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

VLLM - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 钛刻科技 | TCTI.cn - 钛刻 (TCTI.cn) 为您提供最前沿的硬核科技资讯、深度评测和未来技术趋势分析。

共 5 篇相关文章

[Qwen3.6]27B-FP8 SGLang本地部署-venv dev经验分享。

上集回顾 [Qwen3.6]27B-FP8 VLLM本地部署主观个人测评环境： WLS2 Ubuntu22.04 硬件： 4090 48G SGLang太佛系了，快一个月了还不更新正式版，需要好好挖掘pr。目前已经找到4090 48G本地部署的版本。我测试了两个模型。以及DF

相关专题

Trading Contact Technology Sales 专题内容 Story Webinar Internet 专题内容 Machine Automation Tool Update Landing 专题内容 Cheap Dashboard 专题内容 Customer Discount Search Target Contact Coupon 专题内容 Management Spreadsheet Experience 专题内容 Demographic Faq Website Client Mobile Email Sport About 专题内容影视 Budget Planning 专题内容 Optimization Unsubscribe 专题内容 Promotion Strategy Goal Support 专题内容 Image Collaboration Vacation File Success Quality Platform Go...Link Media 专题内容 Prospect Database Networking Project Share Demographic Audien...Accessibility Resolution Media Database Brand Health 专题内容 File Customer Server Subject Change Tracking Luxury 专题内容 Development Conversion Target Support 专题内容 Budget Extension Responsive Food Affordable 专题内容 App Support File Experience URL Sales Task Milestone 专题内容 Search Platform Beauty App About Follow 专题内容 Database 专题内容

tech linux.do 2026-05-04 10:21:26+08:00

xllm 真的比 vllm+plugin 性能好么？

相关专题

Trading Contact Technology Sales 专题内容 Story Webinar Internet 专题内容 Machine Automation Tool Update Landing 专题内容 Cheap Dashboard 专题内容 Customer Discount Search Target Contact Coupon 专题内容 Management Spreadsheet Experience 专题内容 Demographic Faq Website Client Mobile Email Sport About 专题内容影视 Budget Planning 专题内容 Optimization Unsubscribe 专题内容 Promotion Strategy Goal Support 专题内容 Image Collaboration Vacation File Success Quality Platform Go...Link Media 专题内容 Prospect Database Networking Project Share Demographic Audien...Accessibility Resolution Media Database Brand Health 专题内容 File Customer Server Subject Change Tracking Luxury 专题内容 Development Conversion Target Support 专题内容 Budget Extension Responsive Food Affordable 专题内容 App Support File Experience URL Sales Task Milestone 专题内容 Search Platform Beauty App About Follow 专题内容 Database 专题内容

tech www.v2ex.com 2026-04-27 13:46:35+08:00

昇腾910B本地部署DeepSeek-V4-Flash(w8a8量化版)测试

老登们下班了，现在可以霍霍服务器了 vllm-ascend部署文档： DeepSeek-V4 — vllm-ascend 模型： DeepSeek-V4-Flash-w8a8-mtp · 模型库启动成功：先问一下洗车问题：逻辑OK 跑一下文档中的数据集(GSM8K，数学推理

相关专题

Trading Contact Technology Sales 专题内容 Story Webinar Internet 专题内容 Machine Automation Tool Update Landing 专题内容 Cheap Dashboard 专题内容 Customer Discount Search Target Contact Coupon 专题内容 Management Spreadsheet Experience 专题内容 Demographic Faq Website Client Mobile Email Sport About 专题内容影视 Budget Planning 专题内容 Optimization Unsubscribe 专题内容 Promotion Strategy Goal Support 专题内容 Image Collaboration Vacation File Success Quality Platform Go...Link Media 专题内容 Prospect Database Networking Project Share Demographic Audien...Accessibility Resolution Media Database Brand Health 专题内容 File Customer Server Subject Change Tracking Luxury 专题内容 Development Conversion Target Support 专题内容 Budget Extension Responsive Food Affordable 专题内容 App Support File Experience URL Sales Task Milestone 专题内容 Search Platform Beauty App About Follow 专题内容 Database 专题内容

tech linux.do 2026-04-24 18:41:13+08:00

到底怎么才能降低Qwen3.5-35B-A3B的think啊

大佬们最近用最新的gpustack2.1.2版本搭配8个L40装了Qwen3.5-35B-A3B的vllm模型。官方文档说可以靠thinking_budget参数来调节思考的长度，但是根本不好使啊.. system角色写提示词要求它精简思考过程也不行。。。。没招了.. 1

相关专题

Trading Contact Technology Sales 专题内容 Story Webinar Internet 专题内容 Machine Automation Tool Update Landing 专题内容 Cheap Dashboard 专题内容 Customer Discount Search Target Contact Coupon 专题内容 Management Spreadsheet Experience 专题内容 Demographic Faq Website Client Mobile Email Sport About 专题内容影视 Budget Planning 专题内容 Optimization Unsubscribe 专题内容 Promotion Strategy Goal Support 专题内容 Image Collaboration Vacation File Success Quality Platform Go...Link Media 专题内容 Prospect Database Networking Project Share Demographic Audien...Accessibility Resolution Media Database Brand Health 专题内容 File Customer Server Subject Change Tracking Luxury 专题内容 Development Conversion Target Support 专题内容 Budget Extension Responsive Food Affordable 专题内容 App Support File Experience URL Sales Task Milestone 专题内容 Search Platform Beauty App About Follow 专题内容 Database 专题内容

tech linux.do 2026-04-24 12:47:47+08:00

[Qwen3.6]27B-FP8 VLLM本地部署主观个人测评

硬件配置组件规格 CPU Intel i5-13600K GPU RTX 4090 48G + RTX 4070Ti 12G 内存 DDR4-3600 128G (4x32G) 主板华硕 Z690-P D4 系统 Windows 11 LTSC WSL Ubuntu 22.

相关专题

Trading Contact Technology Sales 专题内容 Story Webinar Internet 专题内容 Machine Automation Tool Update Landing 专题内容 Cheap Dashboard 专题内容 Customer Discount Search Target Contact Coupon 专题内容 Management Spreadsheet Experience 专题内容 Demographic Faq Website Client Mobile Email Sport About 专题内容影视 Budget Planning 专题内容 Optimization Unsubscribe 专题内容 Promotion Strategy Goal Support 专题内容 Image Collaboration Vacation File Success Quality Platform Go...Link Media 专题内容 Prospect Database Networking Project Share Demographic Audien...Accessibility Resolution Media Database Brand Health 专题内容 File Customer Server Subject Change Tracking Luxury 专题内容 Development Conversion Target Support 专题内容 Budget Extension Responsive Food Affordable 专题内容 App Support File Experience URL Sales Task Milestone 专题内容 Search Platform Beauty App About Follow 专题内容 Database 专题内容

tech linux.do 2026-04-24 00:28:36+08:00