{"_class_name": "AutoencoderKLTurboVAED", "_diffusers_version": "0.32.0.dev0", "decoder_block_out_channels": [32, 64, 256, 512], "decoder_causal": false, "in_channels": 3, "latent_channels": 48, "decoder_layers_per_block": [3, 2, 2, 2, 3], "out_channels": 3, "patch_size": 2, "patch_size_t": 1, "resnet_norm_eps": 1e-06, "scaling_factor": 1.0, "decoder_spatio_temporal_scaling": [false, true, true, true], "decoder_spatio_only": [false, false, true, false], "decoder_is_dw_conv": [false, false, false, true, true], "decoder_dw_kernel_size": 5, "aligned_feature_projection_mode": "conv-2layer", "aligned_feature_projection_dim": [[512, 1024], [512, 1024]], "aligned_blks_indices": [0, 1], "scale_factor_spatial": 16, "scale_factor_temporal": 4}