Upload 11 files
Browse files- config.json +731 -0
- optimizer.pt +3 -0
- pytorch_model.bin +3 -0
- rng_state.pth +3 -0
- scheduler.pt +3 -0
- special_tokens_map.json +7 -0
- tokenizer.json +0 -0
- tokenizer_config.json +13 -0
- trainer_state.json +148 -0
- training_args.bin +3 -0
- vocab.txt +0 -0
config.json
ADDED
@@ -0,0 +1,731 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "bert-base-multilingual-cased",
|
3 |
+
"architectures": [
|
4 |
+
"BertForTokenClassification"
|
5 |
+
],
|
6 |
+
"attention_probs_dropout_prob": 0.1,
|
7 |
+
"classifier_dropout": null,
|
8 |
+
"directionality": "bidi",
|
9 |
+
"hidden_act": "gelu",
|
10 |
+
"hidden_dropout_prob": 0.1,
|
11 |
+
"hidden_size": 768,
|
12 |
+
"id2label": {
|
13 |
+
"0": "O",
|
14 |
+
"1": "r0:arg0|agt",
|
15 |
+
"2": "r0:arg0|cau",
|
16 |
+
"3": "r0:arg0|exp",
|
17 |
+
"4": "r0:arg0|pat",
|
18 |
+
"5": "r0:arg0|src",
|
19 |
+
"6": "r0:arg1|ext",
|
20 |
+
"7": "r0:arg1|loc",
|
21 |
+
"8": "r0:arg1|pat",
|
22 |
+
"9": "r0:arg1|tem",
|
23 |
+
"10": "r0:arg2|atr",
|
24 |
+
"11": "r0:arg2|ben",
|
25 |
+
"12": "r0:arg2|efi",
|
26 |
+
"13": "r0:arg2|exp",
|
27 |
+
"14": "r0:arg2|ext",
|
28 |
+
"15": "r0:arg2|ins",
|
29 |
+
"16": "r0:arg2|loc",
|
30 |
+
"17": "r0:arg2|tem",
|
31 |
+
"18": "r0:arg3|ben",
|
32 |
+
"19": "r0:arg3|ein",
|
33 |
+
"20": "r0:arg3|exp",
|
34 |
+
"21": "r0:arg3|fin",
|
35 |
+
"22": "r0:arg3|ins",
|
36 |
+
"23": "r0:arg3|loc",
|
37 |
+
"24": "r0:arg3|ori",
|
38 |
+
"25": "r0:arg4|des",
|
39 |
+
"26": "r0:arg4|efi",
|
40 |
+
"27": "r0:argM|LOC",
|
41 |
+
"28": "r0:argM|adv",
|
42 |
+
"29": "r0:argM|atr",
|
43 |
+
"30": "r0:argM|cau",
|
44 |
+
"31": "r0:argM|ext",
|
45 |
+
"32": "r0:argM|fin",
|
46 |
+
"33": "r0:argM|ins",
|
47 |
+
"34": "r0:argM|loc",
|
48 |
+
"35": "r0:argM|mnr",
|
49 |
+
"36": "r0:argM|tmp",
|
50 |
+
"37": "r0:root",
|
51 |
+
"38": "r10:arg0|agt",
|
52 |
+
"39": "r10:arg1|pat",
|
53 |
+
"40": "r10:arg1|tem",
|
54 |
+
"41": "r10:arg2|atr",
|
55 |
+
"42": "r10:arg2|ben",
|
56 |
+
"43": "r10:arg2|efi",
|
57 |
+
"44": "r10:arg2|loc",
|
58 |
+
"45": "r10:arg3|ben",
|
59 |
+
"46": "r10:arg4|des",
|
60 |
+
"47": "r10:argM|adv",
|
61 |
+
"48": "r10:argM|atr",
|
62 |
+
"49": "r10:argM|fin",
|
63 |
+
"50": "r10:argM|loc",
|
64 |
+
"51": "r10:argM|tmp",
|
65 |
+
"52": "r10:root",
|
66 |
+
"53": "r11:arg0|agt",
|
67 |
+
"54": "r11:arg0|cau",
|
68 |
+
"55": "r11:arg1|pat",
|
69 |
+
"56": "r11:arg1|tem",
|
70 |
+
"57": "r11:arg2|atr",
|
71 |
+
"58": "r11:arg2|ben",
|
72 |
+
"59": "r11:arg2|loc",
|
73 |
+
"60": "r11:arg4|des",
|
74 |
+
"61": "r11:argM|adv",
|
75 |
+
"62": "r11:argM|loc",
|
76 |
+
"63": "r11:argM|mnr",
|
77 |
+
"64": "r11:argM|tmp",
|
78 |
+
"65": "r11:root",
|
79 |
+
"66": "r12:arg0|agt",
|
80 |
+
"67": "r12:arg0|cau",
|
81 |
+
"68": "r12:arg1|pat",
|
82 |
+
"69": "r12:arg1|tem",
|
83 |
+
"70": "r12:arg2|atr",
|
84 |
+
"71": "r12:argM|adv",
|
85 |
+
"72": "r12:argM|cau",
|
86 |
+
"73": "r12:argM|loc",
|
87 |
+
"74": "r12:argM|tmp",
|
88 |
+
"75": "r12:root",
|
89 |
+
"76": "r13:arg0|agt",
|
90 |
+
"77": "r13:arg0|cau",
|
91 |
+
"78": "r13:arg1|pat",
|
92 |
+
"79": "r13:arg1|tem",
|
93 |
+
"80": "r13:arg2|atr",
|
94 |
+
"81": "r13:arg2|ben",
|
95 |
+
"82": "r13:argM|adv",
|
96 |
+
"83": "r13:argM|atr",
|
97 |
+
"84": "r13:argM|loc",
|
98 |
+
"85": "r13:root",
|
99 |
+
"86": "r14:arg0|agt",
|
100 |
+
"87": "r14:arg1|pat",
|
101 |
+
"88": "r14:arg2|ben",
|
102 |
+
"89": "r14:argM|adv",
|
103 |
+
"90": "r14:argM|loc",
|
104 |
+
"91": "r14:argM|mnr",
|
105 |
+
"92": "r14:root",
|
106 |
+
"93": "r15:arg0|cau",
|
107 |
+
"94": "r15:arg1|tem",
|
108 |
+
