{ "in_channels": 3, "patch_size": 8, "latent_channels": 32, "enc_width": 1024, "enc_depth": 8, "enc_mlp_ratio": 2.0, "dec_width": 512, "dec_depth": 6, "dec_mlp_ratio": 4.0, "dec_adaln_rank": 128, "dec_skip_blocks": 2, "dec_head_min_channels": 64, "time_embed_dim": 256, "head_dim": 64, "window_radius": 8, "rope_base": 100.0, "latent_mean": [ -4.137261, 2.725146, -1.435803, -2.187023, -4.161275, -1.58918, -2.527523, 3.002302, 2.793705, -2.091501, -2.632896, -3.59589, -3.069669, -4.568713, 3.26668, -0.897648, 2.072541, 5.451158, 1.747871, -2.113284, 2.03444, -1.379274, 0.66369, 2.712183, -0.111181, 6.054838, 1.235881, -2.02777, 3.663243, -2.454632, 1.141335, 1.786136 ], "latent_std": [ 1.030738, 0.951682, 1.079687, 0.957871, 1.198266, 1.102266, 1.213605, 0.947544, 1.085991, 1.093223, 0.9027, 1.010205, 0.920466, 1.044107, 1.023179, 0.992978, 0.991627, 1.073585, 1.114629, 1.056403, 1.093012, 1.015109, 1.007069, 0.976026, 0.982324, 1.093805, 1.036086, 1.005727, 1.104078, 1.042298, 1.095522, 0.960255 ] }