How to use the dataset of 512 * 512

[MingfuYAN]

Hi, thanks for the great work. I want to train with the 512*512 dataset, and I set --image_size to 512, but it doesn't work.

[Wangt-CN]

Hi, thanks for your interests. Which config file did you use?

[MingfuYAN]

Thanks for your reply, here is my config file.

--train_yaml /data0/mfyan/Human_Attribute/composite/train_TiktokDance-coco-single_person-Lindsey_0411_youtube-SHHQ-1.0-deepfashion2-laion_human-masks-single_cap.yaml \
--val_yaml /data0/mfyan/Human_Attribute/composite/val_TiktokDance-coco-single_person-SHHQ-1.0-masks-single_cap.yaml \

[gallenszl]

hello, same question, if i change the image_size to 512 * 512, the model can not converge to a reasonable position while original 256 * 256 can. I only change the img_full_size = (256, 256) img_size = (256, 256) in config/ref_attn_clip_combine_controlnet_attr_pretraining/coco_S256_xformers_tsv_strongrand.py to （512，512）

[gallenszl]

@Wangt-CN