"95": "r15:arg2|atr",
|
109 |
+
"96": "r15:arg3|ben",
|
110 |
+
"97": "r15:root",
|
111 |
+
"98": "r16:arg0|agt",
|
112 |
+
"99": "r16:arg0|cau",
|
113 |
+
"100": "r16:arg1|pat",
|
114 |
+
"101": "r16:arg1|tem",
|
115 |
+
"102": "r16:argM|loc",
|
116 |
+
"103": "r16:argM|tmp",
|
117 |
+
"104": "r16:root",
|
118 |
+
"105": "r1:arg0|agt",
|
119 |
+
"106": "r1:arg0|cau",
|
120 |
+
"107": "r1:arg0|exp",
|
121 |
+
"108": "r1:arg0|src",
|
122 |
+
"109": "r1:arg1|ext",
|
123 |
+
"110": "r1:arg1|loc",
|
124 |
+
"111": "r1:arg1|pat",
|
125 |
+
"112": "r1:arg1|tem",
|
126 |
+
"113": "r1:arg2|atr",
|
127 |
+
"114": "r1:arg2|ben",
|
128 |
+
"115": "r1:arg2|efi",
|
129 |
+
"116": "r1:arg2|exp",
|
130 |
+
"117": "r1:arg2|ext",
|
131 |
+
"118": "r1:arg2|ins",
|
132 |
+
"119": "r1:arg2|loc",
|
133 |
+
"120": "r1:arg3|atr",
|
134 |
+
"121": "r1:arg3|ben",
|
135 |
+
"122": "r1:arg3|des",
|
136 |
+
"123": "r1:arg3|ein",
|
137 |
+
"124": "r1:arg3|exp",
|
138 |
+
"125": "r1:arg3|fin",
|
139 |
+
"126": "r1:arg3|ins",
|
140 |
+
"127": "r1:arg3|ori",
|
141 |
+
"128": "r1:arg4|des",
|
142 |
+
"129": "r1:arg4|efi",
|
143 |
+
"130": "r1:argM|adv",
|
144 |
+
"131": "r1:argM|atr",
|
145 |
+
"132": "r1:argM|cau",
|
146 |
+
"133": "r1:argM|ext",
|
147 |
+
"134": "r1:argM|fin",
|
148 |
+
"135": "r1:argM|ins",
|
149 |
+
"136": "r1:argM|loc",
|
150 |
+
"137": "r1:argM|mnr",
|
151 |
+
"138": "r1:argM|tmp",
|
152 |
+
"139": "r1:root",
|
153 |
+
"140": "r2:arg0|agt",
|
154 |
+
"141": "r2:arg0|cau",
|
155 |
+
"142": "r2:arg0|exp",
|
156 |
+
"143": "r2:arg0|src",
|
157 |
+
"144": "r2:arg1|ext",
|
158 |
+
"145": "r2:arg1|loc",
|
159 |
+
"146": "r2:arg1|pat",
|
160 |
+
"147": "r2:arg1|tem",
|
161 |
+
"148": "r2:arg2|atr",
|
162 |
+
"149": "r2:arg2|ben",
|
163 |
+
"150": "r2:arg2|efi",
|
164 |
+
"151": "r2:arg2|exp",
|
165 |
+
"152": "r2:arg2|ext",
|
166 |
+
"153": "r2:arg2|ins",
|
167 |
+
"154": "r2:arg2|loc",
|
168 |
+
"155": "r2:arg3|atr",
|
169 |
+
"156": "r2:arg3|ben",
|
170 |
+
"157": "r2:arg3|ein",
|
171 |
+
"158": "r2:arg3|exp",
|
172 |
+
"159": "r2:arg3|fin",
|
173 |
+
"160": "r2:arg3|loc",
|
174 |
+
"161": "r2:arg3|ori",
|
175 |
+
"162": "r2:arg4|des",
|
176 |
+
"163": "r2:arg4|efi",
|
177 |
+
"164": "r2:argM|adv",
|
178 |
+
"165": "r2:argM|atr",
|
179 |
+
"166": "r2:argM|cau",
|
180 |
+
"167": "r2:argM|ext",
|
181 |
+
"168": "r2:argM|fin",
|
182 |
+
"169": "r2:argM|ins",
|
183 |
+
"170": "r2:argM|loc",
|
184 |
+
"171": "r2:argM|mnr",
|
185 |
+
"172": "r2:argM|tmp",
|
186 |
+
"173": "r2:root",
|
187 |
+
"174": "r3:arg0|agt",
|
188 |
+
"175": "r3:arg0|cau",
|
189 |
+
"176": "r3:arg0|exp",
|
190 |
+
"177": "r3:arg0|src",
|
191 |
+
"178": "r3:arg1|ext",
|
192 |
+
"179": "r3:arg1|loc",
|
193 |
+
"180": "r3:arg1|pat",
|
194 |
+
"181": "r3:arg1|tem",
|
195 |
+
"182": "r3:arg2|atr",
|
196 |
+
"183": "r3:arg2|ben",
|
197 |
+
"184": "r3:arg2|efi",
|
198 |
+
"185": "r3:arg2|exp",
|
199 |
+
"186": "r3:arg2|ext",
|
200 |
+
"187": "r3:arg2|ins",
|
201 |
+
"188": "r3:arg2|loc",
|
202 |
+
"189": "r3:arg2|tem",
|
203 |
+
"190": "r3:arg3|ben",
|
204 |
+
"191": "r3:arg3|ein",
|
205 |
+
"192": "r3:arg3|fin",
|
206 |
+
"193": "r3:arg3|loc",
|
207 |
+
"194": "r3:arg3|ori",
|
208 |
+
"195": "r3:arg4|des",
|
209 |
+
"196": "r3:arg4|efi",
|
210 |
+
"197": "r3:argM|adv",
|
211 |
+
"198": "r3:argM|atr",
|
212 |
+
"199": "r3:argM|cau",
|
213 |
+
"200": "r3:argM|ext",
|
214 |
+
"201": "r3:argM|fin",
|
215 |
+
"202": "r3:argM|ins",
|
216 |
+
"203": "r3:argM|loc",
|
217 |
+
"204": "r3:argM|mnr",
|
218 |
+
"205": "r3:argM|tmp",
|
219 |
+
"206": "r3:root",
|
220 |
+
"207": "r4:arg0|agt",
|
221 |
+
