bluryar commited on
Commit
39ecb9e
·
verified ·
1 Parent(s): 2d3f9a3

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # UIE ONNX模型使用指南
2
+
3
+ 本指南介绍如何使用和导出UIE (Universal Information Extraction) 模型的ONNX版本。
4
+
5
+ ## 模型说明
6
+
7
+ UIE模型是一个通用信息抽取模型,基于BERT架构。该模型可以导出为ONNX格式以实现更快的推理速度。
8
+
9
+ ## 模型输入
10
+
11
+ 模型接受以下输入张量:
12
+
13
+ - `input_ids`: 形状为 `[batch_size, sequence_length]` 的整型张量
14
+ - `attention_mask`: 形状为 `[batch_size, sequence_length]` 的整型张量
15
+ - `token_type_ids`: 形状为 `[batch_size, sequence_length]` 的整型张量
16
+
17
+ 其中:
18
+
19
+ - `batch_size`: 批处理大小,可变
20
+ - `sequence_length`: 序列长度,可变
21
+ - 所有输入张量的数据类型均为 `int64`
22
+
23
+ ## 模型输出
24
+
25
+ 模型输出以下张量:
26
+
27
+ - `start_prob`: 形状为 `[batch_size, sequence_length]` 的浮点张量,表示每个位置作为实体开始的概率
28
+ - `end_prob`: 形状为 `[batch_size, sequence_length]` 的浮点张量,表示每个位置作为实体结束的概率
29
+
30
+ ## 使用示例
31
+
32
+ ```python
33
+ python
34
+ import onnxruntime as ort
35
+
36
+ tokenizer = AutoTokenizer.from_pretrained("xusenlin/uie-base")
37
+ session = ort.InferenceSession("path/to/model.onnx")
38
+
39
+
40
+ intput = "张三与B公司签订了一份合同,约定了合同金额为100万元,合同期限为一年。"
41
+ schema = ["人名", "公司", "金额", "时间"]
42
+ input_ids_tensor, attention_mask, token_type_ids, offsets_mapping = tokenizer(
43
+ intput,
44
+ schema[0],
45
+ return_tensors="pt",
46
+ return_offsets_mapping=True,
47
+ add_special_tokens=True
48
+ )
49
+ inputs = {
50
+ "input_ids": input_ids_tensor, # shape: [batch_size, sequence_length]
51
+ "attention_mask": attention_mask, # shape: [batch_size, sequence_length]
52
+ "token_type_ids": token_type_ids # shape: [batch_size, sequence_length]
53
+ }
54
+ outputs = session.run(None, inputs)
55
+ start_probs, end_probs = outputs
56
+
57
+ # use offsets_mapping and start_probs, end_probs to get the entities
58
+ # ...
59
+ ```
onnx/model.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:adca8ba807404f82ecf3322ae639cf95c4173b0978bd6ef070bac51cb4078ab5
3
- size 469649896
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8aaba33f81d2868ed07d2c71f95f5d2d718a676a731b451ea76e3154ca612487
3
+ size 469773190
onnx/model_bnb4.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:946ead43384e67887961d7e8055310214ea22dd61235334640ddeaf5764a67c6
3
- size 177698218
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab458f771990521f0202c55db487c2e5338f716a7b88bc035c01fb409b982c89
3
+ size 177816534
onnx/model_fp16.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:10075606e1f8dfd37fa1f71beadea0161169aa54694a9ffdc12d0437ecb551f2
3
- size 234979602
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97cf1f26852109993535d25aa24cdae21c82adf22763b5db2c7c3b095358808b
3
+ size 235061519
onnx/model_int8.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e33af1093b09b55a9a287cf1f5a93d165ada4d8e3267bedb10e99c8070329ba5
3
- size 118454842
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5863a41c0cba2213dfca9170424f8b3180879fcc747a357cbcb67d969d3c44a
3
+ size 118583963
onnx/model_q4.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3fce245c6009719d1a5a36b976477f4329f451a5d56d1ecdb2adee164a5092eb
3
- size 183006106
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5640f3374b7ceb1c8f05ee366329f3683701356ba3c2449518a8d212cce8ea9a
3
+ size 183124504
onnx/model_q4f16.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:72ff1fb57dc3b951a9c91278014814ac1cfbcec36db809344b8a470369998e92
3
- size 112896805
