ok - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 第51页 - 钛刻科技

[程序员] 自建 GPT API 分享

** 登录就送 2 千万 Token ，可用 gpt-5.5 ** https://wgkai.top/

相关专题

tech v2ex.com 2026-04-26 09:14:09+08:00

[Local LLM] 各位推荐一个 32G Macbook air M5 可以跑的 moe 模型

27B/31B 甚至 35B 的 4bit 都可以, 测试了好久, 也下载了几十个了,都不太行, 感觉降智了, 这些刚出来的时候我这个配置能跑到 35tokens/s. 准备直接抄作业, 请给 huggingface 连接, 我的本地推理框架是 omlx, 感谢感谢.

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech v2ex.com 2026-04-26 08:56:08+08:00

[Local LLM] 各位推荐一个 32G Macbook air M5 可以跑的 moe 模型

27B/31B 甚至 35B 的 4bit 都可以, 测试了好久, 也下载了几十个了,都不太行, 感觉降智了, 这些刚出来的时候我这个配置能跑到 35tokens/s. 准备直接抄作业, 请给 huggingface 连接, 我的本地推理框架是 omlx, 感谢感谢.

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech v2ex.com 2026-04-26 08:56:08+08:00

token好烧啊说是 :distorted_face: 烧不动了

现已被封印…并且cc对于deepseek的思维链回传没做…开发体验要起飞了好看是好看，好玩是好玩，token好烧啊… 9 个帖子 - 4 位参与者阅读完整话题

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech linux.do 2026-04-26 08:49:26+08:00

你们一个月上班要用多少 token?

如题，我上个月用了 200M ，主要是 GPT5.4 和 Gemini3.1Pro

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 08:27:52+08:00

你们一个月上班要用多少 token?

如题，我上个月用了 200M ，主要是 GPT5.4 和 Gemini3.1Pro

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 08:27:52+08:00

各位推荐一个 32G Macbook air M5 可以跑的 moe 模型

27B/31B 甚至 35B 的 4bit 都可以, 测试了好久, 也下载了几十个了,都不太行, 感觉降智了, 这些刚出来的时候我这个配置能跑到 35tokens/s. 准备直接抄作业, 请给 huggingface 连接, 我的本地推理框架是 omlx, 感谢感谢.

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 08:22:11+08:00

为了买到便宜靠谱的 Token，少被割韭菜，我做了个比价的网站

最近在开发一个小项目，想买几个 ChatGPT 账号。但是在多个卡网和电报群里转了几圈，发现同一种类型的账号，比如 ChatGPT plus 月卡，就有 N 多种价格。有的几块钱，有的卖 15 ，有的卖 30 ，有的卖 40. 后来我才知道，这些店铺和卖家的来源其实就是那么几

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 08:18:57+08:00

你们一个月上班要用多少 token?

如题，我上个月用了 200M ，主要是 GPT5.4 和 Gemini3.1Pro

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 07:27:52+08:00

为了买到便宜靠谱的 Token，少被割韭菜，我做了个比价的网站

最近在开发一个小项目，想买几个 ChatGPT 账号。但是在多个卡网和电报群里转了几圈，发现同一种类型的账号，比如 ChatGPT plus 月卡，就有 N 多种价格。有的几块钱，有的卖 15 ，有的卖 30 ，有的卖 40. 后来我才知道，这些店铺和卖家的来源其实就是那么几

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 07:18:57+08:00

为了买到便宜靠谱的 Token，少被割韭菜，我做了个比价的网站

最近在开发一个小项目，想买几个 ChatGPT 账号。但是在多个卡网和电报群里转了几圈，发现同一种类型的账号，比如 ChatGPT plus 月卡，就有 N 多种价格。有的几块钱，有的卖 15 ，有的卖 30 ，有的卖 40. 后来我才知道，这些店铺和卖家的来源其实就是那么几

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 07:18:57+08:00

你们一个月上班要用多少 token?

