一、DeepSeek安装1、首先安装ollamahttps://ollama.com/download/linuxwindows安装1下载地址https://ollama.com/download/linux2.通过网盘分享的文件windows链接: https://pan.baidu.com/s/1B14r_HiOg57V2FJ73TnYRQ 提取码: x82slinux安装在线安装curl -fsSL https://ollama.com/install.sh | sh离线安装链接: https://pan.baidu.com/s/10fCl_mYK2_VMSmLVhSjgeQ 提取码: hk69获取有两个文件一个是ollama-linux-amd64.tgz另一个是安装脚本install_ollama.sh2.将两个文件上传到linux服务器放于同一文件夹下3.执行install_ollama.sh文件chmod x install_ollama.sh #给脚本赋予执行权限./install_ollama.sh# 如果报错误 bash: ./build_android.sh/bin/sh^M解释器错误: 没有那个文件或目录执行下面命令后再执行./install_ollama.sh命令sed -i s/\r$// install_ollama.sh3.配置ollamavim /etc/systemd/system/ollama.service[Unit] DescriptionOllama Service Afternetwork-online.target [Service] ExecStart/usr/local/bin/ollama serve Userollama Groupollama Restartalways RestartSec3 EnvironmentPATH/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/root/bin:/usr/local/bin EnvironmentOLLAMA_HOST0.0.0.0:11434 EnvironmentOLLAMA_ORIGINS* [Install] WantedBydefault.target修改后重载服务文件重启ollamasudosystemctl daemon-reloadsudosystemctl restart ollama查看ollama运行状态sudosystemctl status ollama.service4.防火墙配置开放11434端口sudofirewall-cmd--permanent--zonepublic --add-port11434/tcp重新加载sudofirewall-cmd--reload5.访问测试使用其他计算机访问http:/服务器域名11434若返回Ollama is running说明可成功访问2、DeepSeelk部署版本和基础环境以下是整理的DeepSeek-R1 系列模型硬件需求及成本概览表模型名称参数量CPU核心内存DDR4存储M2 SSDGPU 推荐纯CPU推理速度DeepSeek-R1-Distill-Qwen-1.5B1.5B416GB256GB可纯CPU / 可选8GB GPU~20 tokens/sDeepSeek-R1-Distill-Qwen-7B7B832GB256GB可纯CPU / 推荐8GB GPU~5 tokens/sDeepSeek-R1-Distill-Llama-8B8B832GB256GB可纯CPU / 推荐8GB GPU~5 tokens/sDeepSeek-R1-Distill-Qwen-14B14B1264GB512GB可纯CPU / 推荐16GB GPU~0.5 tokens/sDeepSeek-R1-Distill-Qwen-32B32B1696GB512GBA10/A16 / 推荐A100极慢未测DeepSeek-R1-Distill-Llama-70B70B20128GB1TBA100×2 / H100极慢未测DeepSeek-R1-671B671B48768GB (推荐1.5TB)2TBA100×8 (显存≥1000GB)需GPU集群安装deepseekDeepSeek-R1-Distill-Qwen-1.5Bollama run deepseek-r1:1.5bDeepSeek-R1-Distill-Qwen-7Bollama run deepseek-r1:7bDeepSeek-R1-Distill-Llama-8Bollama run deepseek-r1:8bDeepSeek-R1-Distill-Qwen-14Bollama run deepseek-r1:14bDeepSeek-R1-Distill-Qwen-32Bollama run deepseek-r1:32bDeepSeek-R1-Distill-Llama-70Bollama run deepseek-r1:70b下载完成后则可通过终端窗口 进行提问输入 ai是什么 按回确认得到如下回复3、下载chatboxai 实现可视化chatbox 打开后点击设置如下图选择模型至此 chatbox 配置完成。本地测试DeepSeek打开chatbox 输入ai是什么点击发送如下图
