{ "model_config": { "_use_default_values": [ "resnet_out_scale_factor", "use_linear_projection", "projection_class_embeddings_input_dim", "resnet_time_scale_shift", "time_embedding_dim", "class_embeddings_concat", "encoder_hid_dim_type", "num_attention_heads", "transformer_layers_per_block", "addition_embed_type", "encoder_hid_dim", "conv_out_kernel", "upcast_attention", "only_cross_attention", "dropout", "timestep_post_act", "dual_cross_attention", "addition_embed_type_num_heads", "time_cond_proj_dim", "num_class_embeds", "time_embedding_act_fn", "attention_type", "class_embed_type", "resnet_skip_time_act", "reverse_transformer_layers_per_block", "mid_block_only_cross_attention", "cross_attention_norm", "mid_block_type", "time_embedding_type", "conv_in_kernel", "addition_time_embed_dim" ], "_class_name": "UNet2DConditionModel", "_diffusers_version": "0.6.0", "_name_or_path": "runwayml/stable-diffusion-v1-5", "sample_size": 64, "in_channels": 4, "out_channels": 4, "center_input_sample": false, "flip_sin_to_cos": true, "freq_shift": 0, "down_block_types": [ "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D" ], "mid_block_type": "UNetMidBlock2DCrossAttn", "up_block_types": [ "UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D" ], "only_cross_attention": false, "block_out_channels": [ 320, 640, 1280, 1280 ], "layers_per_block": 2, "downsample_padding": 1, "mid_block_scale_factor": 1, "dropout": 0.0, "act_fn": "silu", "norm_num_groups": 32, "norm_eps": 1e-05, "cross_attention_dim": 768, "transformer_layers_per_block": 1, "reverse_transformer_layers_per_block": null, "encoder_hid_dim": null, "encoder_hid_dim_type": null, "attention_head_dim": 8, "num_attention_heads": null, "dual_cross_attention": false, "use_linear_projection": false, "class_embed_type": null, "addition_embed_type": null, "addition_time_embed_dim": null, "num_class_embeds": null, "upcast_attention": false, "resnet_time_scale_shift": "default", "resnet_skip_time_act": false, "resnet_out_scale_factor": 1.0, "time_embedding_type": "positional", "time_embedding_dim": null, "time_embedding_act_fn": null, "timestep_post_act": null, "time_cond_proj_dim": null, "conv_in_kernel": 3, "conv_out_kernel": 3, "projection_class_embeddings_input_dim": null, "attention_type": "default", "class_embeddings_concat": false, "mid_block_only_cross_attention": null, "cross_attention_norm": null, "addition_embed_type_num_heads": 64 }, "module_shapes": { "conv_in": { "type": "conv2d", "out_channels": 320, "in_channels": 4, "kernel_size": [ 3, 3 ] }, "time_embedding.linear_1": { "type": "linear", "out_features": 1280, "in_features": 320 }, "time_embedding.linear_2": { "type": "linear", "out_features": 1280, "in_features": 1280 }, "down_blocks.0.attentions.0.proj_in": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 200, "in_features": 320 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 200, "in_features": 768 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 200, "in_features": 768 }, "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 320, "in_features": 200 }, "down_blocks.0.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 2560, "in_features": 320 }, "down_blocks.0.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 320, "in_features": 1280 }, "down_blocks.0.attentions.0.proj_out": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "down_blocks.0.attentions.1.proj_in": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 320, "in_features": 320 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 200, "in_features": 320 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 200, "in_features": 768 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 200, "in_features": 768 }, "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 320, "in_features": 200 }, "down_blocks.0.attentions.1.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 2560, "in_features": 320 }, "down_blocks.0.attentions.1.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 320, "in_features": 1280 }, "down_blocks.0.attentions.1.proj_out": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "down_blocks.0.resnets.0.conv1": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "down_blocks.0.resnets.0.time_emb_proj": { "type": "linear", "out_features": 320, "in_features": 1280 }, "down_blocks.0.resnets.0.conv2": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "down_blocks.0.resnets.1.conv1": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "down_blocks.0.resnets.1.time_emb_proj": { "type": "linear", "out_features": 320, "in_features": 1280 }, "down_blocks.0.resnets.1.conv2": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "down_blocks.0.downsamplers.0.conv": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "down_blocks.1.attentions.0.proj_in": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 400, "in_features": 640 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 400, "in_features": 768 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 400, "in_features": 768 }, "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 640, "in_features": 400 }, "down_blocks.1.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 5120, "in_features": 640 }, "down_blocks.1.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 640, "in_features": 2560 }, "down_blocks.1.attentions.0.proj_out": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "down_blocks.1.attentions.1.proj_in": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 640, "in_features": 640 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 400, "in_features": 640 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 400, "in_features": 768 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 400, "in_features": 768 }, "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 640, "in_features": 400 }, "down_blocks.1.attentions.1.