jevVIDEO
把 Jev 查询蒸馏成快 10–20 倍的小分类器
先跑 Jev 存结果,npx jimothy train 训出 15–45MB 模型,可服务端或浏览器跑,随时回退 Jev。
Andrew Prifer
@AndrewPrifer2026.09.22收录原推发布于 2026.09.22
Andrew Prifer @AndrewPrifer原推发布于 2026.09.22
You can automatically turn your Jev queries into tiny task-specific classifiers that run 10-20x faster! - Start with Jev, save the results - Run npx jimothy train - Get back a 15-45 MB model - Run on the server or in the browser - Fall back to Jev anytime - Repeat Demo: https://jimothy-r63s.vercel.app GitHub: https://github.com/AndrewPrifer/jimothy
x.com/andrewprifer/status/2102162296739099126采集于 2026.09.22 10:11