"208": "r4:arg0|cau",
|
222 |
+
"209": "r4:arg0|exp",
|
223 |
+
"210": "r4:arg0|src",
|
224 |
+
"211": "r4:arg1|ext",
|
225 |
+
"212": "r4:arg1|loc",
|
226 |
+
"213": "r4:arg1|pat",
|
227 |
+
"214": "r4:arg1|tem",
|
228 |
+
"215": "r4:arg2|atr",
|
229 |
+
"216": "r4:arg2|ben",
|
230 |
+
"217": "r4:arg2|efi",
|
231 |
+
"218": "r4:arg2|exp",
|
232 |
+
"219": "r4:arg2|ext",
|
233 |
+
"220": "r4:arg2|ins",
|
234 |
+
"221": "r4:arg2|loc",
|
235 |
+
"222": "r4:arg3|ben",
|
236 |
+
"223": "r4:arg3|ein",
|
237 |
+
"224": "r4:arg3|exp",
|
238 |
+
"225": "r4:arg3|fin",
|
239 |
+
"226": "r4:arg3|ori",
|
240 |
+
"227": "r4:arg4|des",
|
241 |
+
"228": "r4:arg4|efi",
|
242 |
+
"229": "r4:argM|adv",
|
243 |
+
"230": "r4:argM|atr",
|
244 |
+
"231": "r4:argM|cau",
|
245 |
+
"232": "r4:argM|ext",
|
246 |
+
"233": "r4:argM|fin",
|
247 |
+
"234": "r4:argM|ins",
|
248 |
+
"235": "r4:argM|loc",
|
249 |
+
"236": "r4:argM|mnr",
|
250 |
+
"237": "r4:argM|tmp",
|
251 |
+
"238": "r4:root",
|
252 |
+
"239": "r5:arg0|agt",
|
253 |
+
"240": "r5:arg0|cau",
|
254 |
+
"241": "r5:arg1|ext",
|
255 |
+
"242": "r5:arg1|loc",
|
256 |
+
"243": "r5:arg1|pat",
|
257 |
+
"244": "r5:arg1|tem",
|
258 |
+
"245": "r5:arg2|atr",
|
259 |
+
"246": "r5:arg2|ben",
|
260 |
+
"247": "r5:arg2|efi",
|
261 |
+
"248": "r5:arg2|exp",
|
262 |
+
"249": "r5:arg2|ext",
|
263 |
+
"250": "r5:arg2|loc",
|
264 |
+
"251": "r5:arg3|ben",
|
265 |
+
"252": "r5:arg3|ein",
|
266 |
+
"253": "r5:arg3|fin",
|
267 |
+
"254": "r5:arg3|ins",
|
268 |
+
"255": "r5:arg3|ori",
|
269 |
+
"256": "r5:arg4|des",
|
270 |
+
"257": "r5:arg4|efi",
|
271 |
+
"258": "r5:argM|adv",
|
272 |
+
"259": "r5:argM|atr",
|
273 |
+
"260": "r5:argM|cau",
|
274 |
+
"261": "r5:argM|ext",
|
275 |
+
"262": "r5:argM|fin",
|
276 |
+
"263": "r5:argM|loc",
|
277 |
+
"264": "r5:argM|mnr",
|
278 |
+
"265": "r5:argM|tmp",
|
279 |
+
"266": "r5:root",
|
280 |
+
"267": "r6:arg0|agt",
|
281 |
+
"268": "r6:arg0|cau",
|
282 |
+
"269": "r6:arg1|loc",
|
283 |
+
"270": "r6:arg1|pat",
|
284 |
+
"271": "r6:arg1|tem",
|
285 |
+
"272": "r6:arg2|atr",
|
286 |
+
"273": "r6:arg2|ben",
|
287 |
+
"274": "r6:arg2|efi",
|
288 |
+
"275": "r6:arg2|exp",
|
289 |
+
"276": "r6:arg2|ext",
|
290 |
+
"277": "r6:arg2|loc",
|
291 |
+
"278": "r6:arg3|ben",
|
292 |
+
"279": "r6:arg3|ori",
|
293 |
+
"280": "r6:arg4|des",
|
294 |
+
"281": "r6:argM|adv",
|
295 |
+
"282": "r6:argM|atr",
|
296 |
+
"283": "r6:argM|cau",
|
297 |
+
"284": "r6:argM|ext",
|
298 |
+
"285": "r6:argM|fin",
|
299 |
+
"286": "r6:argM|loc",
|
300 |
+
"287": "r6:argM|mnr",
|
301 |
+
"288": "r6:argM|tmp",
|
302 |
+
"289": "r6:root",
|
303 |
+
"290": "r7:arg0|agt",
|
304 |
+
"291": "r7:arg0|cau",
|
305 |
+
"292": "r7:arg1|loc",
|
306 |
+
"293": "r7:arg1|pat",
|
307 |
+
"294": "r7:arg1|tem",
|
308 |
+
"295": "r7:arg2|atr",
|
309 |
+
"296": "r7:arg2|ben",
|
310 |
+
"297": "r7:arg2|efi",
|
311 |
+
"298": "r7:arg2|loc",
|
312 |
+
"299": "r7:arg3|ben",
|
313 |
+
"300": "r7:arg3|exp",
|
314 |
+
"301": "r7:arg3|fin",
|
315 |
+
"302": "r7:arg3|ori",
|
316 |
+
"303": "r7:arg4|des",
|
317 |
+
"304": "r7:argM|adv",
|
318 |
+
"305": "r7:argM|atr",
|
319 |
+
"306": "r7:argM|cau",
|
320 |
+
"307": "r7:argM|fin",
|
321 |
+
"308": "r7:argM|ins",
|
322 |
+
"309": "r7:argM|loc",
|
323 |
+
"310": "r7:argM|mnr",
|
324 |
+
"311": "r7:argM|tmp",
|
325 |
+
"312": "r7:root",
|
326 |
+
"313": "r8:arg0|agt",
|
327 |
+
"314": "r8:arg0|cau",
|
328 |
+
"315": "r8:arg0|src",
|
329 |
+
"316": "r8:arg1|pat",
|
330 |
+
"317": "r8:arg1|tem",
|
331 |
+
"318": "r8:arg2|atr",
|
332 |
+
"319": "r8:arg2|ben",
|
333 |
+
"320": "r8:arg2|ext",
|
334 |
+
"321": "r8:arg2|loc",
