Upload BeitForSemanticSegmentation
Browse files- config.json +98 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "models/dit-base-layout-detection-v4.5",
|
| 3 |
+
"add_fpn": false,
|
| 4 |
+
"architectures": [
|
| 5 |
+
"BeitForSemanticSegmentation"
|
| 6 |
+
],
|
| 7 |
+
"attention_probs_dropout_prob": 0.0,
|
| 8 |
+
"auxiliary_channels": 256,
|
| 9 |
+
"auxiliary_concat_input": false,
|
| 10 |
+
"auxiliary_loss_weight": 0.4,
|
| 11 |
+
"auxiliary_num_convs": 1,
|
| 12 |
+
"drop_path_rate": 0.1,
|
| 13 |
+
"hidden_act": "gelu",
|
| 14 |
+
"hidden_dropout_prob": 0.0,
|
| 15 |
+
"hidden_size": 768,
|
| 16 |
+
"id2label": {
|
| 17 |
+
"0": "Background",
|
| 18 |
+
"1": "Caption",
|
| 19 |
+
"2": "Footnote",
|
| 20 |
+
"3": "Formula",
|
| 21 |
+
"4": "List-item",
|
| 22 |
+
"5": "Page-footer",
|
| 23 |
+
"6": "Page-header",
|
| 24 |
+
"7": "Picture",
|
| 25 |
+
"8": "Section-header",
|
| 26 |
+
"9": "Table",
|
| 27 |
+
"10": "Text",
|
| 28 |
+
"11": "Title"
|
| 29 |
+
},
|
| 30 |
+
"image_size": 640,
|
| 31 |
+
"initializer_range": 0.02,
|
| 32 |
+
"intermediate_size": 3072,
|
| 33 |
+
"label2id": {
|
| 34 |
+
"Background": 0,
|
| 35 |
+
"Caption": 1,
|
| 36 |
+
"Footnote": 2,
|
| 37 |
+
"Formula": 3,
|
| 38 |
+
"List-item": 4,
|
| 39 |
+
"Page-footer": 5,
|
| 40 |
+
"Page-header": 6,
|
| 41 |
+
"Picture": 7,
|
| 42 |
+
"Section-header": 8,
|
| 43 |
+
"Table": 9,
|
| 44 |
+
"Text": 10,
|
| 45 |
+
"Title": 11
|
| 46 |
+
},
|
| 47 |
+
"layer_norm_eps": 1e-12,
|
| 48 |
+
"layer_scale_init_value": 0.1,
|
| 49 |
+
"model_type": "beit",
|
| 50 |
+
"num_attention_heads": 12,
|
| 51 |
+
"num_channels": 3,
|
| 52 |
+
"num_hidden_layers": 12,
|
| 53 |
+
"out_features": [
|
| 54 |
+
"stage3",
|
| 55 |
+
"stage5",
|
| 56 |
+
"stage7",
|
| 57 |
+
"stage11"
|
| 58 |
+
],
|
| 59 |
+
"out_indices": [
|
| 60 |
+
3,
|
| 61 |
+
5,
|
| 62 |
+
7,
|
| 63 |
+
11
|
| 64 |
+
],
|
| 65 |
+
"patch_size": 16,
|
| 66 |
+
"pool_scales": [
|
| 67 |
+
1,
|
| 68 |
+
2,
|
| 69 |
+
3,
|
| 70 |
+
6
|
| 71 |
+
],
|
| 72 |
+
"reshape_hidden_states": true,
|
| 73 |
+
"semantic_loss_ignore_index": -100,
|
| 74 |
+
"stage_names": [
|
| 75 |
+
"stem",
|
| 76 |
+
"stage1",
|
| 77 |
+
"stage2",
|
| 78 |
+
"stage3",
|
| 79 |
+
"stage4",
|
| 80 |
+
"stage5",
|
| 81 |
+
"stage6",
|
| 82 |
+
"stage7",
|
| 83 |
+
"stage8",
|
| 84 |
+
"stage9",
|
| 85 |
+
"stage10",
|
| 86 |
+
"stage11",
|
| 87 |
+
"stage12"
|
| 88 |
+
],
|
| 89 |
+
"torch_dtype": "float32",
|
| 90 |
+
"transformers_version": "4.41.2",
|
| 91 |
+
"use_absolute_position_embeddings": true,
|
| 92 |
+
"use_auxiliary_head": true,
|
| 93 |
+
"use_mask_token": true,
|
| 94 |
+
"use_mean_pooling": true,
|
| 95 |
+
"use_relative_position_bias": false,
|
| 96 |
+
"use_shared_relative_position_bias": false,
|
| 97 |
+
"vocab_size": 8192
|
| 98 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ffd091de2d9c3cfa7dec50715e297c5ff9c5fb556ab6a0594739a2f4b3647b7a
|
| 3 |
+
size 654506240
|