Skip to main content

CurateClick

github.com favicon

大模型RLHF(ppo奖励模型)训练偏好数据排序助手

一个支持ollama本地模型的工具,用于对大模型生成回答进行排序,生成偏好数据以训练奖励模型,进而优化SFT模型。

Check out 大模型RLHF(ppo奖励模型)训练偏好数据排序助手 on CurateClick

• Copy the embed code to showcase this product on your website

• Share on X to spread the word about this amazing tool

描述:简单来说就是你经过sft微调后,想通过RLHF(ppo奖励模型)训练怎么样的模型,就给你的模型生成回答进行排序,最后再导出偏好数据去训练奖励模型,再用奖励模型去训练sft模型开源地址:https://github.com/zjrwtx/preference_databuilder演示视频地址:https://www.bilibili.com/video/BV1P1421675z/?spm_id_from=333.999.0.0&vd_source=5531fb0981ef79f87198a3c2651dff93

Latest Weekly Picks

freeadhd.com favicon
Weekly Pick

ADHD Test - a free, no-sign-up adult ADHD self-screening tool with instant results.

ADHD Test
pixmira.ai favicon
Weekly Pick

Free AI photo editor and image generator for fast visual creation.

PixMira AI
pslratings.com favicon
Weekly Pick

Free PSL Rating gives an instant AI PSL score with tier and four category breakdowns, no signup.

Free PSL Rating
gothic1remake.com favicon
Weekly Pick

Gothic 1 Remake Guide — Release, Walkthroughs & Quests

gothic 1 remake

Predict video virality before you publish.

Virality Predictor

Powerful Motion Control AI Video Generator

Motion Control AI

PlusGO 是面向中国大陆用户的 ChatGPT Plus 代充值服务平台。使用官方 CDKey,支持支付宝和微信扫码支付,平均 2 分钟完成 Plus 开通,自 2025 年起已为超过 10,000 名用户完成充值。

www.toolcenter.ai favicon
Weekly Pick

Discover the Best AI & Productivity Tools — 1,500+ tools with honest reviews.

ToolCenter