Hugging Face's logo

sachi1
/
PreciseControl

Model card Files
xet
Community
sachi1 commited on
Commit
4df5c8d
·
verified ·
1 Parent(s): ff5c6e9

Upload 4 files

Browse files
Files changed (4) hide show
  1. wt_mapper_70k_sd2_idloss/checkpoints/embeddings_gs-139999.pt +3 -0
  2. wt_mapper_70k_sd2_idloss/checkpoints/last.ckpt +3 -0
  3. wt_mapper_70k_sd2_idloss/configs/wt_mapper_70k_sd2_idloss-lightning.yaml +16 -0
  4. wt_mapper_70k_sd2_idloss/configs/wt_mapper_70k_sd2_idloss-project.yaml +158 -0
wt_mapper_70k_sd2_idloss/checkpoints/embeddings_gs-139999.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b96b78f153f53fba4e8e7d2e66ad2a9ff0d78616931380f7c8bb63785567fc5e
3
+ size 153224563
wt_mapper_70k_sd2_idloss/checkpoints/last.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d030ad8db708280fcae77d87e973102039acd23a11bdecc3db8eb6c0ac940ee1
3
+ size 431
wt_mapper_70k_sd2_idloss/configs/wt_mapper_70k_sd2_idloss-lightning.yaml ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ lightning:
2
+ modelcheckpoint:
3
+ params:
4
+ every_n_train_steps: 20000
5
+ callbacks:
6
+ image_logger:
7
+ target: main.ImageLogger
8
+ params:
9
+ batch_frequency: 10000
10
+ max_images: 8
11
+ increase_log_steps: false
12
+ trainer:
13
+ benchmark: true
14
+ max_steps: 150000
15
+ accelerator: ddp
16
+ gpus: 0,
wt_mapper_70k_sd2_idloss/configs/wt_mapper_70k_sd2_idloss-project.yaml ADDED
@@ -0,0 +1,158 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ base_learning_rate: 5.0e-05
3
+ target: ldm.models.diffusion.ddpm.LatentDiffusion
4
+ params:
5
+ linear_start: 0.00085
6
+ linear_end: 0.012
7
+ num_timesteps_cond: 1
8
+ log_every_t: 200
9
+ timesteps: 1000
10
+ first_stage_key: image
11
+ cond_stage_key: caption
12
+ image_size: 64
13
+ channels: 4
14
+ cond_stage_trainable: true
15
+ conditioning_key: crossattn
16
+ monitor: val/loss_simple_ema
17
+ scale_factor: 0.18215
18
+ use_ema: false
19
+ embedding_reg_weight: 0.0
20
+ unfreeze_model: false
21
+ model_lr: 0.0
22
+ use_face_masking: false
23
+ use_faceid_loss: true
24
+ id_loss_weight: 2
25
+ personalization_config:
26
+ target: ldm.modules.embedding_manager.EmbeddingManagerId
27
+ params:
28
+ placeholder_strings:
29
+ - sks
30
+ - ks
31
+ - ata
32
+ - tre
33
+ - ry
34
+ - bop
35
+ - rn
36
+ - '&'
37
+ - '*'
38
+ - '`'
39
+ initializer_words:
40
+ - face
41
+ - face
42
+ - face
43
+ - face
44
+ - face
45
+ - face
46
+ - face
47
+ - face
48
+ - face
49
+ - face
50
+ max_ids: 20
51
+ num_embeds_per_token: 2
52
+ meta_mlp_depth: 4
53
+ loss_type: none
54
+ meta_inner_dim: 512
55
+ context_dim: 1024
56
+ meta_heads: 1
57
+ use_rm_mlp: false
58
+ test_mode: image
59
+ momentum: 0.99
60
+ save_fp16: false
61
+ use_aligned_faces: true
62
+ use_regularization: true
63
+ regularization_weight: 1.0e-07
64
+ use_norm_reg: false
65
+ use_hspace: false
66
+ use_stylegan_based_mapper: true
67
+ use_timestep_embedder: true
68
+ use_celeb_basis: false
69
+ use_basis_offset: false
70
+ shift_basis: false
71
+ embedding_manager_ckpt: ''
72
+ unet_config:
73
+ target: ldm.modules.diffusionmodules.openaimodel.UNetModel
74
+ params:
75
+ use_fp16: false
76
+ image_size: 32
77
+ in_channels: 4
78
+ out_channels: 4
79
+ model_channels: 320
80
+ attention_resolutions:
81
+ - 4
82
+ - 2
83
+ - 1
84
+ num_res_blocks: 2
85
+ channel_mult:
86
+ - 1
87
+ - 2
88
+ - 4
89
+ - 4
90
+ num_head_channels: 64
91
+ use_spatial_transformer: true
92
+ use_linear_in_transformer: true
93
+ transformer_depth: 1
94
+ context_dim: 1024
95
+ use_checkpoint: true
96
+ legacy: false
97
+ first_stage_config:
98
+ target: ldm.models.autoencoder.AutoencoderKL
99
+ params:
100
+ embed_dim: 4
101
+ monitor: val/rec_loss
102
+ ddconfig:
103
+ double_z: true
104
+ z_channels: 4
105
+ resolution: 512
106
+ in_channels: 3
107
+ out_ch: 3
108
+ ch: 128
109
+ ch_mult:
110
+ - 1
111
+ - 2
112
+ - 4
113
+ - 4
114
+ num_res_blocks: 2
115
+ attn_resolutions: []
116
+ dropout: 0.0
117
+ lossconfig:
118
+ target: torch.nn.Identity
119
+ cond_stage_config:
120
+ target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
121
+ params:
122
+ freeze: true
123
+ layer: penultimate
124
+ use_celeb: true
125
+ use_svd: true
126
+ rm_repeats: true
127
+ celeb_txt: ./infer_images/wiki_names_v2.txt
128
+ n_components: 512
129
+ use_sample_reduce: false
130
+ n_samples: 513
131
+ use_flatten: false
132
+ num_embeds_per_token: 2
133
+ use_clip_face_basis: false
134
+ face_basis_path: ./weights/clip_face_basis100k_pca_wo_mean.pkl
135
+ ckpt_path: ./weights/v2-1_512-ema-pruned.ckpt
136
+ data:
137
+ target: main.DataModuleFromConfig
138
+ params:
139
+ batch_size: 2
140
+ num_workers: 8
141
+ wrap: false
142
+ train:
143
+ target: ldm.data.face_id.FFhq_dataset
144
+ params:
145
+ root_dir: ../../datasets/ffhq/
146
+ split: train
147
+ use_aug: false
148
+ image_size: 512
149
+ limit_dataset_size: 70000
150
+ use_data_interpolation: false
151
+ validation:
152
+ target: ldm.data.face_id.FFhq_dataset
153
+ params:
154
+ root_dir: ../../datasets/ffhq/
155
+ split: val
156
+ image_size: 512
157
+ limit_dataset_size: -1
158
+ use_aug: false