585 lines
23 KiB
Python
585 lines
23 KiB
Python
"""Latent -> RGB preview tables.
|
|
|
|
Linear per-pixel projections (channels x 3, plus bias) so a client can turn a
|
|
raw latent into a rough RGB preview without a VAE. Numeric tables come from
|
|
ComfyUI comfy/latent_formats.py (same VAE families). Layouts: dims 2 = BCHW
|
|
image latents, dims 3 = BCFHW video latents. "reshape" names a packing the
|
|
client must undo first (flux2: 32 channels x 2x2 pixel shuffle).
|
|
"""
|
|
|
|
LATENT_FORMATS = {
|
|
"SD15": {
|
|
"channels": 4, "dims": 2, "spatial": 8, "temporal": 1,
|
|
"reshape": None,
|
|
"bias": None,
|
|
"factors": [
|
|
[0.3512, 0.2297, 0.3227],
|
|
[0.325, 0.4974, 0.235],
|
|
[-0.2829, 0.1762, 0.2721],
|
|
[-0.212, -0.2616, -0.7177],
|
|
],
|
|
},
|
|
"SDXL": {
|
|
"channels": 4, "dims": 2, "spatial": 8, "temporal": 1,
|
|
"reshape": None,
|
|
"bias": [0.1084, -0.0175, -0.0011],
|
|
"factors": [
|
|
[0.3651, 0.4232, 0.4341],
|
|
[-0.2533, -0.0042, 0.1068],
|
|
[0.1076, 0.1111, -0.0362],
|
|
[-0.3165, -0.2492, -0.2188],
|
|
],
|
|
},
|
|
"SD3": {
|
|
"channels": 16, "dims": 2, "spatial": 8, "temporal": 1,
|
|
"reshape": None,
|
|
"bias": [0.2394, 0.2135, 0.1925],
|
|
"factors": [
|
|
[-0.0922, -0.0175, 0.0749],
|
|
[0.0311, 0.0633, 0.0954],
|
|
[0.1994, 0.0927, 0.0458],
|
|
[0.0856, 0.0339, 0.0902],
|
|
[0.0587, 0.0272, -0.0496],
|
|
[-0.0006, 0.1104, 0.0309],
|
|
[0.0978, 0.0306, 0.0427],
|
|
[-0.0042, 0.1038, 0.1358],
|
|
[-0.0194, 0.002, 0.0669],
|
|
[-0.0488, 0.013, -0.0268],
|
|
[0.0922, 0.0988, 0.0951],
|
|
[-0.0278, 0.0524, -0.0542],
|
|
[0.0332, 0.0456, 0.0895],
|
|
[-0.0069, -0.003, -0.081],
|
|
[-0.0596, -0.0465, -0.0293],
|
|
[-0.1448, -0.1463, -0.1189],
|
|
],
|
|
},
|
|
"Flux": {
|
|
"channels": 16, "dims": 2, "spatial": 8, "temporal": 1,
|
|
"reshape": None,
|
|
"bias": [-0.0329, -0.0718, -0.0851],
|
|
"factors": [
|
|
[-0.0346, 0.0244, 0.0681],
|
|
[0.0034, 0.021, 0.0687],
|
|
[0.0275, -0.0668, -0.0433],
|
|
[-0.0174, 0.016, 0.0617],
|
|
[0.0859, 0.0721, 0.0329],
|
|
[0.0004, 0.0383, 0.0115],
|
|
[0.0405, 0.0861, 0.0915],
|
|
[-0.0236, -0.0185, -0.0259],
|
|
[-0.0245, 0.025, 0.118],
|
|
[0.1008, 0.0755, -0.0421],
|
|
[-0.0515, 0.0201, 0.0011],
|
|
[0.0428, -0.0012, -0.0036],
|
|
[0.0817, 0.0765, 0.0749],
|
|
[-0.1264, -0.0522, -0.1103],
|
|
[-0.028, -0.0881, -0.0499],
|
|
[-0.1262, -0.0982, -0.0778],
|
|
],
|
|
},
|
|
"Flux2": {
|
|
"channels": 128, "dims": 2, "spatial": 16, "temporal": 1,
|
|
"reshape": "flux2_2x2",