|
335 |
+
"322": "r8:arg3|ori",
|
336 |
+
"323": "r8:arg4|des",
|
337 |
+
"324": "r8:argM|adv",
|
338 |
+
"325": "r8:argM|cau",
|
339 |
+
"326": "r8:argM|ext",
|
340 |
+
"327": "r8:argM|fin",
|
341 |
+
"328": "r8:argM|loc",
|
342 |
+
"329": "r8:argM|mnr",
|
343 |
+
"330": "r8:argM|tmp",
|
344 |
+
"331": "r8:root",
|
345 |
+
"332": "r9:arg0|agt",
|
346 |
+
"333": "r9:arg0|cau",
|
347 |
+
"334": "r9:arg1|pat",
|
348 |
+
"335": "r9:arg1|tem",
|
349 |
+
"336": "r9:arg2|atr",
|
350 |
+
"337": "r9:arg2|ben",
|
351 |
+
"338": "r9:arg2|ins",
|
352 |
+
"339": "r9:arg2|loc",
|
353 |
+
"340": "r9:arg4|des",
|
354 |
+
"341": "r9:argM|adv",
|
355 |
+
"342": "r9:argM|cau",
|
356 |
+
"343": "r9:argM|fin",
|
357 |
+
"344": "r9:argM|loc",
|
358 |
+
"345": "r9:argM|mnr",
|
359 |
+
"346": "r9:argM|tmp",
|
360 |
+
"347": "r9:root"
|
361 |
+
},
|
362 |
+
"initializer_range": 0.02,
|
363 |
+
"intermediate_size": 3072,
|
364 |
+
"label2id": {
|
365 |
+
"O": 0,
|
366 |
+
"r0:arg0|agt": 1,
|
367 |
+
"r0:arg0|cau": 2,
|
368 |
+
"r0:arg0|exp": 3,
|
369 |
+
"r0:arg0|pat": 4,
|
370 |
+
"r0:arg0|src": 5,
|
371 |
+
"r0:arg1|ext": 6,
|
372 |
+
"r0:arg1|loc": 7,
|
373 |
+
"r0:arg1|pat": 8,
|
374 |
+
"r0:arg1|tem": 9,
|
375 |
+
"r0:arg2|atr": 10,
|
376 |
+
"r0:arg2|ben": 11,
|
377 |
+
"r0:arg2|efi": 12,
|
378 |
+
"r0:arg2|exp": 13,
|
379 |
+
"r0:arg2|ext": 14,
|
380 |
+
"r0:arg2|ins": 15,
|
381 |
+
"r0:arg2|loc": 16,
|
382 |
+
"r0:arg2|tem": 17,
|
383 |
+
"r0:arg3|ben": 18,
|
384 |
+
"r0:arg3|ein": 19,
|
385 |
+
"r0:arg3|exp": 20,
|
386 |
+
"r0:arg3|fin": 21,
|
387 |
+
"r0:arg3|ins": 22,
|
388 |
+
"r0:arg3|loc": 23,
|
389 |
+
"r0:arg3|ori": 24,
|
390 |
+
"r0:arg4|des": 25,
|
391 |
+
"r0:arg4|efi": 26,
|
392 |
+
"r0:argM|LOC": 27,
|
393 |
+
"r0:argM|adv": 28,
|
394 |
+
"r0:argM|atr": 29,
|
395 |
+
"r0:argM|cau": 30,
|
396 |
+
"r0:argM|ext": 31,
|
397 |
+
"r0:argM|fin": 32,
|
398 |
+
"r0:argM|ins": 33,
|
399 |
+
"r0:argM|loc": 34,
|
400 |
+
"r0:argM|mnr": 35,
|
401 |
+
"r0:argM|tmp": 36,
|
402 |
+
"r0:root": 37,
|
403 |
+
"r10:arg0|agt": 38,
|
404 |
+
"r10:arg1|pat": 39,
|
405 |
+
"r10:arg1|tem": 40,
|
406 |
+
"r10:arg2|atr": 41,
|
407 |
+
"r10:arg2|ben": 42,
|
408 |
+
"r10:arg2|efi": 43,
|
409 |
+
"r10:arg2|loc": 44,
|
410 |
+
"r10:arg3|ben": 45,
|
411 |
+
"r10:arg4|des": 46,
|
412 |
+
"r10:argM|adv": 47,
|
413 |
+
"r10:argM|atr": 48,
|
414 |
+
"r10:argM|fin": 49,
|
415 |
+
"r10:argM|loc": 50,
|
416 |
+
"r10:argM|tmp": 51,
|
417 |
+
"r10:root": 52,
|
418 |
+
"r11:arg0|agt": 53,
|
419 |
+
"r11:arg0|cau": 54,
|
420 |
+
"r11:arg1|pat": 55,
|
421 |
+
"r11:arg1|tem": 56,
|
422 |
+
"r11:arg2|atr": 57,
|
423 |
+
"r11:arg2|ben": 58,
|
424 |
+
"r11:arg2|loc": 59,
|
425 |
+
"r11:arg4|des": 60,
|
426 |
+
"r11:argM|adv": 61,
|
427 |
+
"r11:argM|loc": 62,
|
428 |
+
"r11:argM|mnr": 63,
|
429 |
+
"r11:argM|tmp": 64,
|
430 |
+
"r11:root": 65,
|
431 |
+
"r12:arg0|agt": 66,
|
432 |
+
"r12:arg0|cau": 67,
|
433 |
+
"r12:arg1|pat": 68,
|
434 |
+
"r12:arg1|tem": 69,
|
435 |
+
"r12:arg2|atr": 70,
|
436 |
+
"r12:argM|adv": 71,
|
437 |
+
"r12:argM|cau": 72,
|
438 |
+
"r12:argM|loc": 73,
|
439 |
+
"r12:argM|tmp": 74,
|
440 |
+
"r12:root": 75,
|
441 |
+
"r13:arg0|agt": 76,
|
442 |
+
"r13:arg0|cau": 77,
|
443 |
+
"r13:arg1|pat": 78,
|
444 |
+
"r13:arg1|tem": 79,
|
445 |
+
"r13:arg2|atr": 80,
|
446 |
+
"r13:arg2|ben": 81,
|
447 |
+
"r13:argM|adv": 82,
|
448 |
+
"r13:argM|atr": 83,
|
449 |
+
"r13:argM|loc": 84,
|
450 |
+
"r13:root": 85,
|
451 |
+
"r14:arg0|agt": 86,
|
452 |
+
"r14:arg1|pat": 87,
|
453 |
+
"r14:arg2|ben": 88,
|
454 |
+
"r14:argM|adv": 89,
|
455 |
+
"r14:argM|loc": 90,
|
456 |
+
