Training code?

[resurgo97]

Hi, I have a couple of questions:

1. Do you plan to upload the training code as well? If so, could you share when you might do this?
2. From what I understand, you adopted x₀ prediction instead of ϵ prediction. Does the base model (the one you pretrained) originally use ϵ prediction, and did you switch it to x₀ prediction afterward? If so, did you find that changing the prediction target negatively impacted optimization, given that the model had to learn a different objective?

Thanks in advance for any clarification!

[haodong2000]

Hi, thanks so much for your interest!

We do plan to release the training code, but we have to ask our supervisor first.

Please wait for my reply. Thanks.

Best,

[haodong2000]

Hi there, thanks again for your interest!

The training code will be released upon the acceptance of our paper. Same as Marigold.

Upon any questions, please feel free to reach out to us :)

[Cmonsta6]

Greetings Lotus team, @BlingHe @haodong2000

May I request the training logic for my research?

Thank you and happy new year