https://www.instagram.com/p/Db7RCsnkxVJ/?igsh=MTVnM2d0b3p3bzR3bw==
https://www.instagram.com/reel/Db5NSG_S-W0/?igsh=MWE5OXlmbWU0OGVwaA==
https://www.instagram.com/p/Db7Dya7EsuR/?igsh=NThoenk4cnBueTI3
https://www.instagram.com/reel/Db5HAx3zrMK/?igsh=MXRndXRtbWh2dzU2NA==
https://www.threads.com/share/BAgW7lol3o/
Threads
What's that (@hkwwtai) on Threads
What's that (@hkwwtai) on Threads
Oh my pi
https://www.instagram.com/reel/DbcSmW7p_jR/?igsh=MWw2cWN6YTJkZTMyeQ==
https://www.instagram.com/reel/Db3UttuvfwN/?igsh=MW5oMXR2b3YzY2g2MQ==
https://www.instagram.com/p/Db2soJGnFVR/?igsh=bDN1cHQwdGVreWZs
https://www.instagram.com/reel/DbXPdJnyY9t/?igsh=MXQ2NjJva3psanNrbA==
https://www.threads.com/share/BAWOvNrjGW/
#網頁快照
網頁快照
https://www.threads.com/share/BAJa6J6VJC/
https://www.instagram.com/reel/Db0j5WGFUGj/?igsh=MXc5ZTlkcmIwYXhwNw==
Instagram
Instagram
Post might not be available
https://www.instagram.com/reel/Db2sPrwTKEI/?igsh=ZjB1d2NzNGJ0eG41
https://youtu.be/YO5JbD8klHU?si=7padAYeqEXfQ6OJi
YouTube
TurboFieldfare详解：Gemma4 26B大模型只用2GB内存，8GB Macbook Neo也能跑本地AI？| Turo Fieldfare 详解 | 内存价格有救了？！
Turbo Fieldfare repo：https://github.com/drumih/turbo-fieldfare

本地运行大模型，最大的门槛往往不是模型本身，而是内存。

这期视频介绍一个最近很有意思的开源项目：TurboFieldfare。它针对 Gemma 4 26B-A4B 做了专门优化，通过 Swift + Metal 推理引擎，把大部分专家权重放在 SSD 上按需读取，只在内存里保留约 2GB 的核心数据和 KV Cache，让 8GB 内存的 Apple Silicon Mac…
https://www.facebook.com/share/p/198f19T6Qq/?mibextid=wwXIfr
https://www.instagram.com/p/Db2v8H9gZx1/?igsh=aHpxbmJxdDc0MDds
https://www.instagram.com/reel/Db1nnrGgaOo/?igsh=MXhqZ29qdGZ1MHA2Nw==
https://www.facebook.com/share/p/1D9rEQsLwt/
https://www.instagram.com/reel/DbzDu5Tgso4/?igsh=MWdiZzdpcXlvaHd1Yg==
Before
After
Back to Top