"r14:argM|mnr": 91,
|
457 |
+
"r14:root": 92,
|
458 |
+
"r15:arg0|cau": 93,
|
459 |
+
"r15:arg1|tem": 94,
|
460 |
+
"r15:arg2|atr": 95,
|
461 |
+
"r15:arg3|ben": 96,
|
462 |
+
"r15:root": 97,
|
463 |
+
"r16:arg0|agt": 98,
|
464 |
+
"r16:arg0|cau": 99,
|
465 |
+
"r16:arg1|pat": 100,
|
466 |
+
"r16:arg1|tem": 101,
|
467 |
+
"r16:argM|loc": 102,
|
468 |
+
"r16:argM|tmp": 103,
|
469 |
+
"r16:root": 104,
|
470 |
+
"r1:arg0|agt": 105,
|
471 |
+
"r1:arg0|cau": 106,
|
472 |
+
"r1:arg0|exp": 107,
|
473 |
+
"r1:arg0|src": 108,
|
474 |
+
"r1:arg1|ext": 109,
|
475 |
+
"r1:arg1|loc": 110,
|
476 |
+
"r1:arg1|pat": 111,
|
477 |
+
"r1:arg1|tem": 112,
|
478 |
+
"r1:arg2|atr": 113,
|
479 |
+
"r1:arg2|ben": 114,
|
480 |
+
"r1:arg2|efi": 115,
|
481 |
+
"r1:arg2|exp": 116,
|
482 |
+
"r1:arg2|ext": 117,
|
483 |
+
"r1:arg2|ins": 118,
|
484 |
+
"r1:arg2|loc": 119,
|
485 |
+
"r1:arg3|atr": 120,
|
486 |
+
"r1:arg3|ben": 121,
|
487 |
+
"r1:arg3|des": 122,
|
488 |
+
"r1:arg3|ein": 123,
|
489 |
+
"r1:arg3|exp": 124,
|
490 |
+
"r1:arg3|fin": 125,
|
491 |
+
"r1:arg3|ins": 126,
|
492 |
+
"r1:arg3|ori": 127,
|
493 |
+
"r1:arg4|des": 128,
|
494 |
+
"r1:arg4|efi": 129,
|
495 |
+
"r1:argM|adv": 130,
|
496 |
+
"r1:argM|atr": 131,
|
497 |
+
"r1:argM|cau": 132,
|
498 |
+
"r1:argM|ext": 133,
|
499 |
+
"r1:argM|fin": 134,
|
500 |
+
"r1:argM|ins": 135,
|
501 |
+
"r1:argM|loc": 136,
|
502 |
+
"r1:argM|mnr": 137,
|
503 |
+
"r1:argM|tmp": 138,
|
504 |
+
"r1:root": 139,
|
505 |
+
"r2:arg0|agt": 140,
|
506 |
+
"r2:arg0|cau": 141,
|
507 |
+
"r2:arg0|exp": 142,
|
508 |
+
"r2:arg0|src": 143,
|
509 |
+
"r2:arg1|ext": 144,
|
510 |
+
"r2:arg1|loc": 145,
|
511 |
+
"r2:arg1|pat": 146,
|
512 |
+
"r2:arg1|tem": 147,
|
513 |
+
"r2:arg2|atr": 148,
|
514 |
+
"r2:arg2|ben": 149,
|
515 |
+
"r2:arg2|efi": 150,
|
516 |
+
"r2:arg2|exp": 151,
|
517 |
+
"r2:arg2|ext": 152,
|
518 |
+
"r2:arg2|ins": 153,
|
519 |
+
"r2:arg2|loc": 154,
|
520 |
+
"r2:arg3|atr": 155,
|
521 |
+
"r2:arg3|ben": 156,
|
522 |
+
"r2:arg3|ein": 157,
|
523 |
+
"r2:arg3|exp": 158,
|
524 |
+
"r2:arg3|fin": 159,
|
525 |
+
"r2:arg3|loc": 160,
|
526 |
+
"r2:arg3|ori": 161,
|
527 |
+
"r2:arg4|des": 162,
|
528 |
+
"r2:arg4|efi": 163,
|
529 |
+
"r2:argM|adv": 164,
|
530 |
+
"r2:argM|atr": 165,
|
531 |
+
"r2:argM|cau": 166,
|
532 |
+
"r2:argM|ext": 167,
|
533 |
+
"r2:argM|fin": 168,
|
534 |
+
"r2:argM|ins": 169,
|
535 |
+
"r2:argM|loc": 170,
|
536 |
+
"r2:argM|mnr": 171,
|
537 |
+
"r2:argM|tmp": 172,
|
538 |
+
"r2:root": 173,
|
539 |
+
"r3:arg0|agt": 174,
|
540 |
+
"r3:arg0|cau": 175,
|
541 |
+
"r3:arg0|exp": 176,
|
542 |
+
"r3:arg0|src": 177,
|
543 |
+
"r3:arg1|ext": 178,
|
544 |
+
"r3:arg1|loc": 179,
|
545 |
+
"r3:arg1|pat": 180,
|
546 |
+
"r3:arg1|tem": 181,
|
547 |
+
"r3:arg2|atr": 182,
|
548 |
+
"r3:arg2|ben": 183,
|
549 |
+
"r3:arg2|efi": 184,
|
550 |
+
"r3:arg2|exp": 185,
|
551 |
+
"r3:arg2|ext": 186,
|
552 |
+
"r3:arg2|ins": 187,
|
553 |
+
"r3:arg2|loc": 188,
|
554 |
+
"r3:arg2|tem": 189,
|
555 |
+
"r3:arg3|ben": 190,
|
556 |
+
"r3:arg3|ein": 191,
|
557 |
+
"r3:arg3|fin": 192,
|
558 |
+
"r3:arg3|loc": 193,
|
559 |
+
"r3:arg3|ori": 194,
|
560 |
+
"r3:arg4|des": 195,
|
561 |
+
"r3:arg4|efi": 196,
|
562 |
+
"r3:argM|adv": 197,
|
563 |
+
"r3:argM|atr": 198,
|
564 |
+
"r3:argM|cau": 199,
|
565 |
+
"r3:argM|ext": 200,
|
566 |
+
"r3:argM|fin": 201,
|
567 |
+
"r3:argM|ins": 202,
|
568 |
+
"r3:argM|loc": 203,
|
569 |
+
"r3:argM|mnr": 204,
|
570 |
+
"r3:argM|tmp": 205,
|
571 |
+
"r3:root": 206,
|
572 |
+
"r4:arg0|agt": 207,
|
573 |
+
"r4:arg0|cau": 208,
|
574 |
+
"r4:arg0|exp": 209,
|
575 |
+
"r4:arg0|src": 210,
|
576 |
+
"r4:arg1|ext": 211,
|
577 |
+
