Upload unet/config.json with huggingface_hub
Browse files- unet/config.json +102 -0
unet/config.json
ADDED
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_class_name": "UNet2DConditionModel",
|
3 |
+
"_commit_hash": null,
|
4 |
+
"_diffusers_version": "0.23.0",
|
5 |
+
"_name_or_path": "latent-consistency/lcm-sdxl",
|
6 |
+
"act_fn": "silu",
|
7 |
+
"addition_embed_type": "text_time",
|
8 |
+
"addition_embed_type_num_heads": 64,
|
9 |
+
"addition_time_embed_dim": 256,
|
10 |
+
"attention_head_dim": [
|
11 |
+
5,
|
12 |
+
10,
|
13 |
+
20
|
14 |
+
],
|
15 |
+
"attention_type": "default",
|
16 |
+
"block_out_channels": [
|
17 |
+
320,
|
18 |
+
640,
|
19 |
+
1280
|
20 |
+
],
|
21 |
+
"center_input_sample": false,
|
22 |
+
"class_embed_type": null,
|
23 |
+
"class_embeddings_concat": false,
|
24 |
+
"conv_in_kernel": 3,
|
25 |
+
"conv_out_kernel": 3,
|
26 |
+
"cross_attention_dim": 2048,
|
27 |
+
"cross_attention_norm": null,
|
28 |
+
"down_block_types": [
|
29 |
+
"DownBlock2D",
|
30 |
+
"CrossAttnDownBlock2D",
|
31 |
+
"CrossAttnDownBlock2D"
|
32 |
+
],
|
33 |
+
"downsample_padding": 1,
|
34 |
+
"dropout": 0.0,
|
35 |
+
"dual_cross_attention": false,
|
36 |
+
"encoder_hid_dim": null,
|
37 |
+
"encoder_hid_dim_type": null,
|
38 |
+
"flip_sin_to_cos": true,
|
39 |
+
"freq_shift": 0,
|
40 |
+
"in_channels": 4,
|
41 |
+
"layers_per_block": 2,
|
42 |
+
"mid_block_only_cross_attention": null,
|
43 |
+
"mid_block_scale_factor": 1,
|
44 |
+
"mid_block_type": "UNetMidBlock2DCrossAttn",
|
45 |
+
"model_type": "unet",
|
46 |
+
"neuron": {
|
47 |
+
"auto_cast": "matmul",
|
48 |
+
"auto_cast_type": "bf16",
|
49 |
+
"compiler_type": "neuronx-cc",
|
50 |
+
"compiler_version": "2.11.0.34+c5231f848",
|
51 |
+
"disable_fallback": false,
|
52 |
+
"disable_fast_relayout": false,
|
53 |
+
"dynamic_batch_size": false,
|
54 |
+
"input_names": [
|
55 |
+
"sample",
|
56 |
+
"timestep",
|
57 |
+
"encoder_hidden_states",
|
58 |
+
"text_embeds",
|
59 |
+
"time_ids",
|
60 |
+
"timestep_cond"
|
61 |
+
],
|
62 |
+
"output_names": [
|
63 |
+
"sample"
|
64 |
+
],
|
65 |
+
"static_batch_size": 1,
|
66 |
+
"static_height": 128,
|
67 |
+
"static_num_channels": 4,
|
68 |
+
"static_sequence_length": 77,
|
69 |
+
"static_width": 128
|
70 |
+
},
|
71 |
+
"norm_eps": 1e-05,
|
72 |
+
"norm_num_groups": 32,
|
73 |
+
"num_attention_heads": null,
|
74 |
+
"num_class_embeds": null,
|
75 |
+
"only_cross_attention": false,
|
76 |
+
"out_channels": 4,
|
77 |
+
"projection_class_embeddings_input_dim": 2816,
|
78 |
+
"resnet_out_scale_factor": 1.0,
|
79 |
+
"resnet_skip_time_act": false,
|
80 |
+
"resnet_time_scale_shift": "default",
|
81 |
+
"reverse_transformer_layers_per_block": null,
|
82 |
+
"sample_size": 128,
|
83 |
+
"task": "semantic-segmentation",
|
84 |
+
"time_cond_proj_dim": 256,
|
85 |
+
"time_embedding_act_fn": null,
|
86 |
+
"time_embedding_dim": null,
|
87 |
+
"time_embedding_type": "positional",
|
88 |
+
"timestep_post_act": null,
|
89 |
+
"transformer_layers_per_block": [
|
90 |
+
1,
|
91 |
+
2,
|
92 |
+
10
|
93 |
+
],
|
94 |
+
"transformers_version": null,
|
95 |
+
"up_block_types": [
|
96 |
+
"CrossAttnUpBlock2D",
|
97 |
+
"CrossAttnUpBlock2D",
|
98 |
+
"UpBlock2D"
|
99 |
+
],
|
100 |
+
"upcast_attention": null,
|
101 |
+
"use_linear_projection": true
|
102 |
+
}
|