|
|
"bias": [-0.0329, -0.0718, -0.0851],
|
|
"factors": [
|
|
[0.0058, 0.0113, 0.0073],
|
|
[0.0495, 0.0443, 0.0836],
|
|
[-0.0099, 0.0096, 0.0644],
|
|
[0.2144, 0.3009, 0.3652],
|
|
[0.0166, -0.0039, -0.0054],
|
|
[0.0157, 0.0103, -0.016],
|
|
[-0.0398, 0.0902, -0.0235],
|
|
[-0.0052, 0.0095, 0.0109],
|
|
[-0.3527, -0.2712, -0.1666],
|
|
[-0.0301, -0.0356, -0.018],
|
|
[-0.0107, 0.0078, 0.0013],
|
|
[0.0746, 0.009, -0.0941],
|
|
[0.0156, 0.0169, 0.007],
|
|
[-0.0034, -0.004, -0.0114],
|
|
[0.0032, 0.0181, 0.008],
|
|
[-0.0939, -0.0008, 0.0186],
|
|
[0.0018, 0.0043, 0.0104],
|
|
[0.0284, 0.0056, -0.0127],
|
|
[-0.0024, -0.0022, -0.003],
|
|
[0.1207, -0.0026, 0.0065],
|
|
[0.0128, 0.0101, 0.0142],
|
|
[0.0137, -0.0072, -0.0007],
|
|
[0.0095, 0.0092, -0.0059],
|
|
[0.0, -0.0077, -0.0049],
|
|
[-0.0465, -0.0204, -0.0312],
|
|
[0.0095, 0.0012, -0.0066],
|
|
[0.029, -0.0034, 0.0025],
|
|
[0.022, 0.0169, -0.0048],
|
|
[-0.0332, -0.0457, -0.0468],
|
|
[-0.0085, 0.0389, 0.0609],
|
|
[-0.0076, 0.0003, -0.0043],
|
|
[-0.0111, -0.046, -0.0614],
|
|
],
|
|
},
|
|
"Wan21": {
|
|
"channels": 16, "dims": 3, "spatial": 8, "temporal": 4,
|
|
"reshape": None,
|
|
"bias": [-0.1835, -0.0868, -0.336],
|
|
"factors": [
|
|
[-0.1299, -0.1692, 0.2932],
|
|
[0.0671, 0.0406, 0.0442],
|
|
[0.3568, 0.2548, 0.1747],
|
|
[0.0372, 0.2344, 0.142],
|
|
[0.0313, 0.0189, -0.0328],
|
|
[0.0296, -0.0956, -0.0665],
|
|
[-0.3477, -0.4059, -0.2925],
|
|
[0.0166, 0.1902, 0.1975],
|
|
[-0.0412, 0.0267, -0.1364],
|
|
[-0.1293, 0.074, 0.1636],
|
|
[0.068, 0.3019, 0.1128],
|
|
[0.0032, 0.0581, 0.0639],
|
|
[-0.1251, 0.0927, 0.1699],
|
|
[0.006, -0.0633, 0.0005],
|
|
[0.3477, 0.2275, 0.295],
|
|
[0.1984, 0.0913, 0.1861],
|
|
],
|
|
},
|
|
"Wan22": {
|
|
"channels": 48, "dims": 3, "spatial": 16, "temporal": 4,
|
|
"reshape": None,
|
|
"bias": [0.0317, -0.0878, -0.1388],
|
|
"factors": [
|
|
[0.0119, 0.0103, 0.0046],
|
|
[-0.1062, -0.0504, 0.0165],
|
|
[0.014, 0.0409, 0.0491],
|
|
[-0.0813, -0.0677, 0.0607],
|
|
[0.0656, 0.0851, 0.0808],
|
|
[0.0264, 0.0463, 0.0912],
|
|
[0.0295, 0.0326, 0.059],
|
|
[-0.0244, -0.027, 0.0025],
|
|
[0.0443, -0.0102, 0.0288],
|
|
[-0.0465, -0.009, -0.0205],
|
|
[0.0359, 0.0236, 0.0082],
|
|
[-0.0776, 0.0854, 0.1048],
|
|
[0.0564, 0.0264, 0.0561],
|
|
[0.0006, 0.0594, 0.0418],
|
|
[-0.0319, -0.0542, -0.0637],
|
|
[-0.0268, 0.0024, 0.026],
|
|
[0.0539, 0.0265, 0.0358],
|
|
[-0.0359, -0.0312, -0.0287],
|
|
[-0.0285, -0.1032, -0.1237],
|
|
