Commit
•
331d7d7
1
Parent(s):
904e714
Upload unet/config.json with huggingface_hub
Browse files- unet/config.json +111 -0
unet/config.json
ADDED
|
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "UNet2DConditionModel",
|
| 3 |
+
"_commit_hash": null,
|
| 4 |
+
"_diffusers_version": "0.26.3",
|
| 5 |
+
"_name_or_path": "/home/ubuntu/.cache/huggingface/hub/models--SG161222--RealVisXL_V4.0_Lightning/snapshots/2acc78212a1f215f839bda46103df087fa634576/unet",
|
| 6 |
+
"_use_default_values": [
|
| 7 |
+
"attention_type",
|
| 8 |
+
"reverse_transformer_layers_per_block",
|
| 9 |
+
"dropout"
|
| 10 |
+
],
|
| 11 |
+
"act_fn": "silu",
|
| 12 |
+
"addition_embed_type": "text_time",
|
| 13 |
+
"addition_embed_type_num_heads": 64,
|
| 14 |
+
"addition_time_embed_dim": 256,
|
| 15 |
+
"attention_head_dim": [
|
| 16 |
+
5,
|
| 17 |
+
10,
|
| 18 |
+
20
|
| 19 |
+
],
|
| 20 |
+
"attention_type": "default",
|
| 21 |
+
"block_out_channels": [
|
| 22 |
+
320,
|
| 23 |
+
640,
|
| 24 |
+
1280
|
| 25 |
+
],
|
| 26 |
+
"center_input_sample": false,
|
| 27 |
+
"class_embed_type": null,
|
| 28 |
+
"class_embeddings_concat": false,
|
| 29 |
+
"conv_in_kernel": 3,
|
| 30 |
+
"conv_out_kernel": 3,
|
| 31 |
+
"cross_attention_dim": 2048,
|
| 32 |
+
"cross_attention_norm": null,
|
| 33 |
+
"down_block_types": [
|
| 34 |
+
"DownBlock2D",
|
| 35 |
+
"CrossAttnDownBlock2D",
|
| 36 |
+
"CrossAttnDownBlock2D"
|
| 37 |
+
],
|
| 38 |
+
"downsample_padding": 1,
|
| 39 |
+
"dropout": 0.0,
|
| 40 |
+
"dual_cross_attention": false,
|
| 41 |
+
"encoder_hid_dim": null,
|
| 42 |
+
"encoder_hid_dim_type": null,
|
| 43 |
+
"flip_sin_to_cos": true,
|
| 44 |
+
"freq_shift": 0,
|
| 45 |
+
"in_channels": 4,
|
| 46 |
+
"layers_per_block": 2,
|
| 47 |
+
"mid_block_only_cross_attention": null,
|
| 48 |
+
"mid_block_scale_factor": 1,
|
| 49 |
+
"mid_block_type": "UNetMidBlock2DCrossAttn",
|
| 50 |
+
"neuron": {
|
| 51 |
+
"auto_cast": "all",
|
| 52 |
+
"auto_cast_type": "bf16",
|
| 53 |
+
"compiler_type": "neuronx-cc",
|
| 54 |
+
"compiler_version": "2.12.68.0+4480452af",
|
| 55 |
+
"disable_fallback": false,
|
| 56 |
+
"disable_fast_relayout": false,
|
| 57 |
+
"dynamic_batch_size": false,
|
| 58 |
+
"inline_weights_to_neff": true,
|
| 59 |
+
"input_names": [
|
| 60 |
+
"sample",
|
| 61 |
+
"timestep",
|
| 62 |
+
"encoder_hidden_states",
|
| 63 |
+
"text_embeds",
|
| 64 |
+
"time_ids"
|
| 65 |
+
],
|
| 66 |
+
"model_type": "unet",
|
| 67 |
+
"optlevel": "2",
|
| 68 |
+
"output_attentions": false,
|
| 69 |
+
"output_hidden_states": false,
|
| 70 |
+
"output_names": [
|
| 71 |
+
"sample"
|
| 72 |
+
],
|
| 73 |
+
"static_batch_size": 1,
|
| 74 |
+
"static_height": 128,
|
| 75 |
+
"static_num_beams": 1,
|
| 76 |
+
"static_num_channels": 4,
|
| 77 |
+
"static_sequence_length": 77,
|
| 78 |
+
"static_width": 128
|
| 79 |
+
},
|
| 80 |
+
"norm_eps": 1e-05,
|
| 81 |
+
"norm_num_groups": 32,
|
| 82 |
+
"num_attention_heads": null,
|
| 83 |
+
"num_class_embeds": null,
|
| 84 |
+
"only_cross_attention": false,
|
| 85 |
+
"out_channels": 4,
|
| 86 |
+
"projection_class_embeddings_input_dim": 2816,
|
| 87 |
+
"resnet_out_scale_factor": 1.0,
|
| 88 |
+
"resnet_skip_time_act": false,
|
| 89 |
+
"resnet_time_scale_shift": "default",
|
| 90 |
+
"reverse_transformer_layers_per_block": null,
|
| 91 |
+
"sample_size": 128,
|
| 92 |
+
"task": "semantic-segmentation",
|
| 93 |
+
"time_cond_proj_dim": null,
|
| 94 |
+
"time_embedding_act_fn": null,
|
| 95 |
+
"time_embedding_dim": null,
|
| 96 |
+
"time_embedding_type": "positional",
|
| 97 |
+
"timestep_post_act": null,
|
| 98 |
+
"transformer_layers_per_block": [
|
| 99 |
+
1,
|
| 100 |
+
2,
|
| 101 |
+
10
|
| 102 |
+
],
|
| 103 |
+
"transformers_version": null,
|
| 104 |
+
"up_block_types": [
|
| 105 |
+
"CrossAttnUpBlock2D",
|
| 106 |
+
"CrossAttnUpBlock2D",
|
| 107 |
+
"UpBlock2D"
|
| 108 |
+
],
|
| 109 |
+
"upcast_attention": null,
|
| 110 |
+
"use_linear_projection": true
|
| 111 |
+
}
|