Hugging Face's logo

Wawaworker
/
mzskgl

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
+824
-0
Files changed (16) hide show
  1. mzskgl2024-06-24_02-29-11-save-600-60-0.safetensors +3 -0
  2. mzskgl2024-06-24_02-29-11-save-600-60-0.yaml +100 -0
  3. mzskgl2024-06-24_02-38-29-save-900-90-0.safetensors +3 -0
  4. mzskgl2024-06-24_02-38-29-save-900-90-0.yaml +100 -0
  5. mzskgl2024-06-24_02-47-42-save-1200-120-0.safetensors +3 -0
  6. mzskgl2024-06-24_02-47-42-save-1200-120-0.yaml +100 -0
  7. mzskgl2024-06-24_02-56-58-save-1500-150-0.safetensors +3 -0
  8. mzskgl2024-06-24_02-56-58-save-1500-150-0.yaml +100 -0
  9. mzskgl2024-06-24_03-06-22-save-1800-180-0.safetensors +3 -0
  10. mzskgl2024-06-24_03-06-22-save-1800-180-0.yaml +100 -0
  11. mzskgl2024-06-24_03-15-46-save-2100-210-0.safetensors +3 -0
  12. mzskgl2024-06-24_03-15-46-save-2100-210-0.yaml +100 -0
  13. mzskgl2024-06-24_03-25-20-save-2400-240-0.safetensors +3 -0
  14. mzskgl2024-06-24_03-25-20-save-2400-240-0.yaml +100 -0
  15. mzskgl2024-06-24_03-34-48-save-2700-270-0.safetensors +3 -0
  16. mzskgl2024-06-24_03-34-48-save-2700-270-0.yaml +100 -0
mzskgl2024-06-24_02-29-11-save-600-60-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cce3f1afd34f234baef1579da0b333dc63dd21e54203868bed18bd783aade099
3
+ size 6938084280
mzskgl2024-06-24_02-29-11-save-600-60-0.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
mzskgl2024-06-24_02-38-29-save-900-90-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ea7e3309fe5847da177e5025d3cdcac241a824672d6ea5c47d34d210d17a560
3
+ size 6938084280
mzskgl2024-06-24_02-38-29-save-900-90-0.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
mzskgl2024-06-24_02-47-42-save-1200-120-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dea325dfe6f0610f4e6949466df12c17509ff77cdb04fd91e02697bdb668fece
3
+ size 6938084280
mzskgl2024-06-24_02-47-42-save-1200-120-0.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
mzskgl2024-06-24_02-56-58-save-1500-150-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50469766da2b4f0c76c5d8598e65066527bc30a640387f75f33c8a521e2aa9f6
3
+ size 6938084280
mzskgl2024-06-24_02-56-58-save-1500-150-0.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
mzskgl2024-06-24_03-06-22-save-1800-180-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00955de218f8b61a33382f1e7f985416460202a47acf59d39c2d6d53d288ac9e
3
+ size 6938084280
mzskgl2024-06-24_03-06-22-save-1800-180-0.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
mzskgl2024-06-24_03-15-46-save-2100-210-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:374993cfc22f950689cafd335775daa4fb144f07ae3cf6361e2ac9fbbc843e00
3
+ size 6938084280
mzskgl2024-06-24_03-15-46-save-2100-210-0.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
mzskgl2024-06-24_03-25-20-save-2400-240-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:402be8fa858d554acaa1253e283235a305eeb7ad4288a11341a7c6e6169e4048
3
+ size 6938084280
mzskgl2024-06-24_03-25-20-save-2400-240-0.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
mzskgl2024-06-24_03-34-48-save-2700-270-0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5b56777a15824e01c0634ac8c4daf783635af5e2cb9b1604e857c19306c9674
3
+ size 6938084280
mzskgl2024-06-24_03-34-48-save-2700-270-0.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