Hi, I trained JIT-H/16 on my own data using the following commands:
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 torchrun --nproc_per_node=8 --nnodes=1 --node_rank=0 main_jit.py
--model JiT-H/16
--proj_dropout 0.2
--P_mean -0.8 --P_std 0.8
--noise_scale 1.0
--batch_size 3 --blr 5e-5
--epochs 600 --warmup_epochs 5
--gen_bsz 16 --num_images 200 --interval_min 0.1 --interval_max 1.0
--class_num 1 --label_drop_prob 0 --cfg 1.0
--output_dir ${OUTPUT_DIR} --resume ${PRETRAIN_DIR}
--data_path ${IMAGENET_PATH} --online_eval

My data only has one category, human body. After training for 600 epochs, the generated data is very blurry and abstract. How can I optimize it?