docker 本地部署大模型(ollama)

docker 安装 ollama

复制代码
docker search ollama
docker pull ollama/ollama

###docker下载ollama部署
docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama

### 下载模型
docker exec -it ollama ollama pull llama3

### 交互式运行模型
 docker exec -it ollama ollama run llama3

注意有坑 :docker -v ollama:/root/.ollama

此处由于挂载目录使用了相对路径,所以本地文件夹位于 /var/lib/docker/volumes/ollama

而非运行命令的相对路径

测试 api

复制代码
$ curl http://localhost:14444/api/generate -d '{ "model": "llama3", "prompt": "請列出五樣台灣美食", "stream": true, "options": { "seed": 123, "top_k": 20, "top_p": 0.9, "temperature": 0 } }'
{"model":"llama3","created_at":"2024-07-08T03:49:16.611142384Z","response":"Here","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:16.758150085Z","response":" are","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:16.903960154Z","response":" ","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.048831268Z","response":"5","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.192653835Z","response":" must","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.340323632Z","response":"-","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.490090786Z","response":"try","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.630377981Z","response":" Taiwanese","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.79671948Z","response":" dishes","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:17.947959672Z","response":":\n\n","done":false}
.......
{"model":"llama3","created_at":"2024-07-08T03:49:51.670524112Z","response":" dishes","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:51.80466663Z","response":" showcase","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:51.953821627Z","response":" the","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.112141942Z","response":" diversity","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.274051058Z","response":" and","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.41761068Z","response":" richness","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.571736412Z","response":" of","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.719734804Z","response":" Taiwanese","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.857012216Z","response":" cuisine","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:52.998817903Z","response":",","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:53.166612385Z","response":" which","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:53.342996684Z","response":" often","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:53.510572997Z","response":" combines","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:53.668246714Z","response":" Chinese","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:53.836925764Z","response":",","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:53.986648055Z","response":" Japanese","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:54.13517455Z","response":",","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:54.283114108Z","response":" and","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:54.443445171Z","response":" indigenous","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:54.603118958Z","response":" influences","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:54.76832015Z","response":".","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:54.931017326Z","response":" Enjoy","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:55.083288072Z","response":"!","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:55.533841685Z","response":" 🍴","done":false}
{"model":"llama3","created_at":"2024-07-08T03:49:55.671222985Z","response":"","done":true,"done_reason":"stop","context":[128006,882,128007,271,109311,32218,20834,76208,112595,117342,58666,102456,128009,128006,78191,128007,271,8586,527,220,20,2011,12,1568,94037,26863,1473,16,13,3146,3513,830,452,31707,61434,1754,96618,362,11670,94037,12269,11,25309,46895,273,19724,374,264,82651,323,94631,45993,1903,449,6435,23283,11059,25309,11,61260,11,323,24822,627,17,13,3146,46,22604,507,2727,1169,1754,96618,362,5526,8761,3691,304,29389,11,297,22604,297,2727,10145,527,73624,41951,19335,10409,449,628,1538,297,100163,11,70181,11,323,6307,45697,13,328,2841,449,264,10437,323,94760,19737,627,18,13,3146,626,39134,350,1073,84,1754,96618,49386,369,1202,281,2234,306,22843,11,357,39134,80247,374,264,96080,29839,17937,2027,430,596,5655,2269,4588,311,39143,13,1102,596,3629,10434,449,264,86921,19737,323,649,387,5115,57407,4999,19,13,3146,33,93613,85034,97685,1754,96618,362,69130,323,9257,12269,11,20716,4147,36167,36517,374,6435,23283,11059,304,264,21655,315,29839,19737,11,13465,11,323,49659,3156,279,13339,374,28682,323,17503,10980,6847,627,20,13,3146,38,4381,426,3524,1754,96618,7429,3967,439,94037,63452,11,1709,64,103253,527,4179,3690,293,11099,10409,449,20716,4147,36167,11,3820,839,24822,11,323,42393,13,362,18406,323,37154,40459,477,15496,382,9673,26863,35883,279,20057,323,90030,315,94037,36105,11,902,3629,33511,8620,11,11002,11,323,39678,34453,13,23485,0,11410,235,112,128009],"total_duration":43461066356,"load_duration":3482212558,"prompt_eval_count":17,"prompt_eval_duration":916597000,"eval_count":260,"eval_duration":39060126000}