"r4:arg1|loc": 212,
|
578 |
+
"r4:arg1|pat": 213,
|
579 |
+
"r4:arg1|tem": 214,
|
580 |
+
"r4:arg2|atr": 215,
|
581 |
+
"r4:arg2|ben": 216,
|
582 |
+
"r4:arg2|efi": 217,
|
583 |
+
"r4:arg2|exp": 218,
|
584 |
+
"r4:arg2|ext": 219,
|
585 |
+
"r4:arg2|ins": 220,
|
586 |
+
"r4:arg2|loc": 221,
|
587 |
+
"r4:arg3|ben": 222,
|
588 |
+
"r4:arg3|ein": 223,
|
589 |
+
"r4:arg3|exp": 224,
|
590 |
+
"r4:arg3|fin": 225,
|
591 |
+
"r4:arg3|ori": 226,
|
592 |
+
"r4:arg4|des": 227,
|
593 |
+
"r4:arg4|efi": 228,
|
594 |
+
"r4:argM|adv": 229,
|
595 |
+
"r4:argM|atr": 230,
|
596 |
+
"r4:argM|cau": 231,
|
597 |
+
"r4:argM|ext": 232,
|
598 |
+
"r4:argM|fin": 233,
|
599 |
+
"r4:argM|ins": 234,
|
600 |
+
"r4:argM|loc": 235,
|
601 |
+
"r4:argM|mnr": 236,
|
602 |
+
"r4:argM|tmp": 237,
|
603 |
+
"r4:root": 238,
|
604 |
+
"r5:arg0|agt": 239,
|
605 |
+
"r5:arg0|cau": 240,
|
606 |
+
"r5:arg1|ext": 241,
|
607 |
+
"r5:arg1|loc": 242,
|
608 |
+
"r5:arg1|pat": 243,
|
609 |
+
"r5:arg1|tem": 244,
|
610 |
+
"r5:arg2|atr": 245,
|
611 |
+
"r5:arg2|ben": 246,
|
612 |
+
"r5:arg2|efi": 247,
|
613 |
+
"r5:arg2|exp": 248,
|
614 |
+
"r5:arg2|ext": 249,
|
615 |
+
"r5:arg2|loc": 250,
|
616 |
+
"r5:arg3|ben": 251,
|
617 |
+
"r5:arg3|ein": 252,
|
618 |
+
"r5:arg3|fin": 253,
|
619 |
+
"r5:arg3|ins": 254,
|
620 |
+
"r5:arg3|ori": 255,
|
621 |
+
"r5:arg4|des": 256,
|
622 |
+
"r5:arg4|efi": 257,
|
623 |
+
"r5:argM|adv": 258,
|
624 |
+
"r5:argM|atr": 259,
|
625 |
+
"r5:argM|cau": 260,
|
626 |
+
"r5:argM|ext": 261,
|
627 |
+
"r5:argM|fin": 262,
|
628 |
+
"r5:argM|loc": 263,
|
629 |
+
"r5:argM|mnr": 264,
|
630 |
+
"r5:argM|tmp": 265,
|
631 |
+
"r5:root": 266,
|
632 |
+
"r6:arg0|agt": 267,
|
633 |
+
"r6:arg0|cau": 268,
|
634 |
+
"r6:arg1|loc": 269,
|
635 |
+
"r6:arg1|pat": 270,
|
636 |
+
"r6:arg1|tem": 271,
|
637 |
+
"r6:arg2|atr": 272,
|
638 |
+
"r6:arg2|ben": 273,
|
639 |
+
"r6:arg2|efi": 274,
|
640 |
+
"r6:arg2|exp": 275,
|
641 |
+
"r6:arg2|ext": 276,
|
642 |
+
"r6:arg2|loc": 277,
|
643 |
+
"r6:arg3|ben": 278,
|
644 |
+
"r6:arg3|ori": 279,
|
645 |
+
"r6:arg4|des": 280,
|
646 |
+
"r6:argM|adv": 281,
|
647 |
+
"r6:argM|atr": 282,
|
648 |
+
"r6:argM|cau": 283,
|
649 |
+
"r6:argM|ext": 284,
|
650 |
+
"r6:argM|fin": 285,
|
651 |
+
"r6:argM|loc": 286,
|
652 |
+
"r6:argM|mnr": 287,
|
653 |
+
"r6:argM|tmp": 288,
|
654 |
+
"r6:root": 289,
|
655 |
+
"r7:arg0|agt": 290,
|
656 |
+
"r7:arg0|cau": 291,
|
657 |
+
"r7:arg1|loc": 292,
|
658 |
+
"r7:arg1|pat": 293,
|
659 |
+
"r7:arg1|tem": 294,
|
660 |
+
"r7:arg2|atr": 295,
|
661 |
+
"r7:arg2|ben": 296,
|
662 |
+
"r7:arg2|efi": 297,
|
663 |
+
"r7:arg2|loc": 298,
|
664 |
+
"r7:arg3|ben": 299,
|
665 |
+
"r7:arg3|exp": 300,
|
666 |
+
"r7:arg3|fin": 301,
|
667 |
+
"r7:arg3|ori": 302,
|
668 |
+
"r7:arg4|des": 303,
|
669 |
+
"r7:argM|adv": 304,
|
670 |
+
"r7:argM|atr": 305,
|
671 |
+
"r7:argM|cau": 306,
|
672 |
+
"r7:argM|fin": 307,
|
673 |
+
"r7:argM|ins": 308,
|
674 |
+
"r7:argM|loc": 309,
|
675 |
+
"r7:argM|mnr": 310,
|
676 |
+
"r7:argM|tmp": 311,
|
677 |
+
"r7:root": 312,
|
678 |
+
"r8:arg0|agt": 313,
|
679 |
+
"r8:arg0|cau": 314,
|
680 |
+
"r8:arg0|src": 315,
|
681 |
+
"r8:arg1|pat": 316,
|
682 |
+
"r8:arg1|tem": 317,
|
683 |
+
"r8:arg2|atr": 318,
|
684 |
+
"r8:arg2|ben": 319,
|
685 |
+
"r8:arg2|ext": 320,
|
686 |
+
"r8:arg2|loc": 321,
|
687 |
+
"r8:arg3|ori": 322,
|
688 |
+
"r8:arg4|des": 323,
|
689 |
+
"r8:argM|adv": 324,
|
690 |
+
"r8:argM|cau": 325,
|
691 |
+
"r8:argM|ext": 326,
|
692 |
+
"r8:argM|fin": 327,
|
693 |
+
"r8:argM|loc": 328,
|
694 |
+
"r8:argM|mnr": 329,
|
695 |
+
"r8:argM|tmp": 330,
|
696 |
+
"r8:root": 331,
|
697 |
+
"r9:arg0|agt": 332,
|
698 |
+
