Hugging Face's logo

zhengxuanzenwu
/
intervenable_honest_llama2_chat_7B

Transformers
Inference Endpoints
Model card Files Community
zhengxuanzenwu commited on
Commit
5411e78
•
1 Parent(s): 86c7bf7

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +280 -0
config.json ADDED
@@ -0,0 +1,280 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "intervention_constant_sources": [
3
+ true,
4
+ true,
5
+ true,
6
+ true,
7
+ true,
8
+ true,
9
+ true,
10
+ true,
11
+ true,
12
+ true,
13
+ true,
14
+ true,
15
+ true,
16
+ true
17
+ ],
18
+ "intervention_dimensions": [
19
+ null,
20
+ null,
21
+ null,
22
+ null,
23
+ null,
24
+ null,
25
+ null,
26
+ null,
27
+ null,
28
+ null,
29
+ null,
30
+ null,
31
+ null,
32
+ null
33
+ ],
34
+ "intervention_types": [
35
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
36
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
37
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
38
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
39
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
40
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
41
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
42
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
43
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
44
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
45
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
46
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
47
+ "<class 'pyvene.models.interventions.AdditionIntervention'>",
48
+ "<class 'pyvene.models.interventions.AdditionIntervention'>"
49
+ ],
50
+ "mode": "parallel",
51
+ "representations": [
52
+ [
53
+ 0,
54
+ "model.layers[6].self_attn.o_proj.output",
55
+ "pos",
56
+ 1,
57
+ null,
58
+ null,
59
+ null,
60
+ null,
61
+ null,
62
+ null,
63
+ null,
64
+ null,
65
+ null
66
+ ],
67
+ [
68
+ 0,
69
+ "model.layers[9].self_attn.o_proj.output",
70
+ "pos",
71
+ 1,
72
+ null,
73
+ null,
74
+ null,
75
+ null,
76
+ null,
77
+ null,
78
+ null,
79
+ null,
80
+ null
81
+ ],
82
+ [
83
+ 0,
84
+ "model.layers[10].self_attn.o_proj.output",
85
+ "pos",
86
+ 1,
87
+ null,
88
+ null,
89
+ null,
90
+ null,
91
+ null,
92
+ null,
93
+ null,
94
+ null,
95
+ null
96
+ ],
97
+ [
98
+ 0,
99
+ "model.layers[11].self_attn.o_proj.output",
100
+ "pos",
101
+ 1,
102
+ null,
103
+ null,
104
+ null,
105
+ null,
106
+ null,
107
+ null,
108
+ null,
109
+ null,
110
+ null
111
+ ],
112
+ [
113
+ 0,
114
+ "model.layers[12].self_attn.o_proj.output",
115
+ "pos",
116
+ 1,
117
+ null,
118
+ null,
119
+ null,
120
+ null,
121
+ null,
122
+ null,
123
+ null,
124
+ null,
125
+ null
126
+ ],
127
+ [
128
+ 0,
129
+ "model.layers[13].self_attn.o_proj.output",
130
+ "pos",
131
+ 1,
132
+ null,
133
+ null,
134
+ null,
135
+ null,
136
+ null,
137
+ null,
138
+ null,
139
+ null,
140
+ null
141
+ ],
142
+ [
143
+ 0,
144
+ "model.layers[14].self_attn.o_proj.output",
145
+ "pos",
146
+ 1,
147
+ null,
148
+ null,
149
+ null,
150
+ null,
151
+ null,
152
+ null,
153
+ null,
154
+ null,
155
+ null
156
+ ],
157
+ [
158
+ 0,
159
+ "model.layers[15].self_attn.o_proj.output",
160
+ "pos",
161
+ 1,
162
+ null,
163
+ null,
164
+ null,
165
+ null,
166
+ null,
167
+ null,
168
+ null,
169
+ null,
170
+ null
171
+ ],
172
+ [
173
+ 0,
174
+ "model.layers[16].self_attn.o_proj.output",
175
+ "pos",
176
+ 1,
177
+ null,
178
+ null,
179
+ null,
180
+ null,
181
+ null,
182
+ null,
183
+ null,
184
+ null,
185
+ null
186
+ ],
187
+ [
188
+ 0,
189
+ "model.layers[21].self_attn.o_proj.output",
190
+ "pos",
191
+ 1,
192
+ null,
193
+ null,
194
+ null,
195
+ null,
196
+ null,
197
+ null,
198
+ null,
199
+ null,
200
+ null
201
+ ],
202
+ [
203
+ 0,
204
+ "model.layers[26].self_attn.o_proj.output",
205
+ "pos",
206
+ 1,
207
+ null,
208
+ null,
209
+ null,
210
+ null,
211
+ null,
212
+ null,
213
+ null,
214
+ null,
215
+ null
216
+ ],
217
+ [
218
+ 0,
219
+ "model.layers[27].self_attn.o_proj.output",
220
+ "pos",
221
+ 1,
222
+ null,
223
+ null,
224
+ null,
225
+ null,
226
+ null,
227
+ null,
228
+ null,
229
+ null,
230
+ null
231
+ ],
232
+ [
233
+ 0,
234
+ "model.layers[30].self_attn.o_proj.output",
235
+ "pos",
236
+ 1,
237
+ null,
238
+ null,
239
+ null,
240
+ null,
241
+ null,
242
+ null,
243
+ null,
244
+ null,
245
+ null
246
+ ],
247
+ [
248
+ 0,
249
+ "model.layers[31].self_attn.o_proj.output",
250
+ "pos",
251
+ 1,
252
+ null,
253
+ null,
254
+ null,
255
+ null,
256
+ null,
257
+ null,
258
+ null,
259
+ null,
260
+ null
261
+ ]
262
+ ],
263
+ "sorted_keys": [
264
+ "comp.model.layers[6].self_attn.o_proj.output.unit.pos.nunit.1#0",
265
+ "comp.model.layers[9].self_attn.o_proj.output.unit.pos.nunit.1#0",
266
+ "comp.model.layers[10].self_attn.o_proj.output.unit.pos.nunit.1#0",
267
+ "comp.model.layers[11].self_attn.o_proj.output.unit.pos.nunit.1#0",
268
+ "comp.model.layers[12].self_attn.o_proj.output.unit.pos.nunit.1#0",
269
+ "comp.model.layers[13].self_attn.o_proj.output.unit.pos.nunit.1#0",
270
+ "comp.model.layers[14].self_attn.o_proj.output.unit.pos.nunit.1#0",
271
+ "comp.model.layers[15].self_attn.o_proj.output.unit.pos.nunit.1#0",
272
+ "comp.model.layers[16].self_attn.o_proj.output.unit.pos.nunit.1#0",
273
+ "comp.model.layers[21].self_attn.o_proj.output.unit.pos.nunit.1#0",
274
+ "comp.model.layers[26].self_attn.o_proj.output.unit.pos.nunit.1#0",
275
+ "comp.model.layers[27].self_attn.o_proj.output.unit.pos.nunit.1#0",
276
+ "comp.model.layers[30].self_attn.o_proj.output.unit.pos.nunit.1#0",
277
+ "comp.model.layers[31].self_attn.o_proj.output.unit.pos.nunit.1#0"
278
+ ],
279
+ "transformers_version": "4.37.0"
280
+ }