livetalking/llm

1、推理加速

conda create -n vllm python=3.10 conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=12.1 -c pytorch -c nvidia

启动 python -m vllm.entrypoints.openai.api_server --tensor-parallel-size=1 --trust-remote-code --max-model-len 1024 --model THUDM/chatglm3-6b

python -m vllm.entrypoints.openai.api_server --host 127.0.0.1 --port 8101 --tensor-parallel-size=1 --trust-remote-code --max-model-len 1024 --model THUDM/chatglm3-6b

测试 curl http://127.0.0.1:8101/v1/completions
-H "Content-Type: application/json"
-d '{ "model": "THUDM/chatglm3-6b", "prompt": "请用20字内回复我,你今年多大了", "max_tokens": 20, "temperature": 0 }'

多轮对话 curl -X POST "http://127.0.0.1:8101/v1/completions"
-H "Content-Type: application/json"
-d "{"model": "THUDM/chatglm3-6b","prompt": "你叫什么名字", "history": [{"role": "user", "content": "你出生在哪里."}, {"role": "assistant", "content": "出生在北京"}]}"

多轮对话 curl -X POST "http://127.0.0.1:8101/v1/chat/completions"
-H "Content-Type: application/json"
-d "{"model": "THUDM/chatglm3-6b", "messages": [{"role": "system", "content": "You are ChatGLM3, a large language model trained by Zhipu.AI. Follow the user's instructions carefully. Respond using markdown."}, {"role": "user", "content": "你好，给我讲一个故事，大概100字"}], "stream": false, "max_tokens": 100, "temperature": 0.8, "top_p": 0.8}"

参考文档：https://docs.vllm.ai/en/latest/