[0.1041, 0.0537, 0.0622],
|
|
[-0.0086, -0.0374, -0.0051],
|
|
[0.039, 0.067, 0.2863],
|
|
[0.0069, 0.0144, 0.0082],
|
|
[0.0006, -0.0167, 0.0079],
|
|
[0.0313, -0.0574, -0.0232],
|
|
[-0.1454, -0.0902, -0.0481],
|
|
[0.0714, 0.0827, 0.0447],
|
|
[-0.0304, -0.0574, -0.0196],
|
|
[0.0401, 0.0384, 0.0204],
|
|
[-0.0758, -0.0297, -0.0014],
|
|
[0.0568, 0.1307, 0.1372],
|
|
[-0.0055, -0.031, -0.038],
|
|
[0.0239, -0.0305, 0.0325],
|
|
[-0.0663, -0.0673, -0.014],
|
|
[-0.0416, -0.0047, -0.0023],
|
|
[0.0166, 0.0112, -0.0093],
|
|
[-0.0211, 0.0011, 0.0331],
|
|
[0.1833, 0.1466, 0.225],
|
|
[-0.0368, 0.037, 0.0295],
|
|
[-0.3441, -0.3543, -0.2008],
|
|
[-0.0479, -0.0489, -0.042],
|
|
[-0.066, -0.0153, 0.08],
|
|
[-0.0101, 0.0068, 0.0156],
|
|
[-0.069, -0.0452, -0.0927],
|
|
[-0.0145, 0.0041, 0.0015],
|
|
[0.0421, 0.0451, 0.0373],
|
|
[0.0504, -0.0483, -0.0356],
|
|
[-0.0837, 0.0168, 0.0055],
|
|
],
|
|
},
|
|
"LTXV": {
|
|
"channels": 128, "dims": 3, "spatial": 32, "temporal": 8,
|
|
"reshape": None,
|
|
"bias": [-0.0571, -0.1657, -0.2512],
|
|
"factors": [
|
|
[0.011202, -0.00063815, -0.010021],
|
|
[0.086031, 0.065813, 0.00095409],
|
|
[-0.012576, -0.0075734, -0.0040528],
|
|
[0.0094063, -0.0021688, 0.0026093],
|
|
[0.0037636, 0.012765, 0.0091548],
|
|
[0.021024, -0.0052973, 0.0034373],
|
|
[-0.0088896, -0.019703, -0.018761],
|
|
[-0.01316, -0.010523, 0.0019709],
|
|
[-0.0015152, -0.0069891, -0.007581],
|
|
[-0.0017247, 0.0004656, -0.0033839],
|
|
[0.013617, 0.0047077, -0.0020045],
|
|
[0.010256, 0.0077318, 0.013948],
|
|
[-0.016108, -0.0062151, 0.0011561],
|
|
[0.0073407, 0.015628, 0.00044865],
|
|
[0.00095357, -0.0029518, -0.01476],
|
|
[0.019143, 0.010868, 0.012264],
|
|
[0.0044575, 3.6682e-05, -0.0068508],
|
|
[-0.00045681, 0.003257, 0.0077929],
|
|
[0.033902, 0.033405, 0.037454],
|
|
[-0.023001, -0.0024877, -0.0031033],
|
|
[0.050265, 0.038841, 0.033539],
|
|
[-0.0041018, -0.0011095, 0.0015859],
|
|
[-0.12689, -0.13107, -0.21005],
|
|
[0.026276, 0.014189, -0.0035963],
|
|
[-0.0048679, 0.0088486, 0.0078029],
|
|
[-0.001661, -0.0048597, -0.005206],
|
|
[-0.002101, 0.002361, 0.0093796],
|
|
[-0.022482, -0.021305, -0.015087],
|
|
[-0.015753, -0.010646, -0.0065083],
|
|
[-0.0046975, 0.0050288, -0.006739],
|
|
[0.011951, 0.020712, 0.016191],
|
|
[-0.0063704, -0.0084827, -0.0095483],
|
|
[0.007261, -0.0099326, -0.022978],
|
|
[-0.00091904, 0.0062882, 0.009572],
|
|
[-0.037178, -0.037123, -0.056713],
|
|
[-0.13373, -0.1072, -0.053801],
|
|