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d833f118cd3ca1ecf75c4e7535e5158c1fd9fbf3662f3a3206f1317c3af9d80e
3
+ size 112976800
onnx/model_quantized.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e33af1093b09b55a9a287cf1f5a93d165ada4d8e3267bedb10e99c8070329ba5
3
- size 118454842
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5863a41c0cba2213dfca9170424f8b3180879fcc747a357cbcb67d969d3c44a
3
+ size 118583963
onnx/model_uint8.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a9d6f7c0098f7c10f15e5c2ed3b2229c9befab7620a551907b9a57bfcd89ffeb
3
- size 118454840
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cffb9fdf2d31592ef3dd438b152afd958b0dab51766f63e4339e78a0bf9eca45
3
+ size 118583960
test_report.json ADDED
@@ -0,0 +1,1083 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "timestamp": "2024-12-15 14:19:45",
3
+ "model_path": "./uie_base/xusenlin/uie-base/onnx/model_fp16.onnx",
4
+ "model_info": {
5
+ "onnx": "./uie_base/xusenlin/uie-base/onnx/model_fp16.onnx",
6
+ "pytorch": "xusenlin/uie-base"
7
+ },
8
+ "test_results": [
9
+ {
10
+ "case_name": "实体抽取",
11
+ "input": {
12
+ "text": "立冬过后小雪来,但小雪时节却无雪,这让最喜欢雪夜温酒读禁书的世子殿下很遗憾。\n\n 白狐儿脸已经在听潮亭一楼呆了半旬,入定入魔,这份毅力让吃不了苦的徐凤年自惭形秽,但这不耽误徐凤年在王府上找乐子。\n\n 花魁鱼幼薇安定下来,住在一个一夜间被植入棠蕉两种植物的幽静院子,白猫武媚娘似乎很满意新窝,又胖了几分。\n\n 徐凤年给鱼幼薇送去了最上等的貂裘,最精美的食物,但始终没有再度临幸她的凝脂美玉,刻意生疏,那个圆滚滚的禄球儿说得对,养人跟养鹰是一个理儿,得慢慢调教,快了容易失去灵气,慢了就不乖巧。\n 这世上敢这么调戏世子殿下的,明摆着就只有大柱国长女徐脂虎了。\n\n 姐弟两个从小就关系极好,她出嫁前,徐凤年到了十二三岁还被她拉着同床共枕,如果说天下间北凉王徐骁是最护短徐凤年的,徐龙象是最听话的,那徐脂虎绝对是最宠溺欢喜徐凤年的。\n\n 一得到父王书信说弟弟回城,徐脂虎立即就马不停蹄带着一群豪奴恶仆赶回娘家。\n\n",
13
+ "schema": [
14
+ "人名",
15
+ "时间",
16
+ "地点",
17
+ "饰品"
18
+ ]
19
+ },
20
+ "outputs": {
21
+ "onnx": [
22
+ {
23
+ "人名": [
24
+ {
25
+ "text": "徐凤年",
26
+ "start": 76.0,
27
+ "end": 79.0,
28
+ "probability": 0.9734153300214103
29
+ },
30
+ {
31
+ "text": "徐脂虎",
32
+ "start": 284.0,
33
+ "end": 287.0,
34
+ "probability": 0.977874454327079
35
+ },
36
+ {
37
+ "text": "徐凤年",
38
+ "start": 346.0,
39
+ "end": 349.0,
40
+ "probability": 0.9519458791550761
41
+ },
42
+ {
43
+ "text": "徐骁",
44
+ "start": 340.0,
45
+ "end": 342.0,
46
+ "probability": 0.9722451064385211
47
+ },
48
+ {
49
+ "text": "徐凤年",
50
+ "start": 312.0,
51
+ "end": 315.0,
52
+ "probability": 0.9478657699871214
53
+ },
54
+ {
55
+ "text": "徐脂虎",
56
+ "start": 361.0,
57
+ "end": 364.0,
58
+ "probability": 0.9789169051413857
59
+ },
60
+ {
61
+ "text": "徐脂虎",
62
+ "start": 396.0,
63
+ "end": 399.0,
64
+ "probability": 0.9701647997775922
65
+ },
66
+ {
67
+ "text": "徐凤年",
68
+ "start": 372.0,
69
+ "end": 375.0,
70
+ "probability": 0.9323382975404115
71
+ },
72
+ {
73
+ "text": "王",
74
+ "start": 387.0,
75
+ "end": 388.0,
76
+ "probability": 0.5895737007551531
77
+ },
78
+ {
79
+ "text": "鱼幼薇",
80
+ "start": 167.0,
81
+ "end": 170.0,
82
+ "probability": 0.894556804380116
83
+ },
84
+ {
85
+ "text": "徐凤年",
86
+ "start": 163.0,
87
+ "end": 166.0,
88
+ "probability": 0.9348918167083013
89
+ },
90
+ {
91
+ "text": "鱼幼薇",
92
+ "start": 108.0,
93
+ "end": 111.0,
94
+ "probability": 0.7981717829580433
95
+ },
96
+ {
97
+ "text": "武媚娘",
98
+ "start": 140.0,
99
+ "end": 143.0,
100
+ "probability": 0.5757094842778514
101
+ },
102
+ {
103
+ "text": "徐龙象",
104
+ "start": 351.0,
105
+ "end": 354.0,
106
+ "probability": 0.8924126757258222
107
+ },
108
+ {
109
+ "text": "徐凤年",
110
+ "start": 89.0,
111
+ "end": 92.0,
112
+ "probability": 0.9204391283800426
113
+ }
