Continuous conditional label preparation for training stylegen-t

[znavidi]

Hi,

I am trying to prepare my custom dataset with each sample condition provided as a numerical vector similar to the prompt encoding (and not class condition), and have not found a guidance on how to provide that to the model (let's assume I have the numerical sample conditions prepared prior to training). Should I just replace the class label with a vector of continuous vector in the dataset.json file? If so, can you pls prepare an example format? or guide otherwise.

Thanks in advance

[znavidi]

Following up on the above question. Would appreciate any guidance as I could not find more info on Github.