Hello, I am testing with default data_src and dst.
When the model is trained, the phenomenon shown in the picture below appears. how can i solve it?
In the first time, the loss decrease. However after some time, the loss is over 10.
GPU : GTX 1080 ti
CPU : Intel(R) Core(TM) i5-6600 CPU @ 3.30GHz
RAM : 16G
driver version : 531.61(I used the latest version, but the same error occur, so I downgrade the version)
Hello, I am testing with default data_src and dst. When the model is trained, the phenomenon shown in the picture below appears. how can i solve it? In the first time, the loss decrease. However after some time, the loss is over 10.
GPU : GTX 1080 ti CPU : Intel(R) Core(TM) i5-6600 CPU @ 3.30GHz RAM : 16G driver version : 531.61(I used the latest version, but the same error occur, so I downgrade the version)
==================== Model Summary ==================== == == == Model name: new_SAEHD == == == == Current iteration: 12035 == == == ==------------------ Model Options ------------------== == == == resolution: 128 == == face_type: f == == models_opt_on_gpu: True == == archi: liae-ud == == ae_dims: 256 == == e_dims: 64 == == d_dims: 64 == == d_mask_dims: 22 == == masked_training: True == == eyes_mouth_prio: False == == uniform_yaw: False == == blur_out_mask: False == == adabelief: True == == lr_dropout: n == == random_warp: True == == random_hsv_power: 0.0 == == true_face_power: 0.0 == == face_style_power: 0.0 == == bg_style_power: 0.0 == == ct_mode: none == == clipgrad: False == == pretrain: False == == autobackup_hour: 0 == == write_preview_history: False == == target_iter: 0 == == random_src_flip: False == == random_dst_flip: True == == batch_size: 8 == == gan_power: 0.0 == == gan_patch_size: 16 == == gan_dims: 16 == == == ==------------------- Running On --------------------== == == == Device index: 0 == == Name: NVIDIA GeForce GTX 1080 Ti == == VRAM: 9.41GB == == ==