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 5120, "in_features": 640 }, "down_blocks.1.attentions.1.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 640, "in_features": 2560 }, "down_blocks.1.attentions.1.proj_out": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "down_blocks.1.resnets.0.conv1": { "type": "conv2d", "out_channels": 640, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "down_blocks.1.resnets.0.time_emb_proj": { "type": "linear", "out_features": 640, "in_features": 1280 }, "down_blocks.1.resnets.0.conv2": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "down_blocks.1.resnets.0.conv_shortcut": { "type": "conv2d", "out_channels": 640, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "down_blocks.1.resnets.1.conv1": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "down_blocks.1.resnets.1.time_emb_proj": { "type": "linear", "out_features": 640, "in_features": 1280 }, "down_blocks.1.resnets.1.conv2": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "down_blocks.1.downsamplers.0.conv": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "down_blocks.2.attentions.0.proj_in": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 800, "in_features": 1280 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 800, "in_features": 1280 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 800, "in_features": 1280 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 800 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 480, "in_features": 768 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 480, "in_features": 768 }, "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "down_blocks.2.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 10240, "in_features": 1280 }, "down_blocks.2.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 1280, "in_features": 5120 }, "down_blocks.2.attentions.0.proj_out": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "down_blocks.2.attentions.1.proj_in": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 800, "in_features": 1280 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 800, "in_features": 1280 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 800, "in_features": 1280 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 800 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 480, "in_features": 768 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 480, "in_features": 768 }, "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "down_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 10240, "in_features": 1280 }, "down_blocks.2.attentions.1.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 1280, "in_features": 5120 }, "down_blocks.2.attentions.1.proj_out": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "down_blocks.2.resnets.0.conv1": { "type": "conv2d", "out_channels": 1280, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "down_blocks.2.resnets.0.time_emb_proj": { "type": "linear", "out_features": 1280, "in_features": 1280 }, "down_blocks.2.resnets.0.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "down_blocks.2.resnets.0.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "down_blocks.2.resnets.1.conv1": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "down_blocks.2.resnets.1.time_emb_proj": { "type": "linear", "out_features": 1280, "in_features": 1280 }, "down_blocks.2.resnets.1.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "down_blocks.2.downsamplers.0.conv": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "down_blocks.3.resnets.0.conv1": { "type": "conv2d", "out_channels": 480, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "down_blocks.3.resnets.0.time_emb_proj": { "type": "linear", "out_features": 480, "in_features": 1280 }, "down_blocks.3.resnets.0.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 480, "kernel_size": [ 3, 3 ] }, "down_blocks.3.resnets.1.conv1": { "type": "conv2d", "out_channels": 480, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "down_blocks.3.resnets.1.time_emb_proj": { "type": "linear", "out_features": 480, "in_features": 1280 }, "down_blocks.3.resnets.1.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 480, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.0.conv1": { "type": "conv2d", "out_channels": 960, "in_channels": 2560, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.0.time_emb_proj": { "type": "linear", "out_features": 960, "in_features": 1280 }, "up_blocks.0.resnets.0.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 960, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.0.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 1, 1 ] }, "up_blocks.0.resnets.1.conv1": { "type": "conv2d", "out_channels": 960, "in_channels": 2560, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.1.time_emb_proj": { "type": "linear", "out_features": 960, "in_features": 1280 }, "up_blocks.0.resnets.1.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 960, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.1.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 1, 1 ] }, "up_blocks.0.resnets.2.conv1": { "type": "conv2d", "out_channels": 960, "in_channels": 2560, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.2.time_emb_proj": { "type": "linear", "out_features": 960, "in_features": 1280 }, "up_blocks.0.resnets.2.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 960, "kernel_size": [ 3, 3 ] }, "up_blocks.0.resnets.2.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 1, 1 ] }, "up_blocks.0.upsamplers.0.conv": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "up_blocks.1.attentions.0.proj_in": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 800 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 480, "in_features": 768 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 480, "in_features": 768 }, "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "up_blocks.1.