0. 需求說(shuō)明

自用筆記本構(gòu)建一套開發(fā)環(huán)境，用docker 虛擬插件 dev containers,實(shí)現(xiàn)開發(fā)環(huán)境的構(gòu)建，我想構(gòu)建一套LLMs的環(huán)境，由于環(huán)境配置太多，不想污染本地環(huán)境，所以選擇隔離技術(shù)

1. 環(huán)境準(zhǔn)備

• vscode
• docker

2. 步驟

2.1 vscode 安裝插件

• Dev Containers
• Remote Development

2.2 項(xiàng)目根目錄下創(chuàng)建.devcontainer文件

目錄結(jié)構(gòu)

2.2.1 直接用鏡像構(gòu)建環(huán)境

devcontainers.json
VS Code Dev Containers官方文檔
字段說(shuō)明【具體字段參見官方網(wǎng)站字段描述】：
1.name：dev container 名稱
2.image：鏡像名稱
3. customizations：在支持的工具中定義
4. build：dockerfile 構(gòu)建時(shí)需要配置
5. context：構(gòu)建的位置上下文，默認(rèn). 當(dāng)前目錄，.. 表示上級(jí)目錄
6. dockerfile: dockerfile 文件

{"name": "LLMs From Scratch","image": "llms-from-scratch:0.1","customizations": {"vscode": {"extensions": ["ms-python.python","ms-azuretools.vscode-docker","ms-toolsai.jupyter","yahyabatulu.vscode-markdown-alert","tomoki1207.pdf","mechatroner.rainbow-csv"]}}
}

2.2.2 用Dockefile構(gòu)建鏡像

devcontainers.json

{"name": "LLMs From Scratch","build": {"context": "..","dockerfile": "Dockerfile"},"customizations": {"vscode": {"extensions": ["ms-python.python","ms-azuretools.vscode-docker","ms-toolsai.jupyter","yahyabatulu.vscode-markdown-alert","tomoki1207.pdf","mechatroner.rainbow-csv"]}}
}

Dockerfile

FROM pytorch/pytorch:2.0.1-cuda11.7-cudnn8-runtimeRUN apt-get update && \apt-get upgrade -y && \apt-get install -y rsync && \apt-get install -y git && \apt-get install -y curl && \rm -rf /var/lib/apt/lists/*COPY requirements.txt requirements.txt
RUN pip install --no-cache-dir -r requirements.txt

requirements 環(huán)境需要的配置信息

torch >= 2.0.1       
jupyterlab >= 4.0    
tiktoken >= 0.5.1    
matplotlib >= 3.7.1  
tensorflow >= 2.15.0 
tqdm >= 4.66.1       
numpy >= 1.25, < 2.0  
pandas >= 2.2.1       
psutil >= 5.9.5     

3.啟動(dòng)環(huán)境

Command + Shift + P
輸入：Dev Containers: Rebuild and Reopen in Container

啟動(dòng)成功截圖

4. 關(guān)閉環(huán)境

File —>close remote connection