[-0.0053702, 0.0081256, 0.0088397],
|
|
[-0.15247, -0.21437, -0.21843],
|
|
[0.031441, 0.0070335, -0.0097541],
|
|
[0.0021528, -0.0089817, -0.021023],
|
|
[0.0038461, -0.0058957, -0.015014],
|
|
[-0.004347, -0.01294, -0.015972],
|
|
[-0.0054781, -0.010842, -0.0030204],
|
|
[-0.0065347, 0.0030806, -0.010163],
|
|
[-0.0050414, -0.0071503, -0.00089686],
|
|
[-0.0085851, -0.0024351, 0.0010674],
|
|
[-0.0090016, -0.0096493, 0.0015692],
|
|
[0.0050914, 0.012099, 0.019968],
|
|
[0.013758, 0.011669, 0.0081958],
|
|
[-0.010518, -0.011575, -0.0041307],
|
|
[-0.02841, -0.031266, -0.022149],
|
|
[0.0029336, 0.036511, 0.018717],
|
|
[-0.016703, -0.016696, -0.0044529],
|
|
[0.048818, 0.040063, 0.008741],
|
|
[-0.015066, -0.00057328, 0.0029785],
|
|
[-0.017613, -0.0081034, 0.013086],
|
|
[-0.0092633, 0.010803, -0.0063489],
|
|
[0.0030851, 0.0004775, 0.012347],
|
|
[-0.022785, -0.023043, -0.026005],
|
|
[-0.024787, -0.015389, -0.022104],
|
|
[-0.023572, 0.0010544, 0.012361],
|
|
[-0.0078915, -0.0012271, -0.0060968],
|
|
[-0.011478, -0.0012543, 0.0062679],
|
|
[-0.054229, 0.026644, 0.0063394],
|
|
[0.0044216, -0.0073338, -0.010464],
|
|
[-0.0045013, 0.0016082, 0.01442],
|
|
[0.013673, 0.0088877, 0.0041253],
|
|
[-0.010145, 0.0090072, 0.015695],
|
|
[-0.0056234, 0.0011847, 0.0081261],
|
|
[-0.0037171, -0.0053538, 0.001259],
|
|
[0.029476, 0.021424, 0.030424],
|
|
[-0.034925, -0.02434, -0.025316],
|
|
[-0.034127, -0.022406, -0.010589],
|
|
[-0.017342, -0.013249, -0.010719],
|
|
[-0.0021478, -0.0086051, -0.0029878],
|
|
[0.0012089, -0.0042391, -0.0068569],
|
|
[0.00090411, -0.0066886, -6.7547e-05],
|
|
[0.016048, -0.010057, -0.028929],
|
|
[0.001229, 0.010163, 0.018861],
|
|
[0.017264, 0.00027257, 0.013785],
|
|
[-0.013482, -0.0036427, 0.00067481],
|
|
[0.0046782, -0.0052423, 0.0024467],
|
|
[-0.0059113, -0.0062244, -0.0018162],
|
|
[0.015496, 0.014582, 0.0019514],
|
|
[0.0074958, 0.0015886, -0.0082305],
|
|
[0.019086, 0.001636, -0.0039674],
|
|
[-0.0057021, -0.0027307, -0.0041066],
|
|
[0.001745, 0.014602, 0.025794],
|
|
[-0.00082788, 0.0022902, 0.0045161],
|
|
[0.011632, 0.0089193, -0.0072813],
|
|
[0.0075721, 0.0026784, 0.011393],
|
|
[0.0051939, 0.0036903, 0.014049],
|
|
[-0.018383, -0.022529, -0.024477],
|
|
[0.00058842, -0.0057874, -0.01477],
|
|
[-0.016125, -0.0086101, -0.014533],
|
|
[0.02054, 0.020729, 0.0064338],
|
|
[0.0033587, -0.011226, -0.016444],
|
|
[-0.0014742, -0.010489, 0.0017097],
|
|
[0.02813, 0.023546, 0.032791],
|
|
[-0.018532, -0.012842, -0.0087756],