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 10240, "in_features": 1280 }, "up_blocks.1.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 1280, "in_features": 5120 }, "up_blocks.1.attentions.0.proj_out": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.1.attentions.1.proj_in": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 800 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 480, "in_features": 768 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 480, "in_features": 768 }, "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "up_blocks.1.attentions.1.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 10240, "in_features": 1280 }, "up_blocks.1.attentions.1.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 1280, "in_features": 5120 }, "up_blocks.1.attentions.1.proj_out": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.1.attentions.2.proj_in": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 800, "in_features": 1280 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 800 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 480, "in_features": 768 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 480, "in_features": 768 }, "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "up_blocks.1.attentions.2.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 10240, "in_features": 1280 }, "up_blocks.1.attentions.2.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 1280, "in_features": 5120 }, "up_blocks.1.attentions.2.proj_out": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.1.resnets.0.conv1": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 3, 3 ] }, "up_blocks.1.resnets.0.time_emb_proj": { "type": "linear", "out_features": 1280, "in_features": 1280 }, "up_blocks.1.resnets.0.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "up_blocks.1.resnets.0.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 1, 1 ] }, "up_blocks.1.resnets.1.conv1": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 3, 3 ] }, "up_blocks.1.resnets.1.time_emb_proj": { "type": "linear", "out_features": 1280, "in_features": 1280 }, "up_blocks.1.resnets.1.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "up_blocks.1.resnets.1.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 2560, "kernel_size": [ 1, 1 ] }, "up_blocks.1.resnets.2.conv1": { "type": "conv2d", "out_channels": 1280, "in_channels": 1920, "kernel_size": [ 3, 3 ] }, "up_blocks.1.resnets.2.time_emb_proj": { "type": "linear", "out_features": 1280, "in_features": 1280 }, "up_blocks.1.resnets.2.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "up_blocks.1.resnets.2.conv_shortcut": { "type": "conv2d", "out_channels": 1280, "in_channels": 1920, "kernel_size": [ 1, 1 ] }, "up_blocks.1.upsamplers.0.conv": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "up_blocks.2.attentions.0.proj_in": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 400, "in_features": 640 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 400, "in_features": 768 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 400, "in_features": 768 }, "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 640, "in_features": 400 }, "up_blocks.2.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 5120, "in_features": 640 }, "up_blocks.2.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 640, "in_features": 2560 }, "up_blocks.2.attentions.0.proj_out": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.2.attentions.1.proj_in": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 400, "in_features": 640 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 400, "in_features": 768 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 400, "in_features": 768 }, "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 640, "in_features": 400 }, "up_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 5120, "in_features": 640 }, "up_blocks.2.attentions.1.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 640, "in_features": 2560 }, "up_blocks.2.attentions.1.proj_out": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.2.attentions.2.proj_in": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 640, "in_features": 640 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 400, "in_features": 640 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 400, "in_features": 768 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 400, "in_features": 768 }, "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 640, "in_features": 400 }, "up_blocks.2.attentions.2.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 5120, "in_features": 640 }, "up_blocks.2.attentions.2.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 640, "in_features": 2560 }, "up_blocks.2.attentions.2.proj_out": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.2.resnets.0.conv1": { "type": "conv2d", "out_channels": 640, "in_channels": 1920, "kernel_size": [ 3, 3 ] }, "up_blocks.2.resnets.0.time_emb_proj": { "type": "linear", "out_features": 640, "in_features": 1280 }, "up_blocks.2.resnets.0.conv2": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "up_blocks.2.resnets.0.conv_shortcut": { "type": "conv2d", "out_channels": 640, "in_channels": 1920, "kernel_size": [ 1, 1 ] }, "up_blocks.2.resnets.1.conv1": { "type": "conv2d", "out_channels": 640, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "up_blocks.2.resnets.1.time_emb_proj": { "type": "linear", "out_features": 640, "in_features": 1280 }, "up_blocks.2.resnets.1.conv2": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "up_blocks.2.resnets.1.conv_shortcut": { "type": "conv2d", "out_channels": 640, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "up_blocks.2.resnets.2.conv1": { "type": "conv2d", "out_channels": 640, "in_channels": 960, "kernel_size": [ 3, 3 ] }, "up_blocks.2.resnets.2.time_emb_proj": { "type": "linear", "out_features": 640, "in_features": 1280 }, "up_blocks.2.resnets.2.conv2": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "up_blocks.2.resnets.2.conv_shortcut": { "type": "conv2d", "out_channels": 640, "in_channels": 960, "kernel_size": [ 1, 1 ] }, "up_blocks.2.upsamplers.0.conv": { "type": "conv2d", "out_channels": 640, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "up_blocks.3.attentions.0.proj_in": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 200, "in_features": 320 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 200, "in_features": 768 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 200, "in_features": 768 }, "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 320, "in_features": 200 }, "up_blocks.3.