Hugging Face's logo

Wawaworker
/
exsswimsuit

Model card Files Community
1

Upload folder using huggingface_hub

#1
by Wawaworker - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+618
-0
Files changed (12) hide show
  1. exsswimsuit-06-21_23-26-35-save-420-30-0.safetensors +3 -0
  2. exsswimsuit-06-21_23-26-35-save-420-30-0.yaml +100 -0
  3. exsswimsuit-06-21_23-37-39-save-840-60-0.safetensors +3 -0
  4. exsswimsuit-06-21_23-37-39-save-840-60-0.yaml +100 -0
  5. exsswimsuit-06-21_23-48-42-save-1260-90-0.safetensors +3 -0
  6. exsswimsuit-06-21_23-48-42-save-1260-90-0.yaml +100 -0
  7. exsswimsuit-06-21_23-59-43-save-1680-120-0.safetensors +3 -0
  8. exsswimsuit-06-21_23-59-43-save-1680-120-0.yaml +100 -0
  9. exsswimsuit-06-22_00-10-40-save-2100-150-0.safetensors +3 -0
  10. exsswimsuit-06-22_00-10-40-save-2100-150-0.yaml +100 -0
  11. exsswimsuit-06-22_00-21-52-save-2520-180-0.safetensors +3 -0
  12. exsswimsuit-06-22_00-21-52-save-2520-180-0.yaml +100 -0
exsswimsuit-06-21_23-26-35-save-420-30-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61c8239e05c077afd1702e37a0a376f497909c7c116760d3ada74d548ebee7be
3
+ size 6938084280
exsswimsuit-06-21_23-26-35-save-420-30-0.yaml ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ params:
3
+ conditioner_config:
4
+ params:
5
+ emb_models:
6
+ - input_key: txt
7
+ is_trainable: false
8
+ params:
9
+ layer: hidden
10
+ layer_idx: 11
11
+ target: sgm.modules.encoders.modules.FrozenCLIPEmbedder
12
+ - input_key: txt
13
+ is_trainable: false
14
+ params:
15
+ always_return_pooled: true
16
+ arch: ViT-bigG-14
17
+ freeze: true
18
+ layer: penultimate
19
+ legacy: false
20
+ version: laion2b_s39b_b160k
21
+ target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2
22
+ - input_key: original_size_as_tuple
23
+ is_trainable: false
24
+ params:
25
+ outdim: 256
26
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
27
+ - input_key: crop_coords_top_left
28
+ is_trainable: false
29
+ params:
30
+ outdim: 256
31
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
32
+ - input_key: target_size_as_tuple
33
+ is_trainable: false
34
+ params:
35
+ outdim: 256
36
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
37
+ target: sgm.modules.GeneralConditioner
38
+ denoiser_config:
39
+ params:
40
+ discretization_config:
41
+ target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization
42
+ num_idx: 1000
43
+ scaling_config:
44
+ target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling
45
+ weighting_config:
46
+ target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting
47
+ target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser
48
+ disable_first_stage_autocast: true
49
+ first_stage_config:
50
+ params:
51
+ ddconfig:
52
+ attn_resolutions: []
53
+ attn_type: vanilla-xformers
54
+ ch: 128
55
+ ch_mult:
56
+ - 1
57
+ - 2
58
+ - 4
59
+ - 4
60
+ double_z: true
61
+ dropout: 0.0
62
+ in_channels: 3
63
+ num_res_blocks: 2
64
+ out_ch: 3
65
+ resolution: 256
66
+ z_channels: 4
67
+ embed_dim: 4
68
+ lossconfig:
69
+ target: torch.nn.Identity
70
+ monitor: val/rec_loss
71
+ target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper
72
+ network_config:
73
+ params:
74
+ adm_in_channels: 2816
75
+ attention_resolutions:
76
+ - 4
77
+ - 2
78
+ channel_mult:
79
+ - 1
80
+ - 2
81
+ - 4
82
+ context_dim: 2048
83
+ in_channels: 4
84
+ legacy: false
85
+ model_channels: 320
86
+ num_classes: sequential
87
+ num_head_channels: 64
88
+ num_res_blocks: 2
89
+ out_channels: 4
90
+ spatial_transformer_attn_type: softmax-xformers
91
+ transformer_depth:
92
+ - 1
93
+ - 2
94
+ - 10
95
+ use_checkpoint: true
96
