CA - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 第28页 - 钛刻科技

[Local LLM] 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

tech v2ex.com 2026-04-25 03:39:55+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-25 02:39:55+08:00

分享一个咸鱼买的 GPT

还有65刀，可以用 GPT-5.5 https://api.cst9.com sk-cc1a07b67c96b2857b1765b1c891ef22939de14cae0f1ff32387100e2aa5376b 1 个帖子 - 1 位参与者阅读完整话题

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-25 02:23:28+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-25 02:18:07+08:00

【深夜福利】免费送 GLM Coding Plan 体验卡 - 自取

剩余 3/3 张 https://bigmodel.cn/activity/trial-card/1BMR381KZD 3 个帖子 - 3 位参与者阅读完整话题

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-25 02:08:51+08:00

Glm4flash免费用户并发调整成20

如题之前是200来着现在改成20了沉浸式翻译现在我已经转longcat flash了并发20好像也没问题不过我也不确定这是不是快讯我是今天才发现的 4 个帖子 - 3 位参与者阅读完整话题

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-25 01:50:50+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-25 01:40:43+08:00

CPA 内网ip访问 webUI 登录失败: 访问被拒绝，权限不足

docker部署,官网教程部署成功 localhost访问没问题内网ip访问则报标题错误(无论部署机还是其他机) 配置方面: 确认密码没错确认开启了 allow-remote host为 0.0.0.0 localhost进行活动,docker能看到访问日志. ip访问日志都

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-25 00:57:46+08:00

被标记为网络安全风险有啥影响不好怕被封号

此聊天已被标记为可能存在网络安全风险背景：我想codex帮我分析 GitHub - TheMythologist/GenP: Automatically builds GenP executables from source · GitHub 这个项目结果直接报错让我加入什么

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-25 00:29:31+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 23:54:14+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 23:54:14+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 23:54:14+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 23:47:32+08:00

cpa401清洗后存活率9%

Scan Summary checked codex files : 486 unauthorized (401) : 443 quota-exceeded : 0 no-limit/unlimited : 0 non-2xx or errors : 486 #后面发现gpt-5

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-24 23:26:49+08:00

deepseek v4 碎碎念

1、CSA和HCA两种sparse attention 2、mHC 3、Muon优化器 4、GRPO+OPD 想问一下这次的engram去哪里了，没办法scale up？ 1 个帖子 - 1 位参与者阅读完整话题

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-24 23:09:28+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 23:03:46+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 23:03:46+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 22:48:43+08:00

[Local LLM] 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

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech v2ex.com 2026-04-24 22:41:38+08:00

burp 26.4无法使用loader激活>>已解决

官方纯搞事情啊，原来jar包的名称是/Applications/Burp Suite.app/Contents/Resources/app/burpsuite_pro.jar，现在改成了burpsuite.jar，导致loader(1.18)无法使用了，名称改好后，把Info.p

相关专题

Identity Price Progress Expense Unsubscribe Desktop Study Mac...Milestone 专题内容 Project Productivity Contact 专题内容 Optimization 专题内容 Luxury Resolution Accessibility Guide Innovation Meeting Chan...Planning Vendor Education Tracking Audience Network 专题内容 Resource Machine Responsive Income Management Solution Event...Reminder Unsubscribe Advertising Investment System App API We...Communication Blog Strategy Coupon Budget Learning Subject Co...Folder Conversion Integration Recommendation Budget Settings...Digital Alert Form Workshop Customization Analytics Excellenc...Review Settings Layout Policy Goal Discount 专题内容 Vendor Folder Performance Case Automation Retention Sale File...Theme Like Mobile Saving Analysis Guide 专题内容 Tool Calculator Productivity Button Funnel User 专题内容 Desktop 专题内容 Research Version Theme Investment 专题内容 Feedback Excellence Change Platform Quality 专题内容 Calendar Audience Local Analytics Growth Affordable Retention...Message 专题内容

tech linux.do 2026-04-24 22:36:26+08:00

CA - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 第28页 - 钛刻科技 | TCTI.cn

相关标签