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 2560, "in_features": 320 }, "up_blocks.3.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 320, "in_features": 1280 }, "up_blocks.3.attentions.0.proj_out": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "up_blocks.3.attentions.1.proj_in": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 200, "in_features": 320 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 200, "in_features": 768 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 200, "in_features": 768 }, "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 320, "in_features": 200 }, "up_blocks.3.attentions.1.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 2560, "in_features": 320 }, "up_blocks.3.attentions.1.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 320, "in_features": 1280 }, "up_blocks.3.attentions.1.proj_out": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "up_blocks.3.attentions.2.proj_in": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 320, "in_features": 320 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 200, "in_features": 320 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 200, "in_features": 768 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 200, "in_features": 768 }, "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 320, "in_features": 200 }, "up_blocks.3.attentions.2.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 2560, "in_features": 320 }, "up_blocks.3.attentions.2.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 320, "in_features": 1280 }, "up_blocks.3.attentions.2.proj_out": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 1, 1 ] }, "up_blocks.3.resnets.0.conv1": { "type": "conv2d", "out_channels": 320, "in_channels": 960, "kernel_size": [ 3, 3 ] }, "up_blocks.3.resnets.0.time_emb_proj": { "type": "linear", "out_features": 320, "in_features": 1280 }, "up_blocks.3.resnets.0.conv2": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "up_blocks.3.resnets.0.conv_shortcut": { "type": "conv2d", "out_channels": 320, "in_channels": 960, "kernel_size": [ 1, 1 ] }, "up_blocks.3.resnets.1.conv1": { "type": "conv2d", "out_channels": 320, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "up_blocks.3.resnets.1.time_emb_proj": { "type": "linear", "out_features": 320, "in_features": 1280 }, "up_blocks.3.resnets.1.conv2": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "up_blocks.3.resnets.1.conv_shortcut": { "type": "conv2d", "out_channels": 320, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "up_blocks.3.resnets.2.conv1": { "type": "conv2d", "out_channels": 320, "in_channels": 640, "kernel_size": [ 3, 3 ] }, "up_blocks.3.resnets.2.time_emb_proj": { "type": "linear", "out_features": 320, "in_features": 1280 }, "up_blocks.3.resnets.2.conv2": { "type": "conv2d", "out_channels": 320, "in_channels": 320, "kernel_size": [ 3, 3 ] }, "up_blocks.3.resnets.2.conv_shortcut": { "type": "conv2d", "out_channels": 320, "in_channels": 640, "kernel_size": [ 1, 1 ] }, "mid_block.attentions.0.proj_in": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "mid_block.attentions.0.transformer_blocks.0.attn1.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "mid_block.attentions.0.transformer_blocks.0.attn1.to_k": { "type": "linear", "out_features": 480, "in_features": 1280 }, "mid_block.attentions.0.transformer_blocks.0.attn1.to_v": { "type": "linear", "out_features": 480, "in_features": 1280 }, "mid_block.attentions.0.transformer_blocks.0.attn1.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "mid_block.attentions.0.transformer_blocks.0.attn2.to_q": { "type": "linear", "out_features": 480, "in_features": 1280 }, "mid_block.attentions.0.transformer_blocks.0.attn2.to_k": { "type": "linear", "out_features": 480, "in_features": 768 }, "mid_block.attentions.0.transformer_blocks.0.attn2.to_v": { "type": "linear", "out_features": 480, "in_features": 768 }, "mid_block.attentions.0.transformer_blocks.0.attn2.to_out.0": { "type": "linear", "out_features": 1280, "in_features": 480 }, "mid_block.attentions.0.transformer_blocks.0.ff.net.0.proj": { "type": "linear", "out_features": 1024, "in_features": 1280 }, "mid_block.attentions.0.transformer_blocks.0.ff.net.2": { "type": "linear", "out_features": 1280, "in_features": 512 }, "mid_block.attentions.0.proj_out": { "type": "conv2d", "out_channels": 1280, "in_channels": 1280, "kernel_size": [ 1, 1 ] }, "mid_block.resnets.0.conv1": { "type": "conv2d", "out_channels": 608, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "mid_block.resnets.0.time_emb_proj": { "type": "linear", "out_features": 608, "in_features": 1280 }, "mid_block.resnets.0.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 608, "kernel_size": [ 3, 3 ] }, "mid_block.resnets.1.conv1": { "type": "conv2d", "out_channels": 608, "in_channels": 1280, "kernel_size": [ 3, 3 ] }, "mid_block.resnets.1.time_emb_proj": { "type": "linear", "out_features": 608, "in_features": 1280 }, "mid_block.resnets.1.conv2": { "type": "conv2d", "out_channels": 1280, "in_channels": 608, "kernel_size": [ 3, 3 ] }, "conv_out": { "type": "conv2d", "out_channels": 4, "in_channels": 320, "kernel_size": [ 3, 3 ] } }, "torch_version": "2.9.1+cu130" }