
jev-forge
zwliJay
An open training and inference stack for Jev-style decision models. Train models to score dynamic candidate branches from a shared prefix, with support for high-cardinality choice, calibration, and fast batched inference.
Python
Other 最后更新 2026年9月23日活跃165
Stars
15
Forks
6
Watchers
1
Issues
Star 增长
今日0
近 7 天0
近 30 天0
综合评分39.61
默认分支main
暂无 README 内容
项目可能尚未同步完成,请稍后查看