How to Debug and Locate Llama Model Code in SGLang for Profiling Analysis

[ziyuhuang123]

I am using the following code:

python -m sglang.launch_server \
    --model-path /root/zyhuang/sglang/test/llama_config/dir \
    --load-format dummy
python -m sglang.bench_serving \
    --backend sglang \
    --model /root/zyhuang/sglang/test/llama_config/dir \
    --num-prompts 1 \
    --sharegpt-output-len 10 \
    --profile

I want to specifically locate the code file related to the Llama model in SGLang. How can I pinpoint where the model's code is executed? Should I debug it step-by-step? Also, since the host and server are separated into two different windows, is there a simple way to debug it?