pytorch-CycleGAN-and-pix2pix
pytorch-CycleGAN-and-pix2pix copied to clipboard
junyanz
Hello, I am very interested in the work done by this model, but the following situation occurred during GPU training, my graphics card model is RTX3060.
(epoch: 1, iters: 100, time: 0.317, data: 2.125) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 200, time: 0.320, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 300, time: 0.321, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 400, time: 0.428, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 500, time: 0.312, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 600, time: 0.318, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 700, time: 0.323, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan (epoch: 1, iters: 800, time: 0.385, data: 0.000) D_A: nan G_A: nan cycle_A: nan idt_A: nan D_B: nan G_B: nan cycle_B: nan idt_B: nan
Which training command line did you use? Did you train a model on the default datasets?
Thanks for your reply, I use the command line is “python train.py --dataroot ./datasets/maps --name maps_cyclegan --gpu_ids 0 --model cycle_gan”. The same error occurs when I use the command "python train.py --dataroot ./datasets/facades --name facades_pix2pix --model pix2pix --direction BtoA". Download is the default data set, built in Windows 10 above the environment. Pytorch version is '1.4.0+ CU92 '. With CPU training, everything is fine, but with GPU acceleration, the number of parameters explodes and this happens. Looking forward to your reply.