114
+ ],
115
+ "饰品": [
116
+ {
117
+ "text": "凝脂美玉",
118
+ "start": 198.0,
119
+ "end": 202.0,
120
+ "probability": 0.5377234647750875
121
+ }
122
+ ]
123
+ }
124
+ ],
125
+ "pytorch": [
126
+ {
127
+ "人名": [
128
+ {
129
+ "text": "徐凤年",
130
+ "start": 163.0,
131
+ "end": 166.0,
132
+ "probability": 0.9347324967384338
133
+ },
134
+ {
135
+ "text": "徐龙象",
136
+ "start": 351.0,
137
+ "end": 354.0,
138
+ "probability": 0.8920297622680664
139
+ },
140
+ {
141
+ "text": "徐凤年",
142
+ "start": 76.0,
143
+ "end": 79.0,
144
+ "probability": 0.9734014272689819
145
+ },
146
+ {
147
+ "text": "鱼幼薇",
148
+ "start": 108.0,
149
+ "end": 111.0,
150
+ "probability": 0.7986579537391663
151
+ },
152
+ {
153
+ "text": "徐凤年",
154
+ "start": 312.0,
155
+ "end": 315.0,
156
+ "probability": 0.947639524936676
157
+ },
158
+ {
159
+ "text": "徐凤年",
160
+ "start": 346.0,
161
+ "end": 349.0,
162
+ "probability": 0.9518017172813416
163
+ },
164
+ {
165
+ "text": "鱼幼薇",
166
+ "start": 167.0,
167
+ "end": 170.0,
168
+ "probability": 0.8946396708488464
169
+ },
170
+ {
171
+ "text": "徐凤年",
172
+ "start": 372.0,
173
+ "end": 375.0,
174
+ "probability": 0.9319543838500977
175
+ },
176
+ {
177
+ "text": "王",
178
+ "start": 387.0,
179
+ "end": 388.0,
180
+ "probability": 0.5893056392669678
181
+ },
182
+ {
183
+ "text": "徐脂虎",
184
+ "start": 396.0,
185
+ "end": 399.0,
186
+ "probability": 0.9700623750686646
187
+ },
188
+ {
189
+ "text": "徐脂虎",
190
+ "start": 361.0,
191
+ "end": 364.0,
192
+ "probability": 0.9788591265678406
193
+ },
194
+ {
195
+ "text": "徐脂虎",
196
+ "start": 284.0,
197
+ "end": 287.0,
198
+ "probability": 0.9778398275375366
199
+ },
200
+ {
201
+ "text": "徐骁",
202
+ "start": 340.0,
203
+ "end": 342.0,
204
+ "probability": 0.9721871018409729
205
+ },
206
+ {
207
+ "text": "徐凤年",
208
+ "start": 89.0,
209
+ "end": 92.0,
210
+ "probability": 0.9202938675880432
211
+ },
212
+ {
213
+ "text": "武媚娘",
214
+ "start": 140.0,
215
+ "end": 143.0,
216
+ "probability": 0.5761253833770752
217
+ }
218
+ ],
219
+ "饰品": [
220
+ {
221
+ "text": "凝脂美玉",
222
+ "start": 198.0,
223
+ "end": 202.0,
224
+ "probability": 0.5364798307418823
225
+ }
226
+ ]
227
+ }
228
+ ],
229
+ "matches": false,
230
+ "probability_diffs": [
231
+ {
232
+ "path": "[0].人名[0].probability",
233
+ "onnx": 0.9734153300214103,
234
+ "pytorch": 0.9347324967384338,
235
+ "diff": 0.0386828332829765
236
+ },
237
+ {
238
+ "path": "[0].人名[1].probability",
239
+ "onnx": 0.977874454327079,
240
+ "pytorch": 0.8920297622680664,
241
+ "diff": 0.08584469205901257
242
+ },
243
+ {
244
+ "path": "[0].人名[2].probability",
245
+ "onnx": 0.9519458791550761,
246
+ "pytorch": 0.9734014272689819,
247
+ "diff": 0.02145554811390582
248
+ },
249
+ {
250
+ "path": "[0].人名[3].probability",
251
+ "onnx": 0.9722451064385211,
252
+ "pytorch": 0.7986579537391663,
253
+ "diff": 0.17358715269935487
254
+ },
255
+ {
256
+ "path": "[0].人名[4].probability",
257
+ "onnx": 0.9478657699871214,
258
+ "pytorch": 0.947639524936676,
259
+ "diff": 0.0002262450504453284
260
+ },
261
+ {
262
+ "path": "[0].人名[5].probability",
263
+ "onnx": 0.9789169051413857,
264
+ "pytorch": 0.9518017172813416,
265
+ "diff": 0.027115187860044188
266
+ },
267
+ {
268
+ "path": "[0].人名[6].probability",
269
+ "onnx": 0.9701647997775922,
270
+ "pytorch": 0.8946396708488464,
271
+ "diff": 0.07552512892874574
272
+ },
273
+ {
274
+ "path": "[0].人名[7].probability",
