惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Hugging Face - Blog
Hugging Face - Blog
Microsoft Azure Blog
Microsoft Azure Blog
月光博客
月光博客
S
Securelist
J
Java Code Geeks
Recorded Future
Recorded Future
Exploit-DB.com RSS Feed
Exploit-DB.com RSS Feed
K
KPMG report finds enterprise disconnect between AI and its ROI | CIO
M
MIT News - Artificial intelligence
S
Secure Thoughts
Y
Y Combinator Blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
D
Docker
Martin Fowler
Martin Fowler
The Last Watchdog
The Last Watchdog
WordPress大学
WordPress大学
The GitHub Blog
The GitHub Blog
Vercel News
Vercel News
O
OpenAI News
www.infosecurity-magazine.com
www.infosecurity-magazine.com
博客园_首页
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
PCI Perspectives
PCI Perspectives
N
News and Events Feed by Topic
H
Heimdal Security Blog
SecWiki News
SecWiki News
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
博客园 - 【当耐特】
T
Troy Hunt's Blog
L
LINUX DO - 最新话题
Hacker News: Ask HN
Hacker News: Ask HN
Hacker News - Newest:
Hacker News - Newest: "LLM"
N
Netflix TechBlog - Medium
A
Arctic Wolf
The Hacker News
The Hacker News
I
Intezer
S
Schneier on Security
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
Apple Machine Learning Research
Apple Machine Learning Research
L
Lohrmann on Cybersecurity
宝玉的分享
宝玉的分享
P
Privacy & Cybersecurity Law Blog
Stack Overflow Blog
Stack Overflow Blog
T
Tor Project blog
小众软件
小众软件
Simon Willison's Weblog
Simon Willison's Weblog
The Cloudflare Blog
Jina AI
Jina AI

StudyingLover's Blog

Diffusion Policy笔记 rwkv笔记 act笔记 nanovllm-block_manager opencode多智能体 nanobot-pre-train nanobot-rl nanobot-sft nanobot-checkpoint_manager nanobot-gpt nanobot-mid-train Vision Mamba (Vim)笔记 BPE演示 最后一遍学习Transformer YOLOv5 目标检测笔记 下载根服务器解析记录 Dynaseal A Backend-Controlled LLM API Key Distribution Scheme with Constrained Invocation Parameters 判断链表有环 王道25数据结构勘误 关于perplexity的open-sourcing-r1-1776 AI为什么不像人类一样进行多轮对话 新博客改造日记和功能测试 linuxqq只显示登陆背景图 数字设计和计算机体系结构(机械工业出版社)勘误(自制) Dynaseal:面向未来端侧llm agent的llm api key分发机制 A Definitive Guide to Markdown Style This post is using MDX, Where you can embed JSX and Astro components RT-Patch学习 pydantic实现的LLM ReAct fastapi 和 uvicorn 设置监听 ipv6 pydantic+openai+json 控制大模型输出的最佳范式 解决 Matplotlib Scatter 不支持 Marker 列表的问题:mscatter 实现 roofline model zhipuAI接口兼容openai 在docker部署fastapi宝塔里使用nginx反代套上cloudflare获取请求的真实ip clion搭建libbpf-bootstrap开发环境 coze+coze-discord-proxy+ChatNextWebUI实现AI自由 安卓内核时间使用的是UTC时间 colab运行google最新开源模型Gemma Sora技术报告 视频生成模型作为世界模拟器 笔记 archlinux flutter开发踩坑 fastapi集成google auth登录 linux下NTFS磁盘报错输入输出错误 Venn-Abers 预测器 基于Venn-Abers预测器的系统日志异常检测方法_顾兆军 手机平板远程访问kvm虚拟机的windows phi-2弱智吧测评 poe的gemini pro或是百度开发 google gemini api使用 google gemini api申请 构建用于复杂数据处理的高效UDP服务器和客户端 matplotlib中文字体渲染 TruFor笔记和代码复现 深入分析:GitHub Trending 项目 "multipleWindow3dScene" pua大模型 ggml教程|mnist手写体识别量化推理 xgboost2.0最佳实践 xgboost使用GPU最佳实践 马踏棋盘 cloudlflare推理llama2 docker搭建elasticsearch并使用python连接 FreeU-文字生成图片的免费午餐笔记 使用xgboost的c接口推理模型 Archlinux使用CMake调用xgboost的c接口 m2cgen生成机器学习c语言推理代码 xgboost模型序列化存储并推理 speculative-sampling笔记 prompt2model笔记 RoboTAP笔记 自建obsidian同步服务 MediaPipe即将推出图像生成服务 Dual-Stream Diffusion Net for Text-to-Video Generation笔记 ViT在DDPM取代UNet(DiT) arch4edu搞崩了我的flutter LISA(推理分割)笔记 在终端绘制GPU显存使用曲线 GPTBot介绍 arch蓝牙无法连接 GPU部署llama-cpp-python(llama.cpp通用) 花式求GCD 使用llama构建一个蜜罐(前端) 使用llama构建一个蜜罐(后端) llama-cpp-python快速上手 Paper Gestalt笔记 DINO-v2笔记 快速上手llama2.c AnyDoor笔记 Archlinux安装scrcpy加载共享库出错 error while loading shared libraries:libusb-1.0.so.0:wrong ELF class:ELFCLASS32 npc_gzip笔记 python调用c++函数 Filesystem type ntfs3,ntfs not configured in kernel open_clip编码图像和文本 PicGo配置CloudflareR2图片储存 ArchlinuxGnome快捷键打开终端 clip-interrogator代码解析 GroundingDINO安装报错解决 2023华为鲲鹏畅想日暨西安高新国际会议中心零食午饭测评 RoboMaster开源仓库汇总(长期更新) 没有手都可以在腾讯云创建镜像 I3D笔记
快速上手llama2.c(更新版)
About the Author StudyingLover · 2023-07-29 · via StudyingLover's Blog

