GraydientPlatformAPI/grounded-phoenixdown

1{
2  "_class_name": "UNet2DConditionModel",
3  "_diffusers_version": "0.19.3",
4  "act_fn": "silu",
5  "addition_embed_type": null,
6  "addition_embed_type_num_heads": 64,
7  "addition_time_embed_dim": null,
8  "attention_head_dim": 8,
9  "block_out_channels": [
10    320,
11    640,
12    1280,
13    1280
14  ],
15  "center_input_sample": false,
16  "class_embed_type": null,
17  "class_embeddings_concat": false,
18  "conv_in_kernel": 3,
19  "conv_out_kernel": 3,
20  "cross_attention_dim": 768,
21  "cross_attention_norm": null,
22  "down_block_types": [
23    "CrossAttnDownBlock2D",
24    "CrossAttnDownBlock2D",
25    "CrossAttnDownBlock2D",
26    "DownBlock2D"
27  ],
28  "downsample_padding": 1,
29  "dual_cross_attention": false,
30  "encoder_hid_dim": null,
31  "encoder_hid_dim_type": null,
32  "flip_sin_to_cos": true,
33  "freq_shift": 0,
34  "in_channels": 4,
35  "layers_per_block": 2,
36  "mid_block_only_cross_attention": null,
37  "mid_block_scale_factor": 1,
38  "mid_block_type": "UNetMidBlock2DCrossAttn",
39  "norm_eps": 1e-05,
40  "norm_num_groups": 32,
41  "num_attention_heads": null,
42  "num_class_embeds": null,
43  "only_cross_attention": false,
44  "out_channels": 4,
45  "projection_class_embeddings_input_dim": null,
46  "resnet_out_scale_factor": 1.0,
47  "resnet_skip_time_act": false,
48  "resnet_time_scale_shift": "default",
49  "sample_size": 64,
50  "time_cond_proj_dim": null,
51  "time_embedding_act_fn": null,
52  "time_embedding_dim": null,
53  "time_embedding_type": "positional",
54  "timestep_post_act": null,
55  "transformer_layers_per_block": 1,
56  "up_block_types": [
57    "UpBlock2D",
58    "CrossAttnUpBlock2D",
59    "CrossAttnUpBlock2D",
60    "CrossAttnUpBlock2D"
61  ],
62  "upcast_attention": false,
63  "use_linear_projection": false
64}
65