python + ollama

py 复制代码
from ollama import Client 
client = Client(host=f'http://{IP_ADDRESS}:14444') 
response = client.chat(model='llama3', messages=[ { 'role': 'user', 'content': 'Why is the sky blue?', }, ])
response 
复制代码
{'model': 'llama3',
 'created_at': '2024-07-08T04:05:41.013483539Z',
 'message': {'role': 'assistant',
  'content': "The sky appears blue because of a phenomenon called Rayleigh scattering, which is named after the British physicist Lord Rayleigh. Here's what happens:\n\n1. **Sunlight**: When sunlight enters Earth's atmosphere, it contains all the colors of the visible spectrum (red, orange, yellow, green, blue, indigo, and violet).\n2. **Molecules**: The atmosphere is made up of tiny molecules of gases like nitrogen (N2) and oxygen (O2). These molecules are much smaller than the wavelength of light.\n3. **Scattering**: When sunlight hits these small molecules, it scatters in all directions. This scattering effect is more pronounced for shorter wavelengths, like blue and violet light, which have a higher frequency than longer wavelengths, like red and orange light.\n4. **Blue dominance**: As a result of this scattering, the blue light with its shorter wavelength is dispersed throughout the atmosphere, reaching our eyes from all parts of the sky. This is why the sky appears blue during the daytime, especially when the sun is overhead.\n\nThe same principle applies to other colors, but with different intensities:\n\n* **Red and orange**: These longer wavelengths are less affected by scattering and reach our eyes more directly, which is why we see them as dominant at sunrise and sunset.\n* **Green and yellow**: These wavelengths are scattered similarly to blue light, but their intensity decreases with distance from the sun, making them less visible in the sky.\n\nSo, to summarize: the sky appears blue because of the scattering of sunlight by tiny molecules in the atmosphere, which favors shorter wavelengths like blue and violet."},
 'done_reason': 'stop',
 'done': True,
 'total_duration': 52984861518,
 'load_duration': 3257720984,
 'prompt_eval_count': 15,
 'prompt_eval_duration': 829317000,
 'eval_count': 325,
 'eval_duration': 4888172200

流式输出

py 复制代码
import ollama

stream = client.chat(
    model='llama3',
    messages=[{'role': 'user', 'content': 'Why is the sky blue?'}],
    stream=True,
)

for chunk in stream:
    print(chunk['message']['content'], end='', flush=True)

参考

相关推荐
pride.li8 分钟前
第03章 Python 教程-基础语法:注释与变量
服务器·前端·后端·python·算法·计算机视觉
databook11 分钟前
用 Pydantic AI 从零搭建一个能干活的 AI Agent
人工智能·python
康小旺31 分钟前
把 markitdown 做成 Windows 桌面版:拖进文档就出 Markdown,附 233MB→64MB 打包裁剪全过程
python
程序员老赵39 分钟前
Docker 部署 Wiki.js:轻松搭建私有化维基知识库
docker·容器·开源
用户212968352616843 分钟前
开源了一个 Java 多智能体 LLM 框架 Baton:一个依赖,把 Agent 循环、工具调用、记忆、编排全配齐
java·llm
程序员大狐狸1 小时前
Ultra X7 358H 笔记本电脑,本地三个大模型编程能力、速度、上下文全面对比与选型报告
llm·ai编程
Skylar__1 小时前
从 Beam 到 Muse:一个开发者视角的 AI 模型选型决策地图
llm
启观川1 小时前
数据结构与算法-第 3 章 常用算法-贪心算法
数据结构·笔记·python·算法·贪心算法
南笙北梦1 小时前
本地开源大模型多轮对话显存泄漏排查与修复实录
python
万联WANFLOW2 小时前
LLM 调用很快,AI Agent 为什么还是慢?关键路径分析与优化方法
性能优化·llm·aigc