275
+ "onnx": 0.9323382975404115,
276
+ "pytorch": 0.9319543838500977,
277
+ "diff": 0.0003839136903138751
278
+ },
279
+ {
280
+ "path": "[0].人名[8].probability",
281
+ "onnx": 0.5895737007551531,
282
+ "pytorch": 0.5893056392669678,
283
+ "diff": 0.0002680614881853671
284
+ },
285
+ {
286
+ "path": "[0].人名[9].probability",
287
+ "onnx": 0.894556804380116,
288
+ "pytorch": 0.9700623750686646,
289
+ "diff": 0.07550557068854857
290
+ },
291
+ {
292
+ "path": "[0].人名[10].probability",
293
+ "onnx": 0.9348918167083013,
294
+ "pytorch": 0.9788591265678406,
295
+ "diff": 0.04396730985953923
296
+ },
297
+ {
298
+ "path": "[0].人名[11].probability",
299
+ "onnx": 0.7981717829580433,
300
+ "pytorch": 0.9778398275375366,
301
+ "diff": 0.17966804457949337
302
+ },
303
+ {
304
+ "path": "[0].人名[12].probability",
305
+ "onnx": 0.5757094842778514,
306
+ "pytorch": 0.9721871018409729,
307
+ "diff": 0.3964776175631215
308
+ },
309
+ {
310
+ "path": "[0].人名[13].probability",
311
+ "onnx": 0.8924126757258222,
312
+ "pytorch": 0.9202938675880432,
313
+ "diff": 0.027881191862221044
314
+ },
315
+ {
316
+ "path": "[0].人名[14].probability",
317
+ "onnx": 0.9204391283800426,
318
+ "pytorch": 0.5761253833770752,
319
+ "diff": 0.3443137450029674
320
+ },
321
+ {
322
+ "path": "[0].饰品[0].probability",
323
+ "onnx": 0.5377234647750875,
324
+ "pytorch": 0.5364798307418823,
325
+ "diff": 0.0012436340332051543
326
+ }
327
+ ],
328
+ "max_probability_diff": 0.3964776175631215
329
+ }
330
+ },
331
+ {
332
+ "case_name": "实体抽取",
333
+ "input": {
334
+ "text": "2月8日上午北京冬奥会自由式滑雪女子大跳台决赛中中国选手谷爱凌以188.25分获得金牌",
335
+ "schema": [
336
+ "时间",
337
+ "选手",
338
+ "赛事名称"
339
+ ]
340
+ },
341
+ "outputs": {
342
+ "onnx": [
343
+ {
344
+ "时间": [
345
+ {
346
+ "text": "2月8日上午",
347
+ "start": 0.0,
348
+ "end": 6.0,
349
+ "probability": 0.98412900742278
350
+ }
351
+ ],
352
+ "选手": [
353
+ {
354
+ "text": "谷爱凌",
355
+ "start": 28.0,
356
+ "end": 31.0,
357
+ "probability": 0.948812334341909
358
+ }
359
+ ],
360
+ "赛事名称": [
361
+ {
362
+ "text": "北京冬奥会自由式滑雪女子大跳台决赛",
363
+ "start": 6.0,
364
+ "end": 23.0,
365
+ "probability": 0.8410896860125945
366
+ }
367
+ ]
368
+ }
369
+ ],
370
+ "pytorch": [
371
+ {
372
+ "时间": [
373
+ {
374
+ "text": "2月8日上午",
375
+ "start": 0.0,
376
+ "end": 6.0,
377
+ "probability": 0.9841681718826294
378
+ }
379
+ ],
380
+ "选手": [
381
+ {
382
+ "text": "谷爱凌",
383
+ "start": 28.0,
384
+ "end": 31.0,
385
+ "probability": 0.9488540291786194
386
+ }
387
+ ],
388
+ "赛事名称": [
389
+ {
390
+ "text": "北京冬奥会自由式滑雪女子大跳台决赛",
391
+ "start": 6.0,
392
+ "end": 23.0,
393
+ "probability": 0.8410930037498474
394
+ }
395
+ ]
396
+ }
397
+ ],
398
+ "matches": false,
399
+ "probability_diffs": [
400
+ {
401
+ "path": "[0].时间[0].probability",
402
+ "onnx": 0.98412900742278,
403
+ "pytorch": 0.9841681718826294,
404
+ "diff": 3.9164459849416744e-05
405
+ },
406
+ {
407
+ "path": "[0].选手[0].probability",
408
+ "onnx": 0.948812334341909,
409
+ "pytorch": 0.9488540291786194,
410
+ "diff": 4.169483671034868e-05
411
+ },
412
+ {
413
+ "path": "[0].赛事名称[0].probability",
414
+ "onnx": 0.8410896860125945,
415
+ "pytorch": 0.8410930037498474,
416
+ "diff": 3.3177372529280547e-06
417
+ }
418
+ ],
419
+ "max_probability_diff": 4.169483671034868e-05
420
+ }
421
+ },
422
+ {
423
+ "case_name": "关系抽取",
424
+ "input": {
425
+ "text": "2022语言与智能技术竞赛由中国中文信息学会和中国计算机学会联合主办,百度公司、中国中文信息学会评测工作委员会和中国计算机学会自然语言处理专委会承办,已连续举办4届,成为全球最热门的中文NLP赛事之一。",
