{
  "_class_name": "NestedUNet2DConditionModel",
  "_diffusers_version": "0.31.0.dev0",
  "act_fn": "silu",
  "addition_embed_type": "matryoshka",
  "addition_embed_type_num_heads": 64,
  "addition_time_embed_dim": null,
  "attention_head_dim": 8,
  "attention_pre_only": false,
  "attention_type": "default",
  "block_out_channels": [
    64,
    128,
    256
  ],
  "center_input_sample": false,
  "class_embed_type": null,
  "class_embeddings_concat": false,
  "conv_in_kernel": 3,
  "conv_out_kernel": 3,
  "cross_attention_dim": 2048,
  "cross_attention_norm": null,
  "down_block_types": [
    "DownBlock2D",
    "DownBlock2D",
    "DownBlock2D"
  ],
  "downsample_padding": 1,
  "dropout": 0.0,
  "dual_cross_attention": false,
  "encoder_hid_dim": null,
  "encoder_hid_dim_type": null,
  "flip_sin_to_cos": false,
  "freq_shift": 0,
  "in_channels": 3,
  "initialize_inner_with_pretrained": null,
  "inner_config": {
    "addition_embed_type": "matryoshka",
    "block_out_channels": [
      256,
      512,
      768
    ],
    "cross_attention_dim": 2048,
    "down_block_types": [
      "DownBlock2D",
      "CrossAttnDownBlock2D",
      "CrossAttnDownBlock2D"
    ],
    "encoder_hid_dim": 2048,
    "encoder_hid_dim_type": "text_proj",
    "flip_sin_to_cos": false,
    "in_channels": 3,
    "layers_per_block": [
      2,
      2,
      2
    ],
    "masked_cross_attention": false,
    "micro_conditioning_scale": 64,
    "mid_block_type": "UNetMidBlock2DCrossAttn",
    "nesting": true,
    "out_channels": 3,
    "resnet_time_scale_shift": "scale_shift",
    "transformer_layers_per_block": [
      0,
      1,
      5
    ],
    "up_block_types": [
      "CrossAttnUpBlock2D",
      "CrossAttnUpBlock2D",
      "UpBlock2D"
    ],
    "use_attention_ffn": true
  },
  "is_temporal": null,
  "layers_per_block": [
    2,
    2,
    1
  ],
  "masked_cross_attention": true,
  "micro_conditioning_scale": 256,
  "mid_block_only_cross_attention": null,
  "mid_block_scale_factor": 1,
  "mid_block_type": null,
  "nesting": false,
  "norm_eps": 1e-05,
  "norm_num_groups": 32,
  "norm_type": "layer_norm",
  "num_attention_heads": null,
  "num_class_embeds": null,
  "only_cross_attention": false,
  "out_channels": 3,
  "projection_class_embeddings_input_dim": null,
  "resnet_out_scale_factor": 1.0,
  "resnet_skip_time_act": false,
  "resnet_time_scale_shift": "scale_shift",
  "reverse_transformer_layers_per_block": null,
  "sample_size": 256,
  "skip_cond_emb": false,
  "skip_inner_unet_input": false,
  "skip_normalization": true,
  "temporal_mode": false,
  "temporal_spatial_ds": false,
  "time_cond_proj_dim": null,
  "time_embedding_act_fn": null,
  "time_embedding_dim": 1024,
  "time_embedding_type": "positional",
  "timestep_post_act": null,
  "transformer_layers_per_block": [
    0,
    0,
    0
  ],
  "up_block_types": [
    "UpBlock2D",
    "UpBlock2D",
    "UpBlock2D"
  ],
  "upcast_attention": false,
  "use_attention_ffn": false,
  "use_linear_projection": false
}