|
|
[-0.0080533, -0.010771, -0.017536],
|
|
[-0.0039009, 0.01615, 0.033359],
|
|
[-0.0074554, -0.014154, -0.006191],
|
|
[0.0034734, -0.01137, -0.010581],
|
|
[0.011476, 0.0039281, 0.0028231],
|
|
[0.0071639, -0.0014741, -0.0038066],
|
|
[0.002225, -0.0087552, -0.0095719],
|
|
[0.024146, 0.021696, 0.028056],
|
|
[-0.0054365, -0.024291, -0.017802],
|
|
[0.0074263, 0.01051, 0.012705],
|
|
[0.0062669, 0.0062658, 0.019211],
|
|
[0.016378, 0.0094933, 0.0066971],
|
|
[0.017173, 0.023601, 0.023296],
|
|
[-0.014568, -0.0098279, -0.011556],
|
|
[0.014431, 0.01443, 0.0066362],
|
|
[-0.006823, 0.018863, 0.014555],
|
|
[0.0061156, 0.00347, -0.0026662],
|
|
[-0.0026983, -0.0059402, -0.0092276],
|
|
[0.010235, 0.0074173, -0.0076243],
|
|
[-0.013255, 0.019322, -0.00092153],
|
|
[0.0024222, -0.0048039, -0.015759],
|
|
[0.026244, 0.025951, 0.020249],
|
|
[0.015711, 0.018498, 0.0027407],
|
|
[-0.0021714, 0.0047214, -0.022443],
|
|
[-0.0074747, 0.0074166, 0.01443],
|
|
[-0.0083906, -0.0079776, 0.0097927],
|
|
[0.038321, 0.0096622, -0.019268],
|
|
[-0.014605, -0.0067032, 0.0039675],
|
|
],
|
|
},
|
|
"LTXAV": {
|
|
"channels": 128, "dims": 3, "spatial": 32, "temporal": 8,
|
|
"reshape": None,
|
|
"bias": [-0.347892, -0.363814, -0.370287],
|
|
"factors": [
|
|
[0.001135, -0.010555, -0.004925],
|
|
[-0.008019, -0.006231, -0.005564],
|
|
[0.012637, 0.005605, 0.012713],
|
|
[0.023454, 0.020771, 0.017844],
|
|
[-0.01194, -0.000932, 0.009292],
|
|
[0.018602, 0.011018, 0.013969],
|
|
[-0.036369, -0.046631, -0.057898],
|
|
[-0.031919, 0.000131, 0.015214],
|
|
[0.014519, 0.021041, 0.015325],
|
|
[0.018889, 0.016149, -0.002836],
|
|
[-0.003784, -0.006057, -0.008195],
|
|
[0.013262, 0.030259, 0.029775],
|
|
[0.050465, 0.050366, 0.025255],
|
|
[0.018628, 0.007691, 0.002893],
|
|
[-0.015698, -0.008451, -0.000676],
|
|
[-0.0136, -0.012587, -0.004437],
|
|
[0.012482, 0.021469, 0.027913],
|
|
[-0.018241, -0.013488, -0.010975],
|
|
[0.013828, 0.012568, 0.021984],
|
|
[0.017911, 0.006552, 0.005567],
|
|
[0.026769, 0.006803, -0.00936],
|
|
[-0.006794, -0.008447, -0.013921],
|
|
[0.029708, 0.018671, 0.022811],
|
|
[-0.014732, -0.019169, 0.000903],
|
|
[0.019607, 0.032595, 0.053409],
|
|
[-0.003721, 0.003976, 0.010364],
|
|
[-0.020193, -0.026076, -0.036068],
|
|
[-0.002328, 0.006527, 0.013052],
|
|
[0.017171, 0.009224, 0.006548],
|
|
[0.001104, -0.000591, 0.000147],
|
|
[-0.000217, 0.011834, 0.017945],
|
|
[-0.015329, -0.012463, -0.006178],
|
|
[-0.009478, -0.00868, -0.004107],