426
+ "schema": {
427
+ "竞赛名称": [
428
+ "主办方",
429
+ "承办方",
430
+ "已举办次数"
431
+ ]
432
+ }
433
+ },
434
+ "outputs": {
435
+ "onnx": [
436
+ {
437
+ "竞赛名称": [
438
+ {
439
+ "text": "2022语言与智能技术竞赛",
440
+ "start": 0.0,
441
+ "end": 13.0,
442
+ "probability": 0.7823843082569226,
443
+ "relations": {
444
+ "主办方": [
445
+ {
446
+ "text": "中国中文信息学会",
447
+ "start": 14.0,
448
+ "end": 22.0,
449
+ "probability": 0.8419661941400491
450
+ },
451
+ {
452
+ "text": "中国计算机学会",
453
+ "start": 23.0,
454
+ "end": 30.0,
455
+ "probability": 0.7578377954284861
456
+ }
457
+ ],
458
+ "承办方": [
459
+ {
460
+ "text": "中国中文信息学会评测工作委员会",
461
+ "start": 40.0,
462
+ "end": 55.0,
463
+ "probability": 0.7001135169948149
464
+ },
465
+ {
466
+ "text": "中国计算机学会自然语言处理专委会",
467
+ "start": 56.0,
468
+ "end": 72.0,
469
+ "probability": 0.6193604266328165
470
+ },
471
+ {
472
+ "text": "百度公司",
473
+ "start": 35.0,
474
+ "end": 39.0,
475
+ "probability": 0.8291344432635412
476
+ }
477
+ ],
478
+ "已举办次数": [
479
+ {
480
+ "text": "4届",
481
+ "start": 80.0,
482
+ "end": 82.0,
483
+ "probability": 0.4672039099420573
484
+ }
485
+ ]
486
+ }
487
+ }
488
+ ]
489
+ }
490
+ ],
491
+ "pytorch": [
492
+ {
493
+ "竞赛名称": [
494
+ {
495
+ "text": "2022语言与智能技术竞赛",
496
+ "start": 0.0,
497
+ "end": 13.0,
498
+ "probability": 0.7825390696525574,
499
+ "relations": {
500
+ "主办方": [
501
+ {
502
+ "text": "中国中文信息学会",
503
+ "start": 14.0,
504
+ "end": 22.0,
505
+ "probability": 0.8421703577041626
506
+ },
507
+ {
508
+ "text": "中国计算机学会",
509
+ "start": 23.0,
510
+ "end": 30.0,
511
+ "probability": 0.7580800652503967
512
+ }
513
+ ],
514
+ "承办方": [
515
+ {
516
+ "text": "中国计算机学会自然语言处理专委会",
517
+ "start": 56.0,
518
+ "end": 72.0,
519
+ "probability": 0.6193482875823975
520
+ },
521
+ {
522
+ "text": "百度公司",
523
+ "start": 35.0,
524
+ "end": 39.0,
525
+ "probability": 0.8292697668075562
526
+ },
527
+ {
528
+ "text": "中国中文信息学会评测工作委员会",
529
+ "start": 40.0,
530
+ "end": 55.0,
531
+ "probability": 0.7000494003295898
532
+ }
533
+ ],
534
+ "已举办次数": [
535
+ {
536
+ "text": "4届",
537
+ "start": 80.0,
538
+ "end": 82.0,
539
+ "probability": 0.46712979674339294
540
+ }
541
+ ]
542
+ }
543
+ }
544
+ ]
545
+ }
546
+ ],
547
+ "matches": false,
548
+ "probability_diffs": [
549
+ {
550
+ "path": "[0].竞赛名称[0].probability",
551
+ "onnx": 0.7823843082569226,
552
+ "pytorch": 0.7825390696525574,
553
+ "diff": 0.00015476139563475044
554
+ },
555
+ {
556
+ "path": "[0].竞赛名称[0].relations.主办方[0].probability",
557
+ "onnx": 0.8419661941400491,
558
+ "pytorch": 0.8421703577041626,
559
+ "diff": 0.0002041635641134576
560
+ },
561
+ {
562
+ "path": "[0].竞赛名称[0].relations.主办方[1].probability",
563
+ "onnx": 0.7578377954284861,
564
+ "pytorch": 0.7580800652503967,
565
+ "diff": 0.00024226982191066782
566
+ },
567
+ {
568
+ "path": "[0].竞赛名称[0].relations.承办方[0].probability",
569
+ "onnx": 0.7001135169948149,
570
+ "pytorch": 0.6193482875823975,
571
+ "diff": 0.08076522941241748
572
+ },
573
+ {
574
+ "path": "[0].竞赛名称[0].relations.承办方[1].probability",
575
+ "onnx": 0.6193604266328165,
576
+ "pytorch": 0.8292697668075562,
577
+ "diff": 0.2099093401747396
578
+ },
579
+ {
580
+ "path": "[0].竞赛名称[0].relations.承办方[2].probability",
581
+ "onnx": 0.8291344432635412,
582
+ "pytorch": 0.7000494003295898,
583
+ "diff": 0.1290850429339514
584
+ },
585
+ {
586
