Upload folder using huggingface_hub
Browse files- onnx_surya_O2/config.json +99 -0
- onnx_surya_O2/model.onnx +3 -0
- onnx_surya_O2/ort_config.json +39 -0
- onnx_surya_O2/preprocessor_config.json +39 -0
onnx_surya_O2/config.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "onnx_surya_O2",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"SegformerForRegressionMask"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.0,
|
| 7 |
+
"classifier_dropout_prob": 0.1,
|
| 8 |
+
"decoder_hidden_size": 768,
|
| 9 |
+
"decoder_layer_hidden_size": 192,
|
| 10 |
+
"decoder_upsample_rate": 2,
|
| 11 |
+
"depths": [
|
| 12 |
+
3,
|
| 13 |
+
4,
|
| 14 |
+
9,
|
| 15 |
+
3
|
| 16 |
+
],
|
| 17 |
+
"downsampling_rates": [
|
| 18 |
+
1,
|
| 19 |
+
4,
|
| 20 |
+
8,
|
| 21 |
+
16
|
| 22 |
+
],
|
| 23 |
+
"drop_path_rate": 0.1,
|
| 24 |
+
"hidden_act": "gelu",
|
| 25 |
+
"hidden_dropout_prob": 0.0,
|
| 26 |
+
"hidden_sizes": [
|
| 27 |
+
64,
|
| 28 |
+
128,
|
| 29 |
+
320,
|
| 30 |
+
512
|
| 31 |
+
],
|
| 32 |
+
"id2label": {
|
| 33 |
+
"0": "Blank",
|
| 34 |
+
"1": "Caption",
|
| 35 |
+
"2": "Footnote",
|
| 36 |
+
"3": "Formula",
|
| 37 |
+
"4": "List-item",
|
| 38 |
+
"5": "Page-footer",
|
| 39 |
+
"6": "Page-header",
|
| 40 |
+
"7": "Picture",
|
| 41 |
+
"8": "Section-header",
|
| 42 |
+
"9": "Table",
|
| 43 |
+
"10": "Text",
|
| 44 |
+
"11": "Title"
|
| 45 |
+
},
|
| 46 |
+
"image_size": 224,
|
| 47 |
+
"initializer_range": 0.02,
|
| 48 |
+
"label2id": {
|
| 49 |
+
"Blank": 0,
|
| 50 |
+
"Caption": 1,
|
| 51 |
+
"Footnote": 2,
|
| 52 |
+
"Formula": 3,
|
| 53 |
+
"List-item": 4,
|
| 54 |
+
"Page-footer": 5,
|
| 55 |
+
"Page-header": 6,
|
| 56 |
+
"Picture": 7,
|
| 57 |
+
"Section-header": 8,
|
| 58 |
+
"Table": 9,
|
| 59 |
+
"Text": 10,
|
| 60 |
+
"Title": 11
|
| 61 |
+
},
|
| 62 |
+
"layer_norm_eps": 1e-06,
|
| 63 |
+
"mlp_ratios": [
|
| 64 |
+
4,
|
| 65 |
+
4,
|
| 66 |
+
4,
|
| 67 |
+
4
|
| 68 |
+
],
|
| 69 |
+
"model_type": "segformer",
|
| 70 |
+
"num_attention_heads": [
|
| 71 |
+
1,
|
| 72 |
+
2,
|
| 73 |
+
5,
|
| 74 |
+
8
|
| 75 |
+
],
|
| 76 |
+
"num_channels": 3,
|
| 77 |
+
"num_encoder_blocks": 4,
|
| 78 |
+
"patch_sizes": [
|
| 79 |
+
7,
|
| 80 |
+
3,
|
| 81 |
+
3,
|
| 82 |
+
3
|
| 83 |
+
],
|
| 84 |
+
"reshape_last_stage": true,
|
| 85 |
+
"semantic_loss_ignore_index": -1,
|
| 86 |
+
"sr_ratios": [
|
| 87 |
+
8,
|
| 88 |
+
4,
|
| 89 |
+
2,
|
| 90 |
+
1
|
| 91 |
+
],
|
| 92 |
+
"strides": [
|
| 93 |
+
4,
|
| 94 |
+
2,
|
| 95 |
+
2,
|
| 96 |
+
2
|
| 97 |
+
],
|
| 98 |
+
"transformers_version": "4.39.3"
|
| 99 |
+
}
|
onnx_surya_O2/model.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:652ff26db7c55871663333b6f2482a2127c9b3e0b76222b1ee8f6b10e4f15867
