fix: add --max-lora-rank 32 to serve_test.sh

eqbench-ja-run/serve_test.sh

@@ -39,6 +39,7 @@ exec vllm serve "${BASE_MODEL}" \
     --max-num-seqs 16 \
     --enable-prefix-caching \
     --enable-lora \
+    --max-lora-rank 32 \
     --lora-modules "${LORA_NAME}=${LORA_REPO}" \
     --trust-remote-code \
     ${HF_TOKEN:+--hf-token "${HF_TOKEN}"}