the pretrained CelebA 64x64 model gives me noisy sample

[srk1995]

When I extracted image samples using the pretrained CelebA 64x64 model you provided, I obtained noisy images. Is this the correct pretrained weight?

I used arguments below.

python scripts/image_sample.py
--image_size 64
--model_path PATH_TO_CHECKPOINT
--num_channels 192 --num_head_channels 64 --num_res_blocks 3 --attention_resolutions 32,16,8
--resblock_updown True --use_new_attention_order True --learn_sigma True --dropout 0.1
--diffusion_steps 1000 --noise_schedule cosine --use_scale_shift_norm True --batch_size 10 --num_samples 10

If the above arguments are not correct, can you provide me the arguments for sampling CelebA 64 x 64 images?

[srk1995]

Sorry, there were mistakes some arguments I used. The above arguments works well.