adapt to sglang v0.5.2rc1 on dcu

benchmark/json_schema/README.md

+## Run benchmark
+
+### Benchmark sglang
+
+Run Llama-8b
+
+```bash
+python3 -m sglang.launch_server --model-path meta-llama/Llama-3.1-8B-Instruct --port 30000
+```
+
+Benchmark
+
+```bash
+python3 bench_sglang.py
+```