提到 - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 第2页 - 钛刻科技

大项目中大家真的会用 Spec-Driven Development 吗？

https://openai.com/zh-Hans-CN/index/harness-engineering/ 中提到要把 Spec & Planning & Tasks 进度放进 git 仓库中，大家实践中真的会这么做吗？但是我看 codex 仓库内根本没这些

相关专题

tech www.v2ex.com 2026-04-26 03:20:08+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 03:07:34+08:00

大项目中大家真的会用 Spec-Driven Development 吗？

https://openai.com/zh-Hans-CN/index/harness-engineering/ 中提到要把 Spec & Planning & Tasks 进度放进 git 仓库中，大家实践中真的会这么做吗？但是我看 codex 仓库内根本没这些

相关专题

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 02:20:08+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 02:07:34+08:00

大项目中大家真的会用 Spec-Driven Development 吗？

https://openai.com/zh-Hans-CN/index/harness-engineering/ 中提到要把 Spec & Planning & Tasks 进度放进 git 仓库中，大家实践中真的会这么做吗？但是我看 codex 仓库内根本没这些

相关专题

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 01:20:08+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 01:07:34+08:00

大项目中大家真的会用 Spec-Driven Development 吗？

https://openai.com/zh-Hans-CN/index/harness-engineering/ 中提到要把 Spec & Planning & Tasks 进度放进 git 仓库中，大家实践中真的会这么做吗？但是我看 codex 仓库内根本没这些

相关专题

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 00:20:08+08:00

大项目中大家真的会用 Spec-Driven Development 吗？

https://openai.com/zh-Hans-CN/index/harness-engineering/ 中提到要把 Spec & Planning & Tasks 进度放进 git 仓库中，大家实践中真的会这么做吗？但是我看 codex 仓库内根本没这些

相关专题

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-26 00:20:08+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 22:29:34+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 22:29:34+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 22:05:33+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 22:05:33+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 21:45:41+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 21:45:41+08:00

大项目中大家真的会用 Spec-Driven Development 吗？

https://openai.com/zh-Hans-CN/index/harness-engineering/ 中提到要把 Spec & Planning & Tasks 进度放进 git 仓库中，大家实践中真的会这么做吗？但是我看 codex 仓库内根本没这些

相关专题

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 20:45:58+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 20:44:05+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 20:44:05+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

Site Form AI Resource Story Analytics Education Ranking 专题内容 Services 专题内容 Site Search 视频 Conference Performance Forecast 专题内容 Reminder Products 专题内容 Navigation Customer Partner SEO Community 专题内容 Customer 专题内容 Management 专题内容 Investment 专题内容 Event Backup Landing Responsive Policy Document Rating 专题内容 Advertising 专题内容 Update Study Account Tracking Excellence Success Beauty Alert...Fitness Document Food Form Extension Social Health Strategy U...Screen Campaign 专题内容 Analysis Affordable Rating Demographic Discovery Discount Tra...Communication File Form Investment Alert Calendar 专题内容 Demographic Tactic Investment Like Event Collaboration 专题内容 Course Engagement Marketing Image Landing Support Resource 专题内容 Data Milestone Widget Faq 专题内容 Prospect Profit 专题内容 Backup Training Premium Prospect Sale Alert Change Alliance B...

tech www.v2ex.com 2026-04-25 20:44:05+08:00

提到 - 钛刻 - 科技风向旗 - 深度刻画技术趋势,引领数字未来 - 第2页 - 钛刻科技 | TCTI.cn

相关标签