Could you share the training script?

#1
by andysalerno - opened

Hi! Thanks for the great model!

I am also interested in performing DPO tuning on MoE models like your Mixtral_7Bx2_MoE.

I'm successfully run DPO against non-MoE Mistral models using the Huggingface Trl library. But failing at MoE. If you'd be so kind, could you share your technique? Thanks!

Owner

still testing, not ready to share

Sign up or log in to comment