"r9:arg0|cau": 333,
|
699 |
+
"r9:arg1|pat": 334,
|
700 |
+
"r9:arg1|tem": 335,
|
701 |
+
"r9:arg2|atr": 336,
|
702 |
+
"r9:arg2|ben": 337,
|
703 |
+
"r9:arg2|ins": 338,
|
704 |
+
"r9:arg2|loc": 339,
|
705 |
+
"r9:arg4|des": 340,
|
706 |
+
"r9:argM|adv": 341,
|
707 |
+
"r9:argM|cau": 342,
|
708 |
+
"r9:argM|fin": 343,
|
709 |
+
"r9:argM|loc": 344,
|
710 |
+
"r9:argM|mnr": 345,
|
711 |
+
"r9:argM|tmp": 346,
|
712 |
+
"r9:root": 347
|
713 |
+
},
|
714 |
+
"layer_norm_eps": 1e-12,
|
715 |
+
"max_position_embeddings": 512,
|
716 |
+
"model_type": "bert",
|
717 |
+
"num_attention_heads": 12,
|
718 |
+
"num_hidden_layers": 12,
|
719 |
+
"pad_token_id": 0,
|
720 |
+
"pooler_fc_size": 768,
|
721 |
+
"pooler_num_attention_heads": 12,
|
722 |
+
"pooler_num_fc_layers": 3,
|
723 |
+
"pooler_size_per_head": 128,
|
724 |
+
"pooler_type": "first_token_transform",
|
725 |
+
"position_embedding_type": "absolute",
|
726 |
+
"torch_dtype": "float32",
|
727 |
+
"transformers_version": "4.27.4",
|
728 |
+
"type_vocab_size": 2,
|
729 |
+
"use_cache": true,
|
730 |
+
"vocab_size": 119547
|
731 |
+
}
|
optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8d70f5622d7a0415a8ac4e17332756192a062d3c63cea12c6ef97f179d0e22c4
|
3 |
+
size 1420359013
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9541b21ecd9ab7f824a6d855192d0e34336a96f6c94ed60a83f75443d8038027
|
3 |
+
size 710192241
|
rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3c962e153722e566150c79c36b6bd83c2627153553dd74c646e178307c3bed0c
|
3 |
+
size 14503
|
scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8bc9161f4711579afeb0ec838a40ec2eed71060e7b11126ad5772ff1ec638312
|
3 |
+
size 623
|
special_tokens_map.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cls_token": "[CLS]",
|
3 |
+
"mask_token": "[MASK]",
|
4 |
+
"pad_token": "[PAD]",
|
5 |
+
"sep_token": "[SEP]",
|
6 |
+
"unk_token": "[UNK]"
|
7 |
+
}
|
tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
tokenizer_config.json
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cls_token": "[CLS]",
|
3 |
+
"do_lower_case": false,
|
4 |
+
"mask_token": "[MASK]",
|
5 |
+
"model_max_length": 512,
|
6 |
+
"pad_token": "[PAD]",
|
7 |
+
"sep_token": "[SEP]",
|
8 |
+
"special_tokens_map_file": null,
|
9 |
+
"strip_accents": null,
|
10 |
+
"tokenize_chinese_chars": true,
|
11 |
+
"tokenizer_class": "BertTokenizer",
|
12 |
+
"unk_token": "[UNK]"
|
13 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,148 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.26231876015663147,
|
3 |
+
"best_model_checkpoint": "spa_mBERT/checkpoint-4480",
|
4 |
+
"epoch": 6.0,
|
5 |
+
"global_step": 5376,
|
6 |
+
"is_hyper_param_search": false,
|
7 |
+
"is_local_process_zero": true,
|
8 |
+
"is_world_process_zero": true,
|
9 |
+
"log_history": [
|
10 |
+
{
|
11 |
+
"epoch": 0.56,
|
12 |
+
"learning_rate": 1.988839285714286e-05,
|
13 |
+
"loss": 1.0204,
|
14 |
+
"step": 500
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"epoch": 1.0,
|
18 |
+
"eval_accuracy": 0.8996683415089468,
|
19 |
+
"eval_f1": 0.7103438993202774,
|
20 |
+
"eval_loss": 0.3859018087387085,
|
21 |
+
"eval_precision": 0.7230442526373476,
|
22 |
+
"eval_recall": 0.6980820105820106,
|
23 |
+
"eval_runtime": 5.926,
|
24 |
+
"eval_samples_per_second": 279.11,
|
25 |
+
"eval_steps_per_second": 17.55,
|
26 |
+
"step": 896
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"epoch": 1.12,
|
30 |
+
"learning_rate": 1.9776785714285716e-05,
|
31 |
+
"loss": 0.4747,
|
32 |
+
"step": 1000
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"epoch": 1.67,
|
36 |
+
"learning_rate": 1.9665178571428575e-05,
|