+ use_linear_in_transformer: true
97
+ use_spatial_transformer: true
98
+ target: sgm.modules.diffusionmodules.openaimodel.UNetModel
99
+ scale_factor: 0.13025
100
+ target: sgm.models.diffusion.DiffusionEngine
exsswimsuit-06-21_23-37-39-save-840-60-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e32c98811820b3ee5161ee88ebdfe32fb72822089614e1299659f4ec2b00e526
3
+ size 6938084280
exsswimsuit-06-21_23-37-39-save-840-60-0.yaml ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ params:
3
+ conditioner_config:
4
+ params:
5
+ emb_models:
6
+ - input_key: txt
7
+ is_trainable: false
8
+ params:
9
+ layer: hidden
10
+ layer_idx: 11
11
+ target: sgm.modules.encoders.modules.FrozenCLIPEmbedder
12
+ - input_key: txt
13
+ is_trainable: false
14
+ params:
15
+ always_return_pooled: true
16
+ arch: ViT-bigG-14
17
+ freeze: true
18
+ layer: penultimate
19
+ legacy: false
20
+ version: laion2b_s39b_b160k
21
+ target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2
22
+ - input_key: original_size_as_tuple
23
+ is_trainable: false
24
+ params:
25
+ outdim: 256
26
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
27
+ - input_key: crop_coords_top_left
28
+ is_trainable: false
29
+ params:
30
+ outdim: 256
31
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
32
+ - input_key: target_size_as_tuple
33
+ is_trainable: false
34
+ params:
35
+ outdim: 256
36
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
37
+ target: sgm.modules.GeneralConditioner
38
+ denoiser_config:
39
+ params:
40
+ discretization_config:
41
+ target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization
42
+ num_idx: 1000
43
+ scaling_config:
44
+ target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling
45
+ weighting_config:
46
+ target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting
47
+ target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser
48
+ disable_first_stage_autocast: true
49
+ first_stage_config:
50
+ params:
51
+ ddconfig:
52
+ attn_resolutions: []
53
+ attn_type: vanilla-xformers
54
+ ch: 128
55
+ ch_mult:
56
+ - 1
57
+ - 2
58
+ - 4
59
+ - 4
60
+ double_z: true
61
+ dropout: 0.0
62
+ in_channels: 3
63
+ num_res_blocks: 2
64
+ out_ch: 3
65
+ resolution: 256
66
+ z_channels: 4
67
+ embed_dim: 4
68
+ lossconfig:
69
+ target: torch.nn.Identity
70
+ monitor: val/rec_loss
71
+ target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper
72
+ network_config:
73
+ params:
74
+ adm_in_channels: 2816
75
+ attention_resolutions:
76
+ - 4
77
+ - 2
78
+ channel_mult:
79
+ - 1
80
+ - 2
81
+ - 4
82
+ context_dim: 2048
83
+ in_channels: 4
84
+ legacy: false
85
+ model_channels: 320
86
+ num_classes: sequential
87
+ num_head_channels: 64
88
+ num_res_blocks: 2
89
+ out_channels: 4
90
+ spatial_transformer_attn_type: softmax-xformers
91
+ transformer_depth:
92
+ - 1
93
+ - 2
94
+ - 10
95
+ use_checkpoint: true
96
+ use_linear_in_transformer: true
97
+ use_spatial_transformer: true
98
+ target: sgm.modules.diffusionmodules.openaimodel.UNetModel
99
+ scale_factor: 0.13025
100
+ target: sgm.models.diffusion.DiffusionEngine
exsswimsuit-06-21_23-48-42-save-1260-90-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38dd2c1ed1f2b8378d0a74018c76fb6d11c5dbaf2affadae77a9583ff1785f94
3
+ size 6938084280
exsswimsuit-06-21_23-48-42-save-1260-90-0.yaml ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ params:
3
+ conditioner_config:
4
+ params:
5
+ emb_models:
6
+ - input_key: txt
7
+ is_trainable: false
8
+ params:
9
+ layer: hidden
10
+ layer_idx: 11
11
+ target: sgm.modules.encoders.modules.FrozenCLIPEmbedder
12
+ - input_key: txt
13
+ is_trainable: false
14
+ params:
15