|
|
[-0.005565, -0.006006, -0.001493],
|
|
[0.009451, 0.008794, 0.013207],
|
|
[-0.009989, -0.008027, -0.009568],
|
|
[-0.001505, -0.008805, -0.006828],
|
|
[0.001105, 0.008999, 0.009079],
|
|
[0.025935, 0.016426, 0.008036],
|
|
[0.006313, 0.000694, -0.006039],
|
|
[-0.001893, -0.006951, -0.00956],
|
|
[-0.007082, -0.002566, -0.007152],
|
|
[-0.005231, 0.004829, 0.00822],
|
|
[-0.004333, 0.001251, -0.004852],
|
|
[-0.017024, -0.01273, -0.007457],
|
|
[0.024988, 0.032963, 0.036556],
|
|
[0.013697, 0.012278, 0.009979],
|
|
[-0.013751, -0.008369, -0.015446],
|
|
[-0.009348, -0.001047, 0.007622],
|
|
[-0.003135, -0.00335, -0.003766],
|
|
[0.007436, 0.004957, 0.01048],
|
|
[0.018315, 0.022066, 0.021104],
|
|
[-0.005621, -0.00677, -0.008219],
|
|
[-0.007427, 0.001911, -0.001231],
|
|
[-0.007413, 0.000486, -0.006039],
|
|
[-0.014698, -0.00716, 0.006509],
|
|
[0.013775, 0.014185, 0.008203],
|
|
[0.060246, 0.069787, 0.072833],
|
|
[0.009861, 0.00487, 0.001194],
|
|
[-0.00366, 0.003251, 0.008015],
|
|
[0.003696, -0.00368, -0.008851],
|
|
[0.014924, 0.006196, 0.005282],
|
|
[-0.00674, -0.004319, -0.006729],
|
|
[0.020635, 0.015163, 0.012385],
|
|
[-0.032623, -0.006105, 0.010436],
|
|
[-0.058988, -0.030162, -0.037961],
|
|
[-0.035614, -0.021929, -0.011062],
|
|
[-0.023412, -0.011305, -0.005054],
|
|
[-0.002716, -0.005184, -0.004084],
|
|
[0.014591, 0.015294, 0.014045],
|
|
[0.00831, 0.002466, -0.003225],
|
|
[0.005176, 0.001119, 0.000695],
|
|
[-0.021569, -0.030886, -0.044732],
|
|
[0.007517, 0.003891, 0.000551],
|
|
[-0.006793, 0.004059, 0.010184],
|
|
[-0.086481, -0.082033, -0.083414],
|
|
[0.004192, 0.000762, -0.008658],
|
|
[0.01097, 0.009002, 0.007384],
|
|
[0.004042, -0.006732, -0.011031],
|
|
[0.012164, 0.006401, 0.007483],
|
|
[0.029252, 0.01399, 0.011128],
|
|
[0.048452, 0.034648, 0.016269],
|
|
[0.024104, 0.012647, 0.011754],
|
|
[-0.013216, -0.020192, -0.019752],
|
|
[-0.010799, -0.008535, -0.005467],
|
|
[0.005823, 0.001403, 0.00189],
|
|
[0.052393, 0.044771, 0.032777],
|
|
[0.007576, -0.00808, -0.012453],
|
|
[0.00983, 0.004244, 0.001213],
|
|
[-0.025867, -0.013169, -0.010636],
|
|
[0.008494, 0.003135, 0.00079],
|
|
[0.003969, -0.002625, -0.010204],
|
|
[0.006509, 0.008272, 0.020819],
|
|
[-0.004943, -0.013424, -0.015351],
|
|
[0.005541, 0.009136, -0.003666],
|
|
[-0.0143, -0.015864, -0.016853],
|
|
[0.00265, 0.028393, 0.014125],
|
|
[-0.027661, -0.045422, -0.064995],
|
|
[0.00922, 0.015522, 0.010574],
|
|
[-0.002236, 0.002915, 0.004557],