37 |
+
"loss": 0.3496,
|
38 |
+
"step": 1500
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"epoch": 2.0,
|
42 |
+
"eval_accuracy": 0.917184676186126,
|
43 |
+
"eval_f1": 0.764059043027441,
|
44 |
+
"eval_loss": 0.30577903985977173,
|
45 |
+
"eval_precision": 0.7629756644463497,
|
46 |
+
"eval_recall": 0.7651455026455026,
|
47 |
+
"eval_runtime": 5.9393,
|
48 |
+
"eval_samples_per_second": 278.485,
|
49 |
+
"eval_steps_per_second": 17.511,
|
50 |
+
"step": 1792
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"epoch": 2.23,
|
54 |
+
"learning_rate": 1.955357142857143e-05,
|
55 |
+
"loss": 0.3017,
|
56 |
+
"step": 2000
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"epoch": 2.79,
|
60 |
+
"learning_rate": 1.944196428571429e-05,
|
61 |
+
"loss": 0.2568,
|
62 |
+
"step": 2500
|
63 |
+
},
|
64 |
+
{
|
65 |
+
"epoch": 3.0,
|
66 |
+
"eval_accuracy": 0.9256648064663476,
|
67 |
+
"eval_f1": 0.7892974370876299,
|
68 |
+
"eval_loss": 0.28278017044067383,
|
69 |
+
"eval_precision": 0.7913702546373247,
|
70 |
+
"eval_recall": 0.7872354497354498,
|
71 |
+
"eval_runtime": 6.0079,
|
72 |
+
"eval_samples_per_second": 275.304,
|
73 |
+
"eval_steps_per_second": 17.311,
|
74 |
+
"step": 2688
|
75 |
+
},
|
76 |
+
{
|
77 |
+
"epoch": 3.35,
|
78 |
+
"learning_rate": 1.9330357142857144e-05,
|
79 |
+
"loss": 0.22,
|
80 |
+
"step": 3000
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"epoch": 3.91,
|
84 |
+
"learning_rate": 1.9218750000000003e-05,
|
85 |
+
"loss": 0.1981,
|
86 |
+
"step": 3500
|
87 |
+
},
|
88 |
+
{
|
89 |
+
"epoch": 4.0,
|
90 |
+
"eval_accuracy": 0.9319405000695092,
|
91 |
+
"eval_f1": 0.810671256454389,
|
92 |
+
"eval_loss": 0.26689934730529785,
|
93 |
+
"eval_precision": 0.8114232706069441,
|
94 |
+
"eval_recall": 0.8099206349206349,
|
95 |
+
"eval_runtime": 6.0528,
|
96 |
+
"eval_samples_per_second": 273.261,
|
97 |
+
"eval_steps_per_second": 17.182,
|
98 |
+
"step": 3584
|
99 |
+
},
|
100 |
+
{
|
101 |
+
"epoch": 4.46,
|
102 |
+
"learning_rate": 1.910714285714286e-05,
|
103 |
+
"loss": 0.1757,
|
104 |
+
"step": 4000
|
105 |
+
},
|
106 |
+
{
|
107 |
+
"epoch": 5.0,
|
108 |
+
"eval_accuracy": 0.9351975056103906,
|
109 |
+
"eval_f1": 0.8214273897362936,
|
110 |
+
"eval_loss": 0.26231876015663147,
|
111 |
+
"eval_precision": 0.8218896907899093,
|
112 |
+
"eval_recall": 0.8209656084656085,
|
113 |
+
"eval_runtime": 5.8306,
|
114 |
+
"eval_samples_per_second": 283.675,
|
115 |
+
"eval_steps_per_second": 17.837,
|
116 |
+
"step": 4480
|
117 |
+
},
|
118 |
+
{
|
119 |
+
"epoch": 5.02,
|
120 |
+
"learning_rate": 1.8995535714285717e-05,
|
121 |
+
"loss": 0.1648,
|
122 |
+
"step": 4500
|
123 |
+
},
|
124 |
+
{
|
125 |
+
"epoch": 5.58,
|
126 |
+
"learning_rate": 1.8883928571428573e-05,
|
127 |
+
"loss": 0.1388,
|
128 |
+
"step": 5000
|
129 |
+
},
|
130 |
+
{
|
131 |
+
"epoch": 6.0,
|
132 |
+
"eval_accuracy": 0.9327348916648461,
|
133 |
+
"eval_f1": 0.8123542346904777,
|
134 |
+
"eval_loss": 0.2874906361103058,
|
135 |
+
"eval_precision": 0.8185175238351013,
|
136 |
+
"eval_recall": 0.8062830687830688,
|
137 |
+
"eval_runtime": 6.0062,
|
138 |
+
"eval_samples_per_second": 275.382,
|
139 |
+
"eval_steps_per_second": 17.315,
|
140 |
+
"step": 5376
|
141 |
+
}
|
142 |
+
],
|
143 |
+
"max_steps": 89600,
|
144 |
+
"num_train_epochs": 100,
|
145 |
+
"total_flos": 3919265715030336.0,
|
146 |
+
"trial_name": null,
|
147 |
+
"trial_params": null
|
148 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:63d065865deb1cbc1e597d0ee8fff7d3df3466da4bbddf59f6d4f504ff375ffe
|
3 |
+
size 3503
|
vocab.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|