FastChat：开源 LLM 对话机器人平台 2026 版

pip install fschat

python -m fastchat.serve.controller

python -m fastchat.serve.model_worker \
    --model-path /models/vicuna-7b \
    --controller http://localhost:21001

python -m fastchat.serve.openai_api_server \
    --controller http://localhost:21001 \
    --model-list vicuna-7b,gpt-4

python -m fastchat.serve.gradio_web_server

import openai

openai.api_base = "http://localhost:8000/v1"

response = openai.ChatCompletion.create(
    model="vicuna-7b",
    messages=[{"role": "user", "content": "你好"}]
)

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "vicuna-7b", "messages": [{"role": "user", "content": "解释一下机器学习"}]}'

python -m fastchat.serve.model_worker \
    --model-path /models/vicuna-13b \
    --model-names vicuna-13b,vicuna-13b-chat

# 使用 vLLM 高效推理
python -m fastchat.serve.model_worker \
    --model-path /models/vicuna-7b \
    --load-momentum vllm

# 多卡推理
export CUDA_VISIBLE_DEVICES=0,1,2,3
python -m fastchat.serve.model_worker --num-gpus 4

FROM lmstudio/fastchat:latest
COPY models/ /models/
CMD ["python", "-m", "fastchat.serve.controller"]

apiVersion: apps/v1
kind: Deployment
metadata:
  name: fastchat-worker
spec:
  replicas: 3
  template:
    spec:
      containers:
      - name: worker
        image: lmstudio/fastchat
        args: ["python", "-m", "fastchat.serve.model_worker"]