InternLM 腾讯云本地部署指南：Ollama 到 LMDeploy 双路径

1sudo apt-get update
2sudo apt-get install -y ca-certificates curl gnupg lsb-release
3
4sudo install -m 0755 -d /etc/apt/keyrings
5curl -fsSL https://mirrors.cloud.tencent.com/docker-ce/linux/ubuntu/gpg \
6  | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg
7sudo chmod a+r /etc/apt/keyrings/docker.gpg
8
9echo \
10  "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] \
11  https://mirrors.cloud.tencent.com/docker-ce/linux/ubuntu \
12  $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
13
14sudo apt-get update
15sudo apt-get install -y docker-ce docker-ce-cli containerd.io docker-compose-plugin
16
17sudo usermod -aG docker $USER
18newgrp docker

1sudo mkdir -p /etc/docker
2sudo tee /etc/docker/daemon.json <<'EOF'
3{
4  "registry-mirrors": [
5    "https://你的专属地址.mirror.swr.myqcloud.com"
6  ]
7}
8EOF
9sudo systemctl daemon-reload
10sudo systemctl restart docker

1df -h /
2free -h

1docker run -d \
2  --name ollama \
3  --restart unless-stopped \
4  -p 11434:11434 \
5  -v /data/ollama:/root/.ollama \
6  ollama/ollama:latest
7
8curl http://127.0.0.1:11434/api/tags

1docker exec -it ollama ollama pull internlm2:7b

1docker logs -f ollama

1docker exec -it ollama ollama run internlm2:7b "用三句话介绍上海人工智能实验室"

1curl http://127.0.0.1:11434/v1/chat/completions \
2  -H "Content-Type: application/json" \
3  -d '{
4    "model": "internlm2:7b",
5    "messages": [{"role": "user", "content": "总结一下长文本模型的应用场景"}],
6    "temperature": 0.7
7  }'

1# 方式一：启动时用环境变量指定
2docker run -d \
3  --name ollama \
4  --restart unless-stopped \
5  -p 11434:11434 \
6  -v /data/ollama:/root/.ollama \
7  -e OLLAMA_CONTEXT_LENGTH=32768 \
8  ollama/ollama:latest

1// 方式二：请求里带 num_ctx（部分版本支持）
2{
3  "model": "internlm2:7b",
4  "messages": [...],
5  "options": { "num_ctx": 32768 }
6}

1pip install modelscope
2
3python -c "
4from modelscope import snapshot_download
5snapshot_download('Shanghai_AI_Laboratory/internlm2-chat-7b', cache_dir='/data/models')
6"

1export HF_ENDPOINT=https://hf-mirror.com
2huggingface-cli download internlm/internlm2-chat-7b \
3  --local-dir /data/models/internlm2-chat-7b

1docker run -d \
2  --name lmdeploy \
3  --restart unless-stopped \
4  --gpus all \
5  -p 23333:23333 \
6  -v /data/models:/models \
7  --ipc=host \
8  openmmlab/lmdeploy:latest \
9  lmdeploy serve api_server /models/internlm2-chat-7b \
10    --server-port 23333 \
11    --model-name internlm2 \
12    --session-len 32768

1curl http://127.0.0.1:23333/v1/models
2
3curl http://127.0.0.1:23333/v1/chat/completions \
4  -H "Content-Type: application/json" \
5  -d '{
6    "model": "internlm2",
7    "messages": [{"role": "user", "content": "你好，做个自我介绍"}]
8  }'

1python - <<'PY'
2import requests
3# 造一个长文本：重复拼接，模拟长文档
4doc = ("本公司年假制度规定：入职满一年可享 5 天年假，每满一年增加 1 天，上限 15 天。" * 400)
5r = requests.post("http://127.0.0.1:23333/v1/chat/completions", json={
6    "model": "internlm2",
7    "messages": [
8        {"role": "system", "content": "你是制度问答助手，只依据给定文档回答。"},
9        {"role": "user", "content": f"文档内容：\n{doc}\n\n问题：年假上限是多少天？"}
10    ],
11    "max_tokens": 200
12}, timeout=300)
13print(r.json()["choices"][0]["message"]["content"])
14PY

1docker run -d \
2  --name open-webui \
3  --restart unless-stopped \
4  -p 3000:8080 \
5  -e OLLAMA_BASE_URL=http://host.docker.internal:11434 \
6  --add-host=host.docker.internal:host-gateway \
7  -v /data/open-webui:/app/backend/data \
8  ghcr.io/open-webui/open-webui:main

1# 只监听本机
2docker run -d ... -p 127.0.0.1:11434:11434 ollama/ollama:latest