Training cLR-GAN or cVAE-GAN about bicyclegan HOT 3 CLOSED

You can do it by changing the lambda hyper-parameters.
For cLR-GAN: set --lambda_GAN 0 and --lambda_kl 0.
For cVAE-GAN: set --lambda_GAN2 0 and --lambda_z 0.

from bicyclegan.

Thank you very much for your response!
It is very simple :)

from bicyclegan.

Hi,

Thank you for sharing your great work!

As your above suggestion, I try to train cVAE-GAN model by setting --lambda_GAN2 0 and --lambda_z 0.
python .\train.py^
--display_id 1^
--dataroot .\datasets\polyp_612^
--model bicycle_gan^
--which_direction BtoA^
--checkpoints_dir .\checkpoints\polyp_612^
--loadSize 256^
--fineSize 256^
--nz 8^
--input_nc 1^
--niter 30^
--niter_decay 30^
--use_dropout^
--lambda_GAN2 0^
--lambda_z 0

But, I got this error:
create web directory .\checkpoints\polyp_612\web...
Traceback (most recent call last):
File ".\train.py", line 36, in
model.optimize_parameters()
File "E:\BicycleGAN\models\bicycle_gan_model.py", line 212, in optimize_parameters
self.update_G_and_E()
File "E:\BicycleGAN\models\bicycle_gan_model.py", line 200, in update_G_and_E
self.backward_EG()
File "E:\BicycleGAN\models\bicycle_gan_model.py", line 156, in backward_EG
self.loss_G_GAN2 = self.backward_G_GAN(self.fake_data_random, self.netD2, self.opt.lambda_GAN2)
AttributeError: 'BiCycleGANModel' object has no attribute 'netD2'

If I train original model(BicycleGAN), i.e., didn't set --lambda_GAN2 and --lambda_z, training is working.

Could you help me solve this problem or any other way to train cVAE-GAN model?

Thank you.

from bicyclegan.