|
| 3 |
+
size 116990389
|
onnx_surya_O2/ort_config.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"one_external_file": true,
|
| 3 |
+
"opset": null,
|
| 4 |
+
"optimization": {
|
| 5 |
+
"disable_attention": null,
|
| 6 |
+
"disable_attention_fusion": false,
|
| 7 |
+
"disable_bias_gelu": null,
|
| 8 |
+
"disable_bias_gelu_fusion": false,
|
| 9 |
+
"disable_bias_skip_layer_norm": null,
|
| 10 |
+
"disable_bias_skip_layer_norm_fusion": false,
|
| 11 |
+
"disable_embed_layer_norm": true,
|
| 12 |
+
"disable_embed_layer_norm_fusion": true,
|
| 13 |
+
"disable_gelu": null,
|
| 14 |
+
"disable_gelu_fusion": false,
|
| 15 |
+
"disable_group_norm_fusion": true,
|
| 16 |
+
"disable_layer_norm": null,
|
| 17 |
+
"disable_layer_norm_fusion": false,
|
| 18 |
+
"disable_packed_kv": true,
|
| 19 |
+
"disable_rotary_embeddings": false,
|
| 20 |
+
"disable_shape_inference": true,
|
| 21 |
+
"disable_skip_layer_norm": null,
|
| 22 |
+
"disable_skip_layer_norm_fusion": false,
|
| 23 |
+
"enable_gelu_approximation": false,
|
| 24 |
+
"enable_gemm_fast_gelu_fusion": false,
|
| 25 |
+
"enable_transformers_specific_optimizations": true,
|
| 26 |
+
"fp16": false,
|
| 27 |
+
"no_attention_mask": false,
|
| 28 |
+
"optimization_level": 2,
|
| 29 |
+
"optimize_for_gpu": false,
|
| 30 |
+
"optimize_with_onnxruntime_only": null,
|
| 31 |
+
"use_mask_index": false,
|
| 32 |
+
"use_multi_head_attention": false,
|
| 33 |
+
"use_raw_attention_mask": false
|
| 34 |
+
},
|
| 35 |
+
"optimum_version": "1.20.0.dev0",
|
| 36 |
+
"quantization": {},
|
| 37 |
+
"transformers_version": "4.39.3",
|
| 38 |
+
"use_external_data_format": false
|
| 39 |
+
}
|
onnx_surya_O2/preprocessor_config.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_valid_processor_keys": [
|
| 3 |
+
"images",
|
| 4 |
+
"segmentation_maps",
|
| 5 |
+
"do_resize",
|
| 6 |
+
"size",
|
| 7 |
+
"resample",
|
| 8 |
+
"do_rescale",
|
| 9 |
+
"rescale_factor",
|
| 10 |
+
"do_normalize",
|
| 11 |
+
"image_mean",
|
| 12 |
+
"image_std",
|
| 13 |
+
"do_reduce_labels",
|
| 14 |
+
"return_tensors",
|
| 15 |
+
"data_format",
|
| 16 |
+
"input_data_format"
|
| 17 |
+
],
|
| 18 |
+
"do_normalize": true,
|
| 19 |
+
"do_reduce_labels": true,
|
| 20 |
+
"do_rescale": true,
|
| 21 |
+
"do_resize": true,
|
| 22 |
+
"image_mean": [
|
| 23 |
+
0.485,
|
| 24 |
+
0.456,
|
| 25 |
+
0.406
|
| 26 |
+
],
|
| 27 |
+
"image_processor_type": "SegformerFeatureExtractor",
|
| 28 |
+
"image_std": [
|
| 29 |
+
0.229,
|
| 30 |
+
0.224,
|
| 31 |
+
0.225
|
| 32 |
+
],
|
| 33 |
+
"resample": 2,
|
| 34 |
+
"rescale_factor": 0.00392156862745098,
|
| 35 |
+
"size": {
|
| 36 |
+
"height": 1024,
|
| 37 |
+
"width": 1024
|
| 38 |
+
}
|
| 39 |
+
}
|