df-udt 320 162 ，万能模型，购买时老板附加赠送的模型
。老板给的转换后视频效果图，截取的一张
这是我自己随便拿个dst合成的一张
配置有点高，低配电脑带不动，我也不清楚这换脸的人是谁，有知道的可以说下。
所以这模型只当送了，积攒一丢丢灵石对话的遮罩。
320isabelleharis.f.162W.dfud
当前迭代: 1628728
________________________模型参数_________________________
模型架构: df-udt
人脸类型: f
模型分辨率: 320
自动编码器维度 ae_dims: 352
编码器维度 e_dims: 88
解码器维度 d_dims: 88
解码器掩码维度 d_mask_dims: 28
BS大小: 16
学习率lr: 5e-05
遮罩约束训练范围: 开启
均衡侧脸训练权重: 开启
素材增强-翻转dst: 开启
RG优化器: 开启
信仰优化器 adabelief: 开启
模型优化器放置于GPU: 开启
梯度剪裁: 开启
=============== Model Summary ===============
== ==
== Model name: DF-UDT320F_SAEHD ==
== ==
== Current iteration: 1628728 ==
== ==
==------------- Model Options -------------==
== ==
== resolution: 320 ==
== face_type: f ==
== models_opt_on_gpu: True ==
== archi: df-udt ==
== ae_dims: 352 ==
== e_dims: 88 ==
== d_dims: 88 ==
== d_mask_dims: 28 ==
== masked_training: True ==
== eyes_mouth_prio: False ==
== uniform_yaw: True ==
== blur_out_mask: False ==
== adabelief: True ==
== lr_dropout: n ==
== random_warp: False ==
== true_face_power: 0.0 ==
== face_style_power: 0.0 ==
== bg_style_power: 0.0 ==
== ct_mode: none ==
== clipgrad: True ==
== pretrain: False ==
== autobackup_hour: 0 ==
== write_preview_history: False ==
== target_iter: 0 ==
== random_src_flip: False ==
== random_dst_flip: True ==
== batch_size: 16 ==
== gan_power: 0.0 ==
== gan_patch_size: 60 ==
== gan_dims: 30 ==
== random_hsv_power: 0.0 ==
== ==
==-------------- Running On ---------------==
== ==
== Device index: 0 ==
== Name: Tesla P40 ==
== VRAM: 20.20GB ==
== ==
=============================================
F脸告辞了
页:
[1]