+ always_return_pooled: true
16
+ arch: ViT-bigG-14
17
+ freeze: true
18
+ layer: penultimate
19
+ legacy: false
20
+ version: laion2b_s39b_b160k
21
+ target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2
22
+ - input_key: original_size_as_tuple
23
+ is_trainable: false
24
+ params:
25
+ outdim: 256
26
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
27
+ - input_key: crop_coords_top_left
28
+ is_trainable: false
29
+ params:
30
+ outdim: 256
31
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
32
+ - input_key: target_size_as_tuple
33
+ is_trainable: false
34
+ params:
35
+ outdim: 256
36
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
37
+ target: sgm.modules.GeneralConditioner
38
+ denoiser_config:
39
+ params:
40
+ discretization_config:
41
+ target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization
42
+ num_idx: 1000
43
+ scaling_config:
44
+ target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling
45
+ weighting_config:
46
+ target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting
47
+ target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser
48
+ disable_first_stage_autocast: true
49
+ first_stage_config:
50
+ params:
51
+ ddconfig:
52
+ attn_resolutions: []
53
+ attn_type: vanilla-xformers
54
+ ch: 128
55
+ ch_mult:
56
+ - 1
57
+ - 2
58
+ - 4
59
+ - 4
60
+ double_z: true
61
+ dropout: 0.0
62
+ in_channels: 3
63
+ num_res_blocks: 2
64
+ out_ch: 3
65
+ resolution: 256
66
+ z_channels: 4
67
+ embed_dim: 4
68
+ lossconfig:
69
+ target: torch.nn.Identity
70
+ monitor: val/rec_loss
71
+ target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper
72
+ network_config:
73
+ params:
74
+ adm_in_channels: 2816
75
+ attention_resolutions:
76
+ - 4
77
+ - 2
78
+ channel_mult:
79
+ - 1
80
+ - 2
81
+ - 4
82
+ context_dim: 2048
83
+ in_channels: 4
84
+ legacy: false
85
+ model_channels: 320
86
+ num_classes: sequential
87
+ num_head_channels: 64
88
+ num_res_blocks: 2
89
+ out_channels: 4
90
+ spatial_transformer_attn_type: softmax-xformers
91
+ transformer_depth:
92
+ - 1
93
+ - 2
94
+ - 10
95
+ use_checkpoint: true
96
+ use_linear_in_transformer: true
97
+ use_spatial_transformer: true
98
+ target: sgm.modules.diffusionmodules.openaimodel.UNetModel
99
+ scale_factor: 0.13025
100
+ target: sgm.models.diffusion.DiffusionEngine
exsswimsuit-06-21_23-59-43-save-1680-120-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8187b2746617d0a4ac59b2ab2d425c1c9c9a5dec66b368782a655d6bd5e5ded0
3
+ size 6938084280
exsswimsuit-06-21_23-59-43-save-1680-120-0.yaml ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ params:
3
+ conditioner_config:
4
+ params:
5
+ emb_models:
6
+ - input_key: txt
7
+ is_trainable: false
8
+ params:
9
+ layer: hidden
10
+ layer_idx: 11
11
+ target: sgm.modules.encoders.modules.FrozenCLIPEmbedder
12
+ - input_key: txt
13
+ is_trainable: false
14
+ params:
15
+ always_return_pooled: true
16
+ arch: ViT-bigG-14
17
+ freeze: true
18
+ layer: penultimate
19
+ legacy: false
20
+ version: laion2b_s39b_b160k
21
+ target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2
22
+ - input_key: original_size_as_tuple
23
+ is_trainable: false
24
+ params:
25
+ outdim: 256
26
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
27
+ - input_key: crop_coords_top_left
28
+ is_trainable: false
29
+ params:
30
+ outdim: 256
31
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
32
+ - input_key: target_size_as_tuple
33
+ is_trainable: false
34
+ params:
35
+ outdim: 256
36
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
37
+ target: sgm.modules.GeneralConditioner
38
+ denoiser_config:
39
+ params:
40
+ discretization_config:
41
