vae/config.json
2.0 KB · 97 lines · json Raw
1 {
2 "_class_name": "AutoencoderKLMiniMaxH3",
3 "_diffusers_version": "0.36.0.dev0",
4 "in_channels": 3,
5 "out_channels": 3,
6 "latent_channels": 24,
7 "block_out_channels": [
8 128,
9 256,
10 256,
11 512,
12 512,
13 1024
14 ],
15 "layers_per_block": 2,
16 "spatial_downsample_factors": [
17 2,
18 2,
19 2,
20 2,
21 1,
22 1
23 ],
24 "temporal_downsample_factors": [
25 1,
26 2,
27 2,
28 1,
29 1,
30 1
31 ],
32 "norm_num_groups": 32,
33 "norm_eps": 1e-06,
34 "spatial_padding_mode": "reflect",
35 "decoder_num_layers": 36,
36 "decoder_num_attention_heads": 32,
37 "decoder_attention_head_dim": 64,
38 "decoder_num_register_tokens": 4,
39 "decoder_ffn_mult": 4,
40 "decoder_rope_theta": 100.0,
41 "decoder_rope_dim_ratio": 0.75,
42 "decoder_norm_eps": 1e-05,
43 "clip_length": 17,
44 "token_drop": 3,
45 "latents_mean": [
46 0.858090341091156,
47 -0.9606591463088989,
48 1.0661640167236328,
49 -0.5090325474739075,
50 -0.2727581858634949,
51 -1.3675414323806763,
52 -0.2553254961967468,
53 -0.26907554268836975,
54 -0.5376840829849243,
55 -0.0464097298681736,
56 0.6657370328903198,
57 0.19690127670764923,
58 -0.5460608005523682,
59 -0.4035342037677765,
60 -0.23683024942874908,
61 0.25928452610969543,
62 -0.30133944749832153,
63 0.211341992020607,
64 -1.1206848621368408,
65 0.3581933379173279,
66 -0.04225143790245056,
67 0.2604829967021942,
68 0.22864092886447906,
69 0.7056031823158264
70 ],
71 "latents_std": [
72 1.2223774194717407,
73 1.2767263650894165,
74 1.6831774711608887,
75 1.7549455165863037,
76 1.5636216402053833,
77 2.194143533706665,
78 0.9653137922286987,
79 1.0569885969161987,
80 0.841948926448822,
81 0.7729952931404114,
82 1.8955937623977661,
83 0.946841835975647,
84 0.7996809482574463,
85 0.44988900423049927,
86 0.7197399735450745,
87 0.6936293244361877,
88 2.961095094680786,
89 2.7694199085235596,
90 3.0496184825897217,
91 2.1088054180145264,
92 3.276226282119751,
93 3.1627357006073,
94 2.2816812992095947,
95 2.6127843856811523
96 ]
97 }