tutsmybarreh said:
Share it with the settings in the model sharing thread too if you want me to add it to the list:
https://mrdeepfakes.com/forums/thre...-thread-for-sharing-pretrained-trained-models
Did you pretrain it on custom HEAD face type dataset or used included WF FFHQ one?
I'm not him, but I checked his settings, and it's 128 resolution - probably too low to be of any use for head type. And if it was trained on the included facesets, they'd be useless for head anyway. I suggest you put a sticky note (if there isn't one already) that head type requires a custom faceset. I have uploaded a couple of those.
Here are his settings:
=============== Model Summary ===============
== ==
== Model name: Head Model_SAEHD ==
== ==
== Current iteration: 705937 ==
== ==
==------------- Model Options -------------==
== ==
== resolution: 128 ==
== face_type: head ==
== models_opt_on_gpu: True ==
== archi: df-ud ==
== ae_dims: 256 ==
== e_dims: 64 ==
== d_dims: 64 ==
== d_mask_dims: 22 ==
== masked_training: True ==
== eyes_prio: True ==
== uniform_yaw: True ==
== lr_dropout: n ==
== random_warp: True ==
== gan_power: 0.0 ==
== true_face_power: 0.0 ==
== face_style_power: 0.0 ==
== bg_style_power: 0.0 ==
== ct_mode: none ==
== clipgrad: True ==
== pretrain: True ==
== autobackup_hour: 1 ==
== write_preview_history: False ==
== target_iter: 1000000 ==
== random_flip: True ==
== batch_size: 8 ==
== ==
==-------------- Running On ---------------==
== ==
== Device index: 0 ==
== Name: GeForce RTX 2080 ==
== VRAM: 8.00GB ==
== ==
=============================================