Please reconsider to share either the model or the code|technique to reproduce it.

#4
by mindkrypted - opened

Hello Mr. Cheung, I tried this modified model on your HF space, I really feel that it should be shared with the community. (Either here or another platform of your choice that respects what you believe in.) A code sample or explanations on how to reproduce your achievement could be another option.

Thanks :)

CausalLM org

Some details here: https://x.com/RealJosephus/status/1794380925318680590

Basically data driven, and the RL method should be quite similar to the original MiniCPM.

JosephusCheung changed discussion status to closed

Sign up or log in to comment