+ "path": "[0].竞赛名称[0].relations.已举办次数[0].probability",
587
+ "onnx": 0.4672039099420573,
588
+ "pytorch": 0.46712979674339294,
589
+ "diff": 7.411319866434951e-05
590
+ }
591
+ ],
592
+ "max_probability_diff": 0.2099093401747396
593
+ }
594
+ },
595
+ {
596
+ "case_name": "事件抽取",
597
+ "input": {
598
+ "text": "中国地震台网正式测定:5月16日06时08分在云南临沧市凤庆县(北纬24.34度,东经99.98度)发生3.5级地震,震源深度10千米。",
599
+ "schema": {
600
+ "地震触发词": [
601
+ "地震强度",
602
+ "时间",
603
+ "震中位置",
604
+ "震源深度"
605
+ ]
606
+ }
607
+ },
608
+ "outputs": {
609
+ "onnx": [
610
+ {
611
+ "地震触发词": [
612
+ {
613
+ "text": "地震",
614
+ "start": 56.0,
615
+ "end": 58.0,
616
+ "probability": 0.997742496069165,
617
+ "relations": {
618
+ "地震强度": [
619
+ {
620
+ "text": "3.5级",
621
+ "start": 52.0,
622
+ "end": 56.0,
623
+ "probability": 0.9980807537133956
624
+ }
625
+ ],
626
+ "时间": [
627
+ {
628
+ "text": "5月16日06时08分",
629
+ "start": 11.0,
630
+ "end": 22.0,
631
+ "probability": 0.9852992667393039
632
+ }
633
+ ],
634
+ "震中位置": [
635
+ {
636
+ "text": "云南临沧市凤庆县(北纬24.34度,东经99.98度)",
637
+ "start": 23.0,
638
+ "end": 50.0,
639
+ "probability": 0.7874488113591553
640
+ }
641
+ ],
642
+ "震源深度": [
643
+ {
644
+ "text": "10千米",
645
+ "start": 63.0,
646
+ "end": 67.0,
647
+ "probability": 0.9937999405099731
648
+ }
649
+ ]
650
+ }
651
+ }
652
+ ]
653
+ }
654
+ ],
655
+ "pytorch": [
656
+ {
657
+ "地震触发词": [
658
+ {
659
+ "text": "地震",
660
+ "start": 56.0,
661
+ "end": 58.0,
662
+ "probability": 0.9977425336837769,
663
+ "relations": {
664
+ "地震强度": [
665
+ {
666
+ "text": "3.5级",
667
+ "start": 52.0,
668
+ "end": 56.0,
669
+ "probability": 0.9980801939964294
670
+ }
671
+ ],
672
+ "时间": [
673
+ {
674
+ "text": "5月16日06时08分",
675
+ "start": 11.0,
676
+ "end": 22.0,
677
+ "probability": 0.9853299856185913
678
+ }
679
+ ],
680
+ "震中位置": [
681
+ {
682
+ "text": "云南临沧市凤庆县(北纬24.34度,东经99.98度)",
683
+ "start": 23.0,
684
+ "end": 50.0,
685
+ "probability": 0.7874014973640442
686
+ }
687
+ ],
688
+ "震源深度": [
689
+ {
690
+ "text": "10千米",
691
+ "start": 63.0,
692
+ "end": 67.0,
693
+ "probability": 0.9937973022460938
694
+ }
695
+ ]
696
+ }
697
+ }
698
+ ]
699
+ }
700
+ ],
701
+ "matches": false,
702
+ "probability_diffs": [
703
+ {
704
+ "path": "[0].地震触发词[0].probability",
705
+ "onnx": 0.997742496069165,
706
+ "pytorch": 0.9977425336837769,
707
+ "diff": 3.761461186968518e-08
708
+ },
709
+ {
710
+ "path": "[0].地震触发词[0].relations.地震强度[0].probability",
711
+ "onnx": 0.9980807537133956,
712
+ "pytorch": 0.9980801939964294,
713
+ "diff": 5.597169661086809e-07
714
+ },
715
+ {
716
+ "path": "[0].地震触发词[0].relations.时间[0].probability",
717
+ "onnx": 0.9852992667393039,
718
+ "pytorch": 0.9853299856185913,
719
+ "diff": 3.07188792874058e-05
720
+ },
721
+ {
722
+ "path": "[0].地震触发词[0].relations.震中位置[0].probability",
723
+ "onnx": 0.7874488113591553,
724
+ "pytorch": 0.7874014973640442,
725
+ "diff": 4.731399511115342e-05
726
+ },
727
+ {
728
+ "path": "[0].地震触发词[0].relations.震源深度[0].probability",
729
+ "onnx": 0.9937999405099731,
730
+ "pytorch": 0.9937973022460938,
731
+ "diff": 2.638263879362057e-06
732
+ }
733
+ ],
734
+ "max_probability_diff": 4.731399511115342e-05
735
+ }
736
+ },
737
+ {
738
+ "case_name": "情感分析",
739
+ "input": {
740
+ "text": "店面干净,很清静,服务员服务热情,性价比很高,发现收银台有排队",
741
+ "schema": {
742
+ "评价维度": [
743
+ "观点词",
744
+ "情感倾向[正向,负向]"
745
+ ]
746
+ }
747
+ },
748
+ "outputs": {
749
+ "onnx": [
750
+ {
751