快速上手llama2.c(更新版)

在上一次我同时在我的博客和知乎发布了快速上手llama2.c 之后,我一个小透明也收获了不少收藏,并收到了人生中第一个这样的留言(其实我感觉是机器人)。

当然,之前的llama2.c也有一些不好的地方,例如不能添加自己的prompt,所以我提了这样的一个issue,今天收到了贡献者的回复说是可以用了。那我们来看一下。

首先还是克隆整个仓库,编译并下载模型,这里以15m参数的模型作为示例

git clone https://github.com/karpathy/llama2.c.git
cd llama2.c
make run
wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories15M.bin

接下来我们就可以使用编译出来的run 运行了,要使用自己的prompt,需要指定温度和 步长,这里温度设置成1.0,步长设置256,prompt在双引号写,我这里写的是One day morning , I don't want to go to school .

./run stories15M.bin 1.0 256 "One day morning , I don't want to go to school"

这里给出我的运行结果,也就3秒种不到

<s>
One day morning , I don't want to go to school, so he packed his trunk lid to pack. memorized his chores, he thought about what his mom would like him to stay home and not do all day. She wanted him to in a very competitive way.
"Come and play in the puddle, it'll be more fun!"He begged.
Mom shook her head. "No, we haven't seen coming for sure," she said thought. 
Thumper and Mom just shrugged.
"See," she said. "Come on now. Let's go and find some fun ways to clean the world!"
The little boy was relieved and ran out to the yard. He had found a great idea to share his day with his mom instead. They scattered around the yard and had fun playing until their tired eyes were aching.
<s>
Once upon a time, there was a little boy named Tim. Tim was very excited because he was going on a trip with his family. He saw a big bus that helped them get off at their destination.
As the bus drove along, Tim noticed an unusual looking man sitting next to it. Tim asked the
achieved tok/s: 175.378267

当然为了获得更好的效果,我们可以使用更大模型

下载42m参数模型

wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories42M.bin

下载110m参数模型

wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories110M.bin