vae/config.json
| 1 | { |
| 2 | "_class_name": "AutoencoderKLWan", |
| 3 | "_diffusers_version": "0.35.0.dev0", |
| 4 | "attn_scales": [], |
| 5 | "base_dim": 160, |
| 6 | "clip_output": false, |
| 7 | "decoder_base_dim": 256, |
| 8 | "dim_mult": [ |
| 9 | 1, |
| 10 | 2, |
| 11 | 4, |
| 12 | 4 |
| 13 | ], |
| 14 | "dropout": 0.0, |
| 15 | "in_channels": 12, |
| 16 | "is_residual": true, |
| 17 | "latents_mean": [ |
| 18 | -0.2289, |
| 19 | -0.0052, |
| 20 | -0.1323, |
| 21 | -0.2339, |
| 22 | -0.2799, |
| 23 | 0.0174, |
| 24 | 0.1838, |
| 25 | 0.1557, |
| 26 | -0.1382, |
| 27 | 0.0542, |
| 28 | 0.2813, |
| 29 | 0.0891, |
| 30 | 0.157, |
| 31 | -0.0098, |
| 32 | 0.0375, |
| 33 | -0.1825, |
| 34 | -0.2246, |
| 35 | -0.1207, |
| 36 | -0.0698, |
| 37 | 0.5109, |
| 38 | 0.2665, |
| 39 | -0.2108, |
| 40 | -0.2158, |
| 41 | 0.2502, |
| 42 | -0.2055, |
| 43 | -0.0322, |
| 44 | 0.1109, |
| 45 | 0.1567, |
| 46 | -0.0729, |
| 47 | 0.0899, |
| 48 | -0.2799, |
| 49 | -0.123, |
| 50 | -0.0313, |
| 51 | -0.1649, |
| 52 | 0.0117, |
| 53 | 0.0723, |
| 54 | -0.2839, |
| 55 | -0.2083, |
| 56 | -0.052, |
| 57 | 0.3748, |
| 58 | 0.0152, |
| 59 | 0.1957, |
| 60 | 0.1433, |
| 61 | -0.2944, |
| 62 | 0.3573, |
| 63 | -0.0548, |
| 64 | -0.1681, |
| 65 | -0.0667 |
| 66 | ], |
| 67 | "latents_std": [ |
| 68 | 0.4765, |
| 69 | 1.0364, |
| 70 | 0.4514, |
| 71 | 1.1677, |
| 72 | 0.5313, |
| 73 | 0.499, |
| 74 | 0.4818, |
| 75 | 0.5013, |
| 76 | 0.8158, |
| 77 | 1.0344, |
| 78 | 0.5894, |
| 79 | 1.0901, |
| 80 | 0.6885, |
| 81 | 0.6165, |
| 82 | 0.8454, |
| 83 | 0.4978, |
| 84 | 0.5759, |
| 85 | 0.3523, |
| 86 | 0.7135, |
| 87 | 0.6804, |
| 88 | 0.5833, |
| 89 | 1.4146, |
| 90 | 0.8986, |
| 91 | 0.5659, |
| 92 | 0.7069, |
| 93 | 0.5338, |
| 94 | 0.4889, |
| 95 | 0.4917, |
| 96 | 0.4069, |
| 97 | 0.4999, |
| 98 | 0.6866, |
| 99 | 0.4093, |
| 100 | 0.5709, |
| 101 | 0.6065, |
| 102 | 0.6415, |
| 103 | 0.4944, |
| 104 | 0.5726, |
| 105 | 1.2042, |
| 106 | 0.5458, |
| 107 | 1.6887, |
| 108 | 0.3971, |
| 109 | 1.06, |
| 110 | 0.3943, |
| 111 | 0.5537, |
| 112 | 0.5444, |
| 113 | 0.4089, |
| 114 | 0.7468, |
| 115 | 0.7744 |
| 116 | ], |
| 117 | "num_res_blocks": 2, |
| 118 | "out_channels": 12, |
| 119 | "patch_size": 2, |
| 120 | "scale_factor_spatial": 16, |
| 121 | "scale_factor_temporal": 4, |
| 122 | "temperal_downsample": [ |
| 123 | false, |
| 124 | true, |
| 125 | true |
| 126 | ], |
| 127 | "z_dim": 48 |
| 128 | } |
| 129 | |