" \
     "http://${HOST}:${PORT}/v1/chat/completions" \
     -H "Content-Type: application/json" \
     -d "{
@@ -37,6 +37,11 @@ echo "Response Time: ${ELAPSED_MS}ms"
 echo ""
 
 if [ "$HTTP_CODE" = "200" ]; then
+    if [ "$ELAPSED_MS" -gt 60000 ]; then
+        echo "FAIL — Response time ${ELAPSED_MS}ms exceeds 60s limit"
+        echo "TIME=${ELAPSED_MS}"
+        exit 1
+    fi
     CONTENT=$(echo "$BODY" | python3 -c "import sys,json; print(json.load(sys.stdin)['choices'][0]['message']['content'].strip())" 2>/dev/null)
     echo "Model Output: ${CONTENT}"
     echo ""