unet/config.json
1.9 KB · 79 lines · json Raw
1 {
2 "_class_name": "UNet2DConditionModel",
3 "_diffusers_version": "0.21.0.dev0",
4 "_name_or_path": "valhalla/sdxl-inpaint-ema",
5 "act_fn": "silu",
6 "addition_embed_type": "text_time",
7 "addition_embed_type_num_heads": 64,
8 "addition_time_embed_dim": 256,
9 "attention_head_dim": [
10 5,
11 10,
12 20
13 ],
14 "attention_type": "default",
15 "block_out_channels": [
16 320,
17 640,
18 1280
19 ],
20 "center_input_sample": false,
21 "class_embed_type": null,
22 "class_embeddings_concat": false,
23 "conv_in_kernel": 3,
24 "conv_out_kernel": 3,
25 "cross_attention_dim": 2048,
26 "cross_attention_norm": null,
27 "decay": 0.9999,
28 "down_block_types": [
29 "DownBlock2D",
30 "CrossAttnDownBlock2D",
31 "CrossAttnDownBlock2D"
32 ],
33 "downsample_padding": 1,
34 "dual_cross_attention": false,
35 "encoder_hid_dim": null,
36 "encoder_hid_dim_type": null,
37 "flip_sin_to_cos": true,
38 "freq_shift": 0,
39 "in_channels": 9,
40 "inv_gamma": 1.0,
41 "layers_per_block": 2,
42 "mid_block_only_cross_attention": null,
43 "mid_block_scale_factor": 1,
44 "mid_block_type": "UNetMidBlock2DCrossAttn",
45 "min_decay": 0.0,
46 "norm_eps": 1e-05,
47 "norm_num_groups": 32,
48 "num_attention_heads": null,
49 "num_class_embeds": null,
50 "only_cross_attention": false,
51 "optimization_step": 37000,
52 "out_channels": 4,
53 "power": 0.6666666666666666,
54 "projection_class_embeddings_input_dim": 2816,
55 "resnet_out_scale_factor": 1.0,
56 "resnet_skip_time_act": false,
57 "resnet_time_scale_shift": "default",
58 "sample_size": 128,
59 "time_cond_proj_dim": null,
60 "time_embedding_act_fn": null,
61 "time_embedding_dim": null,
62 "time_embedding_type": "positional",
63 "timestep_post_act": null,
64 "transformer_layers_per_block": [
65 1,
66 2,
67 10
68 ],
69 "up_block_types": [
70 "CrossAttnUpBlock2D",
71 "CrossAttnUpBlock2D",
72 "UpBlock2D"
73 ],
74 "upcast_attention": null,
75 "update_after_step": 0,
76 "use_ema_warmup": false,
77 "use_linear_projection": true
78 }
79