xianyu

icon indicating an email [email protected]

icon company Zhejiang University icon location Beijing icon location Founder of OpenLLMAI. Do something cool and useful

Results 36 comments of xianyu

[BUG] ZERO++ | AssertionError: ZeRO parameter intra parallel group is already initialized

same error

How to use customized reward for PPO with golden answers

just use remote_rm_url

[Feature] Support RM API

great gob!

[WIP] support Ascend NPU backend

> 其实这么大的PR更适合开个自研hard fork 感觉可以，或者单开一个非cuda的分支吧，方便大家试用。目前来看已经可以试试了~

Support rl logging board

> I'm not sure if OpenRLHF is interested in integrating [this visualizaition tool](https://github.com/HarderThenHarder/RLLoggingBoard/tree/main), this tool has helped me a lot to analysis PPO training. > > So I try to...

[Roadmap] OpenRLHF Development Roadmap

More discussion: I personally pay great attention to the numbers 9 (moe related) and 4. Friends who are interested can get in touch with me.