|
|
[-0.020269, -0.008212, -0.000532],
|
|
[0.019294, 0.003655, -0.002809],
|
|
[0.007116, -0.002784, 1.7e-05],
|
|
[0.057277, 0.07327, 0.074401],
|
|
[-0.002616, -0.001696, -0.000498],
|
|
[0.007248, 0.009793, 0.022829],
|
|
[-0.00259, -0.005601, -0.000436],
|
|
[-0.007681, 0.003893, -0.004119],
|
|
[-0.057392, -0.045545, -0.02529],
|
|
[0.045188, 0.047985, 0.054059],
|
|
[0.000937, -0.008861, -0.038406],
|
|
[-0.010192, -0.008036, -0.005385],
|
|
[-0.030222, -0.027498, -0.030765],
|
|
[-0.008359, 0.013247, 0.010918],
|
|
[0.004102, 0.002093, 0.006934],
|
|
[0.039461, 0.027339, 0.008284],
|
|
[-0.075747, -0.07634, -0.071625],
|
|
[0.002692, 0.005096, -0.002247],
|
|
[-0.002453, -0.002785, -0.010483],
|
|
[0.012265, 0.005481, 0.001729],
|
|
[0.017755, 0.008655, 0.003532],
|
|
[0.05556, 0.049128, 0.044137],
|
|
[-0.025861, -0.023798, -0.018815],
|
|
[-0.014876, -0.01077, -0.010713],
|
|
[-0.017315, -0.012599, -0.008661],
|
|
[-0.008461, -0.00621, -0.007744],
|
|
[-0.040175, -0.042255, -0.048119],
|
|
[-0.019355, -0.021055, -0.021919],
|
|
],
|
|
},
|
|
"MiniMaxH3Video": {
|
|
"channels": 24, "dims": 3, "spatial": 16, "temporal": 4,
|
|
"reshape": None,
|
|
"bias": [0.057426, -0.022078, -0.071449],
|
|
"factors": [
|
|
[-0.018555, 0.024344, -0.017536],
|
|
[0.150164, 0.137244, 0.129221],
|
|
[0.027367, -0.050369, -0.208606],
|
|
[-0.000793, -0.164622, -0.323161],
|
|
[-0.048556, 0.01397, -0.074286],
|
|
[0.01174, 0.014172, -0.006906],
|
|
[0.061517, 0.061212, 0.110025],
|
|
[0.035321, 0.086879, 0.110059],
|
|
[-0.017426, 0.002997, 0.035356],
|
|
[0.531539, 0.548819, 0.624404],
|
|
[-0.024968, -0.040234, -0.034302],
|
|
[-0.032549, -0.029096, -0.017221],
|
|
[0.022609, 0.020286, 0.050661],
|
|
[-0.084001, -0.038131, -0.020805],
|
|
[-0.01883, 0.010412, 0.06112],
|
|
[0.020777, 0.011196, -0.030994],
|
|
[-0.00839, -0.012201, -0.025687],
|
|
[-0.013281, -0.002924, 0.006331],
|
|
[0.00026, 0.001833, -0.011038],
|
|
[0.105471, 0.100482, 0.132106],
|
|
[0.016529, 0.015213, 0.009999],
|
|
[-0.014015, -0.017438, -0.019134],
|
|
[-0.033787, -0.009984, -0.019725],
|
|
[0.004224, 0.017284, 0.027196],
|
|
],
|
|
},
|
|
"HunyuanVideo": {
|
|
"channels": 16, "dims": 3, "spatial": 8, "temporal": 4,
|
|
"reshape": None,
|
|
"bias": [0.0259, -0.0192, -0.0761],
|
|
"factors": [
|
|
[-0.0395, -0.0331, 0.0445],
|
|
[0.0696, 0.0795, 0.0518],
|
|
[0.0135, -0.0945, -0.0282],
|
|
[0.0108, -0.025, -0.0765],
|
|
[-0.0209, 0.0032, 0.0224],
|
|
