tolgacangoz
commited on
Commit
•
d69c46d
1
Parent(s):
d2c526f
move files
Browse files- unet/config.json +0 -173
unet/config.json
DELETED
@@ -1,173 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"_class_name": "NestedUNet2DConditionModel",
|
3 |
-
"_diffusers_version": "0.31.0.dev0",
|
4 |
-
"act_fn": "silu",
|
5 |
-
"addition_embed_type": "matryoshka",
|
6 |
-
"addition_embed_type_num_heads": 64,
|
7 |
-
"addition_time_embed_dim": null,
|
8 |
-
"attention_head_dim": 8,
|
9 |
-
"attention_pre_only": false,
|
10 |
-
"attention_type": "default",
|
11 |
-
"block_out_channels": [
|
12 |
-
32,
|
13 |
-
32,
|
14 |
-
64
|
15 |
-
],
|
16 |
-
"center_input_sample": false,
|
17 |
-
"class_embed_type": null,
|
18 |
-
"class_embeddings_concat": false,
|
19 |
-
"conv_in_kernel": 3,
|
20 |
-
"conv_out_kernel": 3,
|
21 |
-
"cross_attention_dim": 2048,
|
22 |
-
"cross_attention_norm": null,
|
23 |
-
"down_block_types": [
|
24 |
-
"DownBlock2D",
|
25 |
-
"DownBlock2D",
|
26 |
-
"DownBlock2D"
|
27 |
-
],
|
28 |
-
"downsample_padding": 1,
|
29 |
-
"dropout": 0.0,
|
30 |
-
"dual_cross_attention": false,
|
31 |
-
"encoder_hid_dim": null,
|
32 |
-
"encoder_hid_dim_type": null,
|
33 |
-
"flip_sin_to_cos": false,
|
34 |
-
"freq_shift": 0,
|
35 |
-
"in_channels": 3,
|
36 |
-
"initialize_inner_with_pretrained": null,
|
37 |
-
"inner_config": {
|
38 |
-
"addition_embed_type": "matryoshka",
|
39 |
-
"block_out_channels": [
|
40 |
-
64,
|
41 |
-
128,
|
42 |
-
256
|
43 |
-
],
|
44 |
-
"cross_attention_dim": 2048,
|
45 |
-
"down_block_types": [
|
46 |
-
"DownBlock2D",
|
47 |
-
"DownBlock2D",
|
48 |
-
"DownBlock2D"
|
49 |
-
],
|
50 |
-
"flip_sin_to_cos": false,
|
51 |
-
"in_channels": 3,
|
52 |
-
"initialize_inner_with_pretrained": null,
|
53 |
-
"inner_config": {
|
54 |
-
"addition_embed_type": "matryoshka",
|
55 |
-
"block_out_channels": [
|
56 |
-
256,
|
57 |
-
512,
|
58 |
-
768
|
59 |
-
],
|
60 |
-
"cross_attention_dim": 2048,
|
61 |
-
"down_block_types": [
|
62 |
-
"DownBlock2D",
|
63 |
-
"CrossAttnDownBlock2D",
|
64 |
-
"CrossAttnDownBlock2D"
|
65 |
-
],
|
66 |
-
"encoder_hid_dim": 2048,
|
67 |
-
"encoder_hid_dim_type": "text_proj",
|
68 |
-
"flip_sin_to_cos": false,
|
69 |
-
"in_channels": 3,
|
70 |
-
"layers_per_block": [
|
71 |
-
2,
|
72 |
-
2,
|
73 |
-
2
|
74 |
-
],
|
75 |
-
"masked_cross_attention": false,
|
76 |
-
"micro_conditioning_scale": 64,
|
77 |
-
"mid_block_type": "UNetMidBlock2DCrossAttn",
|
78 |
-
"nesting": true,
|
79 |
-
"out_channels": 3,
|
80 |
-
"resnet_time_scale_shift": "scale_shift",
|
81 |
-
"temporal_mode": false,
|
82 |
-
"temporal_spatial_ds": false,
|
83 |
-
"time_embedding_dim": 1024,
|
84 |
-
