kwonminki / Asyrp_official

official repo for Asyrp : Diffusion Models already have a Semantic Latent Space (ICLR2023)
MIT License
249 stars 20 forks source link

do ./checkpoint/* work? #11

Closed Lifedecoder closed 1 year ago

Lifedecoder commented 1 year ago

Hi! I've tried several checkpoints saved in ./checkpoint, like smiling, angry, young, it seems that they couldn't change corresponding attributes. Are there any problems of these checkpoints you have uploaded?

Lifedecoder commented 1 year ago

my test codes are as follows: sh_file_name="run.sh" gpu="3" config="celeba.yml" # for custom.yml, you need to set custom_train_dataset_dir and custom_test_dataset_dir. If not, vice versa. guid="angry" # we don't use it but need to run main.py

CUDA_VISIBLE_DEVICES=$gpu python main.py --run_test \ --config $config \ --exp ./runs/test \ --edit_attr $guid \ --do_train 0 \ --do_test 1 \ --n_test_img 10 \ --bs_test 1 \ --save_x0 \ --t_0 999 \ --n_inv_step 40 \ --n_test_step 40 \ --get_h_num 1 \ --train_delta_block \ --add_noise_from_xt \ --lpips_addnoise_th 1.2 \ --lpips_edit_th 0.33 \ --sh_file_name $sh_file_name \ --use_x0_tensor \ --hs_coeff_delta_h 1.0 \ --model_path "/pretrained_models/celeba_hq.ckpt" \ --load_from_checkpoint "checkpoint/angry_LC_CelebA_HQ_t999_ninv40_ngen40_0.pth"

Lifedecoder commented 1 year ago

got it, might increase hs_coeff_delta_h to get obvious results