如题，我上个月用了 200M ，主要是 GPT5.4 和 Gemini3.1Pro

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 06:27:52+08:00

为了买到便宜靠谱的 Token，少被割韭菜，我做了个比价的网站

最近在开发一个小项目，想买几个 ChatGPT 账号。但是在多个卡网和电报群里转了几圈，发现同一种类型的账号，比如 ChatGPT plus 月卡，就有 N 多种价格。有的几块钱，有的卖 15 ，有的卖 30 ，有的卖 40. 后来我才知道，这些店铺和卖家的来源其实就是那么几

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 06:18:57+08:00

分享一个机场

27年4月到期，还有1tb https://s-dywrwizazu.cn-shanghai.fcapp.run/okz/sub?token=487099b89235ad9fe234908444d2d6e5 1 个帖子 - 1 位参与者阅读完整话题

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech linux.do 2026-04-26 06:17:13+08:00

你们一个月上班要用多少 token?

如题，我上个月用了 200M ，主要是 GPT5.4 和 Gemini3.1Pro

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 05:27:52+08:00

为了买到便宜靠谱的 Token，少被割韭菜，我做了个比价的网站

最近在开发一个小项目，想买几个 ChatGPT 账号。但是在多个卡网和电报群里转了几圈，发现同一种类型的账号，比如 ChatGPT plus 月卡，就有 N 多种价格。有的几块钱，有的卖 15 ，有的卖 30 ，有的卖 40. 后来我才知道，这些店铺和卖家的来源其实就是那么几

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 05:18:57+08:00

你们一个月上班要用多少 token?

如题，我上个月用了 200M ，主要是 GPT5.4 和 Gemini3.1Pro

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 04:27:52+08:00

为了买到便宜靠谱的 Token，少被割韭菜，我做了个比价的网站

最近在开发一个小项目，想买几个 ChatGPT 账号。但是在多个卡网和电报群里转了几圈，发现同一种类型的账号，比如 ChatGPT plus 月卡，就有 N 多种价格。有的几块钱，有的卖 15 ，有的卖 30 ，有的卖 40. 后来我才知道，这些店铺和卖家的来源其实就是那么几

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 04:18:57+08:00

我做了个工具让 8GB 显卡跑 30B 模型从 3 tok/s 提到 21 tok/s，记录一下技术发现

最近在折腾本地大模型，发现一个核心问题：Ollama 和 LM Studio 能让模型跑起来，但参数全靠猜——上下文长度、KV cache 类型、MoE expert 放哪、ubatch 多大……用默认参数基本是在浪费显卡。于是做了个工具自动找最优配置，过程中踩了不少坑，记录一

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 04:07:34+08:00

我做了个工具让 8GB 显卡跑 30B 模型从 3 tok/s 提到 21 tok/s，记录一下技术发现

最近在折腾本地大模型，发现一个核心问题：Ollama 和 LM Studio 能让模型跑起来，但参数全靠猜——上下文长度、KV cache 类型、MoE expert 放哪、ubatch 多大……用默认参数基本是在浪费显卡。于是做了个工具自动找最优配置，过程中踩了不少坑，记录一

相关专题

Education App Health Software Design Technology Training Repo...Seminar Revenue 专题内容 Desktop Market Careers Blog Fitness Message About Software UR...Mobile 专题内容 Alliance Partner Forecast Reporting Support Customer Identity...Saving Deadline Products Web Services Project Faq Data Suppor...Alert Sales Workshop Local Quality Web Advertising 专题内容 Study Alert Experience Network 专题内容 Sport Blog Cheap Community 专题内容 Analytics App Dashboard Spreadsheet Education Business 专题内容 Support Objective Follow Alert Efficiency Team Rating 专题内容 Desktop Workshop 专题内容 User Machine Reporting Mobile Template Community Ranking Navi...Landing Traffic Email Schedule Mobile Productivity Blog Platf...Solution Button Mobile Affordable Technology Logo Budget Team...Customer Education Investment Behavior Sale 专题内容 Affordable 专题内容 Objective Digital API 专题内容 Server Platform Hosting Digital 视频 Security Internet 专题内容 Saving Web Folder Support Local Presentation Register 专题内容

tech www.v2ex.com 2026-04-26 04:07:34+08:00

ok - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 第51页 - 钛刻科技 | TCTI.cn

相关标签