sumorday
sumorday
I guess that's because resnet would be better than simple cnn architecture. and tf/torch even different gpu, or even you changed seeds will get a different result.
Hi,I met same question but solved. let me show you my codes. /home/edward/archive/T2I_CL-main/AttnGAN+CL/code/main.py --cfg /home/edward/archive/T2I_CL-main/AttnGAN+CL/code/cfg/bird_attn2.yml --**gpu 0** almost same like the author, but make sure you have gpu .I guess...
update new error for training pretrain_DAMSM.py Traceback (most recent call last): File "/home/edward/archive/T2I_CL-main/AttnGAN+CL/code/pretrain_DAMSM.py", line 336, in mask = mask_correlated_samples_2(batch_size) File "/lustre7/home/edward/archive/T2I_CL-main/AttnGAN+CL/code/masks.py", line 4, in mask_correlated_samples_2 mask = torch.ones((args.batch_size * 2,...
update use this mask.py https://github.com/huiyegit/T2I_CL/blob/ff6f82e07907dd1194af48fed24c8ba24487874c/DM-GAN%2BCL/code/masks.py
Sorry, can not open the colab link
same there. Any update?
https://github.com/huiyegit/T2I_CL/issues/10
batchsize？