+ "评价维度": [
752
+ {
753
+ "text": "性价比",
754
+ "start": 17.0,
755
+ "end": 20.0,
756
+ "probability": 0.9816947685707831,
757
+ "relations": {
758
+ "观点词": [
759
+ {
760
+ "text": "高",
761
+ "start": 21.0,
762
+ "end": 22.0,
763
+ "probability": 0.9573563787963479
764
+ }
765
+ ],
766
+ "情感倾向[正向,负向]": [
767
+ {
768
+ "text": "正向",
769
+ "probability": 0.9966081212676805
770
+ }
771
+ ]
772
+ }
773
+ },
774
+ {
775
+ "text": "店面",
776
+ "start": 0.0,
777
+ "end": 2.0,
778
+ "probability": 0.9697329670369754,
779
+ "relations": {
780
+ "观点词": [
781
+ {
782
+ "text": "干净",
783
+ "start": 2.0,
784
+ "end": 4.0,
785
+ "probability": 0.9945395308698579
786
+ }
787
+ ],
788
+ "情感倾向[正向,负向]": [
789
+ {
790
+ "text": "正向",
791
+ "probability": 0.9982160421894335
792
+ }
793
+ ]
794
+ }
795
+ }
796
+ ]
797
+ }
798
+ ],
799
+ "pytorch": [
800
+ {
801
+ "评价维度": [
802
+ {
803
+ "text": "性价比",
804
+ "start": 17.0,
805
+ "end": 20.0,
806
+ "probability": 0.9817039370536804,
807
+ "relations": {
808
+ "观点词": [
809
+ {
810
+ "text": "高",
811
+ "start": 21.0,
812
+ "end": 22.0,
813
+ "probability": 0.9573964476585388
814
+ }
815
+ ],
816
+ "情感倾向[正向,负向]": [
817
+ {
818
+ "text": "正向",
819
+ "probability": 0.9966142773628235
820
+ }
821
+ ]
822
+ }
823
+ },
824
+ {
825
+ "text": "店面",
826
+ "start": 0.0,
827
+ "end": 2.0,
828
+ "probability": 0.9696847200393677,
829
+ "relations": {
830
+ "观点词": [
831
+ {
832
+ "text": "干净",
833
+ "start": 2.0,
834
+ "end": 4.0,
835
+ "probability": 0.9945318102836609
836
+ }
837
+ ],
838
+ "情感倾向[正向,负向]": [
839
+ {
840
+ "text": "正向",
841
+ "probability": 0.9982153177261353
842
+ }
843
+ ]
844
+ }
845
+ }
846
+ ]
847
+ }
848
+ ],
849
+ "matches": false,
850
+ "probability_diffs": [
851
+ {
852
+ "path": "[0].评价维度[0].probability",
853
+ "onnx": 0.9816947685707831,
854
+ "pytorch": 0.9817039370536804,
855
+ "diff": 9.168482897337071e-06
856
+ },
857
+ {
858
+ "path": "[0].评价维度[0].relations.观点词[0].probability",
859
+ "onnx": 0.9573563787963479,
860
+ "pytorch": 0.9573964476585388,
861
+ "diff": 4.006886219087846e-05
862
+ },
863
+ {
864
+ "path": "[0].评价维度[0].relations.情感倾向[正向,负向][0].probability",
865
+ "onnx": 0.9966081212676805,
866
+ "pytorch": 0.9966142773628235,
867
+ "diff": 6.156095142983986e-06
868
+ },
869
+ {
870
+ "path": "[0].评价维度[1].probability",
871
+ "onnx": 0.9697329670369754,
872
+ "pytorch": 0.9696847200393677,
873
+ "diff": 4.82469976077482e-05
874
+ },
875
+ {
876
+ "path": "[0].评价维度[1].relations.观点词[0].probability",
877
+ "onnx": 0.9945395308698579,
878
+ "pytorch": 0.9945318102836609,
879
+ "diff": 7.720586197024204e-06
880
+ },
881
+ {
882
+ "path": "[0].评价维度[1].relations.情感倾向[正向,负向][0].probability",
883
+ "onnx": 0.9982160421894335,
884
+ "pytorch": 0.9982153177261353,
885
+ "diff": 7.244632982406074e-07
886
+ }
887
+ ],
888
+ "max_probability_diff": 4.82469976077482e-05
889
+ }
890
+ },
891
+ {
892
+ "case_name": "分类任务",
893
+ "input": {
894
+ "text": "这个产品用起来真的很流畅,我非常喜欢",
895
+ "schema": "情感倾向[正向,负向]"
896
+ },
897
+ "outputs": {
898
+ "onnx": [
899
+ {
900
+ "情感倾向[正向,负向]": [
901
+ {
902
+ "text": "正向",
903
+ "probability": 0.9990008568783395
904
+ }
905
+ ]
906
+ }
907
+ ],
908
+ "pytorch": [
909
+ {
910
+ "情感倾向[正向,负向]": [
911
+ {
912
+ "text": "正向",
913
+ "probability": 0.9990023970603943
914
+ }
915
+ ]
916
+ }
917
+ ],
918
+ "matches": false,
919
+ "probability_diffs": [
920
+ {
921
