sglang0.4.5.post1/benchmark/mmlu

Download data

bash download_data.sh

Run benchmark

Benchmark sglang

python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000

python3 bench_sglang.py --nsub 10

# OpenAI models
python3 bench_sglang.py --backend gpt-3.5-turbo --parallel 8

Benchmark vllm

python3 -m vllm.entrypoints.api_server --tokenizer-mode auto --model meta-llama/Llama-2-7b-chat-hf --disable-log-requests --port 21000

python3 bench_other.py --nsub 10 --backend vllm

Benchmark lightllm

# A10G
python -m lightllm.server.api_server --tokenizer_mode auto --model_dir ~/model_weights/llama-2-7b-chat-hf --max_total_token_num 16000 --port 22000

# V100
python -m lightllm.server.api_server --tokenizer_mode auto --model_dir ~/model_weights/llama-2-7b-chat-hf --max_total_token_num 4500 --port 22000

python3 bench_other.py --nsub 10 --backend lightllm

Benchmark guidance

python3 bench_other.py --nsub 10 --backend guidance --parallel 1 --n-ctx 4096 --model-path path/to/gguf

Benchmark lmql

CUDA_VISIBLE_DEVICES=0,1 lmql serve-model meta-llama/Llama-2-7b-chat-hf --cuda --port 23000

python3 bench_other.py --nsub 10 --backend lmql --parallel 2