[-0.0804, -0.0254, -0.0639],
|
|
[-0.0991, 0.0271, -0.0669],
|
|
[-0.0646, -0.0422, -0.04],
|
|
[-0.0696, -0.0595, -0.0894],
|
|
[-0.0799, -0.0208, -0.0375],
|
|
[0.1166, 0.1627, 0.0962],
|
|
[0.1165, 0.0432, 0.0407],
|
|
[-0.2315, -0.192, -0.1355],
|
|
[-0.027, 0.0401, -0.0821],
|
|
[-0.0616, -0.0997, -0.0727],
|
|
[0.0249, -0.0469, -0.1703],
|
|
],
|
|
},
|
|
"Cosmos1CV8x8x8": {
|
|
"channels": 16, "dims": 3, "spatial": 8, "temporal": 8,
|
|
"reshape": None,
|
|
"bias": [-0.1223, -0.1889, -0.1976],
|
|
"factors": [
|
|
[0.1817, 0.2284, 0.2423],
|
|
[-0.0586, -0.0862, -0.3108],
|
|
[-0.4703, -0.4255, -0.3995],
|
|
[0.0803, 0.1963, 0.1001],
|
|
[-0.082, -0.105, 0.04],
|
|
[0.2511, 0.3098, 0.2787],
|
|
[-0.183, -0.2117, -0.004],
|
|
[-0.0621, -0.2187, -0.0939],
|
|
[0.3619, 0.1082, 0.1455],
|
|
[0.3164, 0.3922, 0.2575],
|
|
[0.1152, 0.0231, -0.0462],
|
|
[-0.1434, -0.3609, -0.3665],
|
|
[0.0635, 0.1471, 0.168],
|
|
[-0.3635, -0.1963, -0.3248],
|
|
[-0.1865, 0.0365, 0.2346],
|
|
[0.0447, 0.0994, 0.0881],
|
|
],
|
|
},
|
|
"Flux3Image": {
|
|
# FLUX.3 image VAE (FLUX.2 packed f8 autoencoder, BN-normalized): 32 ch x 2x2 shuffle.
|
|
# least-squares fit on flux3-small latents (2026-09-12), r2 ~ [0.74, 0.76, 0.76]
|
|
"channels": 128, "dims": 2, "spatial": 16, "temporal": 1,
|
|
"reshape": "flux2_2x2",
|
|
"bias": [-0.01281, -0.10207, -0.2469],
|
|
"factors": [
|
|
[0.01443, 0.01462, 0.02385],
|
|
[-0.01563, -0.05629, 0.03518],
|
|
[-0.00366, 0.01501, 0.0998],
|
|
[0.17186, 0.27951, 0.40425],
|
|
[0.05145, -0.01879, -0.02143],
|
|
[0.01929, -0.00069, -0.03722],
|
|
[-0.0942, 0.11153, -0.09854],
|
|
[0.03436, 0.01086, -0.00353],
|
|
[-0.33485, -0.22427, -0.04648],
|
|
[0.01356, -0.00215, 0.03131],
|
|
[-0.00078, 0.0166, -0.01035],
|
|
[0.05476, 0.04449, -0.06473],
|
|
[0.00259, -0.00427, -0.01085],
|
|
[0.02877, 0.00309, -0.02534],
|
|
[-0.01333, 0.0106, 0.01833],
|
|
[-0.0832, 0.09934, 0.07942],
|
|
[-0.00951, -0.00139, 0.0027],
|
|
[0.01677, -0.01178, -0.02627],
|
|
[-0.01436, -0.02519, -0.01887],
|
|
[0.13145, -0.0941, 0.02256],
|
|
[-0.0128, -0.00109, 0.01468],
|
|
[0.03173, -0.01365, 0.01746],
|
|
[-0.05829, 0.01014, -0.02845],
|
|
[-0.02415, 0.01729, 0.03453],
|
|
[0.0132, 0.03026, 0.00655],
|
|
[0.004, 0.00867, -0.01574],
|
|
[0.07235, -0.00765, 0.03589],
|
|
[0.05798, 0.01141, 0.00962],
|
|
[-0.01311, -0.02256, -0.04102],
|
|
[-0.03182, -0.02579, 0.00077],
|
|
[0.02911, 0.01766, -0.01212],
|
|
[0.01937, -0.00168, -0.02601],
|
|
],
|
|
},
|
|
}
|