+ "path": "[0].情感倾向[正向,负向][0].probability",
922
+ "onnx": 0.9990008568783395,
923
+ "pytorch": 0.9990023970603943,
924
+ "diff": 1.5401820547822354e-06
925
+ }
926
+ ],
927
+ "max_probability_diff": 1.5401820547822354e-06
928
+ }
929
+ },
930
+ {
931
+ "case_name": "复杂关系抽取",
932
+ "input": {
933
+ "text": "北京市海淀区人民法院\n 民事判决书\n (199x)建初字第xxx号\n 原告:张三。\n 委托代理人李四,北京市 A律师事务所律师。\n 被告:B公司,法定代表人王五,开发公司总经理。\n 委托代理人赵六,北京市 C律师事务所律师。",
934
+ "schema": [
935
+ "法院",
936
+ {
937
+ "原告": "委托代理人"
938
+ },
939
+ {
940
+ "被告": "委托代理人"
941
+ }
942
+ ]
943
+ },
944
+ "outputs": {
945
+ "onnx": [
946
+ {
947
+ "法院": [
948
+ {
949
+ "text": "北京市海淀区人民法院",
950
+ "start": 0.0,
951
+ "end": 10.0,
952
+ "probability": 0.9219728006197414
953
+ }
954
+ ],
955
+ "原告": [
956
+ {
957
+ "text": "张三",
958
+ "start": 71.0,
959
+ "end": 73.0,
960
+ "probability": 0.994995078985994,
961
+ "relations": {
962
+ "委托代理人": [
963
+ {
964
+ "text": "李四",
965
+ "start": 92.0,
966
+ "end": 94.0,
967
+ "probability": 0.7952536469989475
968
+ }
969
+ ]
970
+ }
971
+ }
972
+ ],
973
+ "被告": [
974
+ {
975
+ "text": "B公司",
976
+ "start": 124.0,
977
+ "end": 127.0,
978
+ "probability": 0.8431729295214296,
979
+ "relations": {
980
+ "委托代理人": [
981
+ {
982
+ "text": "赵六",
983
+ "start": 162.0,
984
+ "end": 164.0,
985
+ "probability": 0.7268286253941483
986
+ }
987
+ ]
988
+ }
989
+ }
990
+ ]
991
+ }
992
+ ],
993
+ "pytorch": [
994
+ {
995
+ "法院": [
996
+ {
997
+ "text": "北京市海淀区人民法院",
998
+ "start": 0.0,
999
+ "end": 10.0,
1000
+ "probability": 0.9221080541610718
1001
+ }
1002
+ ],
1003
+ "原告": [
1004
+ {
1005
+ "text": "张三",
1006
+ "start": 71.0,
1007
+ "end": 73.0,
1008
+ "probability": 0.9949812889099121,
1009
+ "relations": {
1010
+ "委托代理人": [
1011
+ {
1012
+ "text": "李四",
1013
+ "start": 92.0,
1014
+ "end": 94.0,
1015
+ "probability": 0.7956846356391907
1016
+ }
1017
+ ]
1018
+ }
1019
+ }
1020
+ ],
1021
+ "被告": [
1022
+ {
1023
+ "text": "B公司",
1024
+ "start": 124.0,
1025
+ "end": 127.0,
1026
+ "probability": 0.8437343239784241,
1027
+ "relations": {
1028
+ "委托代理人": [
1029
+ {
1030
+ "text": "赵六",
1031
+ "start": 162.0,
1032
+ "end": 164.0,
1033
+ "probability": 0.7267114520072937
1034
+ }
1035
+ ]
1036
+ }
1037
+ }
1038
+ ]
1039
+ }
1040
+ ],
1041
+ "matches": false,
1042
+ "probability_diffs": [
1043
+ {
1044
+ "path": "[0].法院[0].probability",
1045
+ "onnx": 0.9219728006197414,
1046
+ "pytorch": 0.9221080541610718,
1047
+ "diff": 0.00013525354133037126
1048
+ },
1049
+ {
1050
+ "path": "[0].原告[0].probability",
1051
+ "onnx": 0.994995078985994,
1052
+ "pytorch": 0.9949812889099121,
1053
+ "diff": 1.3790076081932057e-05
1054
+ },
1055
+ {
1056
+ "path": "[0].原告[0].relations.委托代理人[0].probability",
1057
+ "onnx": 0.7952536469989475,
1058
+ "pytorch": 0.7956846356391907,
1059
+ "diff": 0.0004309886402431573
1060
+ },
1061
+ {
1062
+ "path": "[0].被告[0].probability",
1063
+ "onnx": 0.8431729295214296,
1064
+ "pytorch": 0.8437343239784241,
1065
+ "diff": 0.0005613944569944351
1066
+ },
1067
+ {
1068
+ "path": "[0].被告[0].relations.委托代理人[0].probability",
1069
+ "onnx": 0.7268286253941483,
1070
+ "pytorch": 0.7267114520072937,
1071
+ "diff": 0.00011717338685457435
1072
+ }
1073
+ ],
1074
+ "max_probability_diff": 0.0005613944569944351
1075
+ }
1076
+ }
1077
+ ],
1078
+ "summary": {
1079
+ "total_cases": 7,
1080
+ "matching_outputs": 0,
1081
+ "max_probability_diff": 0.3964776175631215
1082
+ }
1083
+ }