+ target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization
42
+ num_idx: 1000
43
+ scaling_config:
44
+ target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling
45
+ weighting_config:
46
+ target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting
47
+ target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser
48
+ disable_first_stage_autocast: true
49
+ first_stage_config:
50
+ params:
51
+ ddconfig:
52
+ attn_resolutions: []
53
+ attn_type: vanilla-xformers
54
+ ch: 128
55
+ ch_mult:
56
+ - 1
57
+ - 2
58
+ - 4
59
+ - 4
60
+ double_z: true
61
+ dropout: 0.0
62
+ in_channels: 3
63
+ num_res_blocks: 2
64
+ out_ch: 3
65
+ resolution: 256
66
+ z_channels: 4
67
+ embed_dim: 4
68
+ lossconfig:
69
+ target: torch.nn.Identity
70
+ monitor: val/rec_loss
71
+ target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper
72
+ network_config:
73
+ params:
74
+ adm_in_channels: 2816
75
+ attention_resolutions:
76
+ - 4
77
+ - 2
78
+ channel_mult:
79
+ - 1
80
+ - 2
81
+ - 4
82
+ context_dim: 2048
83
+ in_channels: 4
84
+ legacy: false
85
+ model_channels: 320
86
+ num_classes: sequential
87
+ num_head_channels: 64
88
+ num_res_blocks: 2
89
+ out_channels: 4
90
+ spatial_transformer_attn_type: softmax-xformers
91
+ transformer_depth:
92
+ - 1
93
+ - 2
94
+ - 10
95
+ use_checkpoint: true
96
+ use_linear_in_transformer: true
97
+ use_spatial_transformer: true
98
+ target: sgm.modules.diffusionmodules.openaimodel.UNetModel
99
+ scale_factor: 0.13025
100
+ target: sgm.models.diffusion.DiffusionEngine
exsswimsuit-06-22_00-10-40-save-2100-150-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:634785ab3dc1ec7baab100bfc81e91f48cae5d10e15907d7f583d4f0a08506d4
3
+ size 6938084280
exsswimsuit-06-22_00-10-40-save-2100-150-0.yaml ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ params:
3
+ conditioner_config:
4
+ params:
5
+ emb_models:
6
+ - input_key: txt
7
+ is_trainable: false
8
+ params:
9
+ layer: hidden
10
+ layer_idx: 11
11
+ target: sgm.modules.encoders.modules.FrozenCLIPEmbedder
12
+ - input_key: txt
13
+ is_trainable: false
14
+ params:
15
+ always_return_pooled: true
16
+ arch: ViT-bigG-14
17
+ freeze: true
18
+ layer: penultimate
19
+ legacy: false
20
+ version: laion2b_s39b_b160k
21
+ target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2
22
+ - input_key: original_size_as_tuple
23
+ is_trainable: false
24
+ params:
25
+ outdim: 256
26
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
27
+ - input_key: crop_coords_top_left
28
+ is_trainable: false
29
+ params:
30
+ outdim: 256
31
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
32
+ - input_key: target_size_as_tuple
33
+ is_trainable: false
34
+ params:
35
+ outdim: 256
36
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
37
+ target: sgm.modules.GeneralConditioner
38
+ denoiser_config:
39
+ params:
40
+ discretization_config:
41
+ target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization
42
+ num_idx: 1000
43
+ scaling_config:
44
+ target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling
45
+ weighting_config:
46
+ target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting
47
+ target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser
48
+ disable_first_stage_autocast: true
49
+ first_stage_config:
50
+ params:
51
+ ddconfig:
52
+ attn_resolutions: []
53
+ attn_type: vanilla-xformers
54
+ ch: 128
55
+ ch_mult:
56
+ - 1
57
+ - 2
58
+ - 4
59
+ - 4
60
+ double_z: true
61
+ dropout: 0.0
62
+ in_channels: 3
63
+ num_res_blocks: 2
64
+ out_ch: 3
65
+ resolution: 256
66
+ z_channels: 4
67
+ embed_dim: 4
68
+ lossconfig:
69
+ target: torch.nn.Identity
70
+ monitor: val/rec_loss
