search

THUDM / ImageReward

[NeurIPS 2023] ImageReward: Learning and Evaluating Human Preferences for Text-to-image Generation
Apache License 2.0
1.15k stars 64 forks source link

How to train stabilityai/stable-diffusion-xl-base-1.0 using ImageReward model #77

Open emelpolaris opened 6 months ago

emelpolaris commented 6 months ago

I found the code snippet for training CompVis/stable-diffusion-v-1-4 using ImageReward. Based on this, I tried to make the code for training stabilityai/stable-diffusion-xl-base-1.0 but failed. Is it possible to train stabilityai/stable-diffusion-xl-base-1.0 using ImageReward model? Thanks

xujz18 commented 6 months ago

Hello, thanks for your attention! You can refer to our ReFL code and just add the ReFL action similarly in the SDXL fine-tuning script.

emelpolaris commented 6 months ago

Thanks for your response. I tried to add the ReFL action in the SDXL fine-tuning, but no luck.

emelpolaris commented 6 months ago

@xujz18 stuck with an error. could you provide a code for it?

emelpolaris commented 6 months ago

If anyone is interested in it and resolve this issue, please reach out to me https://join.slack.com/t/xrunnergroup/shared_invite/zt-2h89gocpx-KjI4Vf0Z8ZtA1kP2DvsX1Q

amulyaprasanth commented 6 months ago

Hello, thanks for your attention! You can refer to our ReFL code and just add the ReFL action similarly in the SDXL fine-tuning script.

I was trying the same thing but no luck. Could you please elaborate the steps