"transformer_layers_per_block": [
|
85 |
-
0,
|
86 |
-
1,
|
87 |
-
5
|
88 |
-
],
|
89 |
-
"up_block_types": [
|
90 |
-
"CrossAttnUpBlock2D",
|
91 |
-
"CrossAttnUpBlock2D",
|
92 |
-
"UpBlock2D"
|
93 |
-
],
|
94 |
-
"use_attention_ffn": true
|
95 |
-
},
|
96 |
-
"layers_per_block": [
|
97 |
-
2,
|
98 |
-
2,
|
99 |
-
1
|
100 |
-
],
|
101 |
-
"masked_cross_attention": true,
|
102 |
-
"micro_conditioning_scale": 256,
|
103 |
-
"mid_block_type": null,
|
104 |
-
"nesting": true,
|
105 |
-
"out_channels": 3,
|
106 |
-
"resnet_time_scale_shift": "scale_shift",
|
107 |
-
"skip_inner_unet_input": false,
|
108 |
-
"skip_normalization": false,
|
109 |
-
"temporal_mode": false,
|
110 |
-
"temporal_spatial_ds": false,
|
111 |
-
"time_embedding_dim": 1024,
|
112 |
-
"transformer_layers_per_block": [
|
113 |
-
0,
|
114 |
-
0,
|
115 |
-
0
|
116 |
-
],
|
117 |
-
"up_block_types": [
|
118 |
-
"UpBlock2D",
|
119 |
-
"UpBlock2D",
|
120 |
-
"UpBlock2D"
|
121 |
-
],
|
122 |
-
"use_attention_ffn": false
|
123 |
-
},
|
124 |
-
"is_temporal": null,
|
125 |
-
"layers_per_block": [
|
126 |
-
2,
|
127 |
-
2,
|
128 |
-
1
|
129 |
-
],
|
130 |
-
"masked_cross_attention": true,
|
131 |
-
"micro_conditioning_scale": 1024,
|
132 |
-
"mid_block_only_cross_attention": null,
|
133 |
-
"mid_block_scale_factor": 1,
|
134 |
-
"mid_block_type": null,
|
135 |
-
"nest_ratio": null,
|
136 |
-
"nesting": false,
|
137 |
-
"norm_eps": 1e-05,
|
138 |
-
"norm_num_groups": 32,
|
139 |
-
"norm_type": "layer_norm",
|
140 |
-
"num_attention_heads": null,
|
141 |
-
"num_class_embeds": null,
|
142 |
-
"only_cross_attention": false,
|
143 |
-
"out_channels": 3,
|
144 |
-
"projection_class_embeddings_input_dim": null,
|
145 |
-
"resnet_out_scale_factor": 1.0,
|
146 |
-
"resnet_skip_time_act": false,
|
147 |
-
"resnet_time_scale_shift": "scale_shift",
|
148 |
-
"reverse_transformer_layers_per_block": null,
|
149 |
-
"sample_size": null,
|
150 |
-
"skip_cond_emb": false,
|
151 |
-
"skip_inner_unet_input": false,
|
152 |
-
"skip_normalization": true,
|
153 |
-
"temporal_mode": false,
|
154 |
-
"temporal_spatial_ds": false,
|
155 |
-
"time_cond_proj_dim": null,
|
156 |
-
"time_embedding_act_fn": null,
|
157 |
-
"time_embedding_dim": 1024,
|
158 |
-
"time_embedding_type": "positional",
|
159 |
-
"timestep_post_act": null,
|
160 |
-
"transformer_layers_per_block": [
|
161 |
-
0,
|
162 |
-
0,
|
163 |
-
0
|
164 |
-
],
|
165 |
-
"up_block_types": [
|
166 |
-
"UpBlock2D",
|
167 |
-
"UpBlock2D",
|
168 |
-
"UpBlock2D"
|
169 |
-
],
|
170 |
-
"upcast_attention": false,
|
171 |
-
"use_attention_ffn": false,
|
172 |
-
"use_linear_projection": false
|
173 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|