71
+ target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper
72
+ network_config:
73
+ params:
74
+ adm_in_channels: 2816
75
+ attention_resolutions:
76
+ - 4
77
+ - 2
78
+ channel_mult:
79
+ - 1
80
+ - 2
81
+ - 4
82
+ context_dim: 2048
83
+ in_channels: 4
84
+ legacy: false
85
+ model_channels: 320
86
+ num_classes: sequential
87
+ num_head_channels: 64
88
+ num_res_blocks: 2
89
+ out_channels: 4
90
+ spatial_transformer_attn_type: softmax-xformers
91
+ transformer_depth:
92
+ - 1
93
+ - 2
94
+ - 10
95
+ use_checkpoint: true
96
+ use_linear_in_transformer: true
97
+ use_spatial_transformer: true
98
+ target: sgm.modules.diffusionmodules.openaimodel.UNetModel
99
+ scale_factor: 0.13025
100
+ target: sgm.models.diffusion.DiffusionEngine
exsswimsuit-06-22_00-21-52-save-2520-180-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb6d608dd5be29690ab4b6c92a45d4ca6d9cfcff654422177ad4281d32c111f2
3
+ size 6938084280
exsswimsuit-06-22_00-21-52-save-2520-180-0.yaml ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ params:
3
+ conditioner_config:
4
+ params:
5
+ emb_models:
6
+ - input_key: txt
7
+ is_trainable: false
8
+ params:
9
+ layer: hidden
10
+ layer_idx: 11
11
+ target: sgm.modules.encoders.modules.FrozenCLIPEmbedder
12
+ - input_key: txt
13
+ is_trainable: false
14
+ params:
15
+ always_return_pooled: true
16
+ arch: ViT-bigG-14
17
+ freeze: true
18
+ layer: penultimate
19
+ legacy: false
20
+ version: laion2b_s39b_b160k
21
+ target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2
22
+ - input_key: original_size_as_tuple
23
+ is_trainable: false
24
+ params:
25
+ outdim: 256
26
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
27
+ - input_key: crop_coords_top_left
28
+ is_trainable: false
29
+ params:
30
+ outdim: 256
31
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
32
+ - input_key: target_size_as_tuple
33
+ is_trainable: false
34
+ params:
35
+ outdim: 256
36
+ target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND
37
+ target: sgm.modules.GeneralConditioner
38
+ denoiser_config:
39
+ params:
40
+ discretization_config:
41
+ target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization
42
+ num_idx: 1000
43
+ scaling_config:
44
+ target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling
45
+ weighting_config:
46
+ target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting
47
+ target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser
48
+ disable_first_stage_autocast: true
49
+ first_stage_config:
50
+ params:
51
+ ddconfig:
52
+ attn_resolutions: []
53
+ attn_type: vanilla-xformers
54
+ ch: 128
55
+ ch_mult:
56
+ - 1
57
+ - 2
58
+ - 4
59
+ - 4
60
+ double_z: true
61
+ dropout: 0.0
62
+ in_channels: 3
63
+ num_res_blocks: 2
64
+ out_ch: 3
65
+ resolution: 256
66
+ z_channels: 4
67
+ embed_dim: 4
68
+ lossconfig:
69
+ target: torch.nn.Identity
70
+ monitor: val/rec_loss
71
+ target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper
72
+ network_config:
73
+ params:
74
+ adm_in_channels: 2816
75
+ attention_resolutions:
76
+ - 4
77
+ - 2
78
+ channel_mult:
79
+ - 1
80
+ - 2
81
+ - 4
82
+ context_dim: 2048
83
+ in_channels: 4
84
+ legacy: false
85
+ model_channels: 320
86
+ num_classes: sequential
87
+ num_head_channels: 64
88
+ num_res_blocks: 2
89
+ out_channels: 4
90
+ spatial_transformer_attn_type: softmax-xformers
91
+ transformer_depth:
92
+ - 1
93
+ - 2
94
+ - 10
95
+ use_checkpoint: true
96
+ use_linear_in_transformer: true
97
+ use_spatial_transformer: true
98
+ target: sgm.modules.diffusionmodules.openaimodel.UNetModel
99
+ scale_factor: 0.13025
100
+ target: sgm.models.diffusion.DiffusionEngine