Hugging Face's logo

anonym-repos
/
calcformer-selftrain-ipo-0.99-lora-32

Safetensors
Model card Files Community
anonym-repos commited on
Commit
19ceae9
•
1 Parent(s): fe8f2ea

Upload T5ForConditionalGeneration

Browse files
Files changed (3) hide show
  1. adapter_config.json +458 -0
  2. adapter_model.safetensors +3 -0
  3. generation_config.json +6 -0
adapter_config.json ADDED
@@ -0,0 +1,458 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "MU-NLPC/calcformer-instruct-flan-xl_step-128k",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layers_pattern": null,
10
+ "layers_to_transform": null,
11
+ "loftq_config": {},
12
+ "lora_alpha": 32,
13
+ "lora_dropout": 0.1,
14
+ "megatron_config": null,
15
+ "megatron_core": "megatron.core",
16
+ "modules_to_save": null,
17
+ "peft_type": "LORA",
18
+ "r": 32,
19
+ "rank_pattern": {},
20
+ "revision": null,
21
+ "target_modules": [
22
+ "encoder.block.6.layer.1.DenseReluDense.wi_0",
23
+ "decoder.block.12.layer.0.SelfAttention.o",
24
+ "decoder.block.4.layer.0.SelfAttention.k",
25
+ "decoder.block.9.layer.1.EncDecAttention.o",
26
+ "decoder.block.4.layer.0.SelfAttention.v",
27
+ "encoder.block.3.layer.0.SelfAttention.o",
28
+ "encoder.block.4.layer.1.DenseReluDense.wi_0",
29
+ "decoder.block.20.layer.1.EncDecAttention.k",
30
+ "encoder.block.7.layer.0.SelfAttention.v",
31
+ "decoder.block.15.layer.2.DenseReluDense.wi_1",
32
+ "encoder.block.15.layer.0.SelfAttention.o",
33
+ "decoder.block.17.layer.0.SelfAttention.v",
34
+ "encoder.block.12.layer.1.DenseReluDense.wi_1",
35
+ "encoder.block.3.layer.1.DenseReluDense.wi_1",
36
+ "encoder.block.22.layer.1.DenseReluDense.wi_1",
37
+ "decoder.block.9.layer.2.DenseReluDense.wo",
38
+ "decoder.block.12.layer.0.SelfAttention.q",
39
+ "encoder.block.2.layer.1.DenseReluDense.wi_0",
40
+ "encoder.block.17.layer.0.SelfAttention.k",
41
+ "encoder.block.22.layer.0.SelfAttention.v",
42
+ "decoder.block.22.layer.1.EncDecAttention.k",
43
+ "decoder.block.16.layer.0.SelfAttention.o",
44
+ "decoder.block.3.layer.1.EncDecAttention.k",
45
+ "decoder.block.10.layer.2.DenseReluDense.wi_1",
46
+ "decoder.block.15.layer.0.SelfAttention.q",
47
+ "decoder.block.19.layer.1.EncDecAttention.q",
48
+ "encoder.block.3.layer.0.SelfAttention.v",
49
+ "decoder.block.22.layer.1.EncDecAttention.q",
50
+ "decoder.block.22.layer.1.EncDecAttention.o",
51
+ "decoder.block.5.layer.0.SelfAttention.o",
52
+ "encoder.block.11.layer.1.DenseReluDense.wi_1",
53
+ "decoder.block.13.layer.2.DenseReluDense.wi_0",
54
+ "decoder.block.14.layer.0.SelfAttention.o",
55
+ "encoder.block.6.layer.1.DenseReluDense.wo",
56
+ "decoder.block.17.layer.2.DenseReluDense.wi_1",
57
+ "decoder.block.0.layer.1.EncDecAttention.q",
58
+ "decoder.block.6.layer.1.EncDecAttention.v",
59
+ "decoder.block.3.layer.0.SelfAttention.k",
60
+ "encoder.block.5.layer.0.SelfAttention.o",
61
+ "encoder.block.8.layer.0.SelfAttention.v",
62
+ "decoder.block.19.layer.2.DenseReluDense.wi_1",
63
+ "decoder.block.22.layer.0.SelfAttention.o",
64
+ "encoder.block.11.layer.0.SelfAttention.k",
65
+ "decoder.block.10.layer.2.DenseReluDense.wi_0",
66
+ "decoder.block.11.layer.0.SelfAttention.v",
67
+ "decoder.block.7.layer.1.EncDecAttention.k",
68
+ "decoder.block.21.layer.0.SelfAttention.k",
69
+ "decoder.block.3.layer.1.EncDecAttention.q",
70
+ "decoder.block.20.layer.1.EncDecAttention.v",
71
+ "encoder.block.9.layer.0.SelfAttention.o",
72
+ "decoder.block.11.layer.1.EncDecAttention.v",
73
+ "encoder.block.15.layer.0.SelfAttention.q",
74
+ "encoder.block.9.layer.1.DenseReluDense.wi_1",
75
+ "encoder.block.20.layer.0.SelfAttention.o",
76
+ "decoder.block.19.layer.1.EncDecAttention.o",
77
+ "decoder.block.0.layer.0.SelfAttention.v",
78
+ "decoder.block.19.layer.0.SelfAttention.o",
79
+ "decoder.block.23.layer.0.SelfAttention.q",
80
+ "encoder.block.16.layer.0.SelfAttention.q",
81
+ "decoder.block.6.layer.2.DenseReluDense.wo",
82
+ "decoder.block.20.layer.1.EncDecAttention.o",
83
+ "decoder.block.15.layer.0.SelfAttention.v",
84
+ "decoder.block.16.layer.1.EncDecAttention.v",
85
+ "decoder.block.0.layer.0.SelfAttention.q",
86
+ "decoder.block.21.layer.2.DenseReluDense.wi_1",
87
+ "encoder.block.3.layer.1.DenseReluDense.wi_0",
88
+ "encoder.block.5.layer.0.SelfAttention.q",
89
+ "decoder.block.15.layer.0.SelfAttention.o",
90
+ "encoder.block.18.layer.0.SelfAttention.o",
91
+ "decoder.block.21.layer.1.EncDecAttention.q",
92
+ "encoder.block.18.layer.0.SelfAttention.q",
93
+ "decoder.block.19.layer.0.SelfAttention.k",
94
+ "encoder.block.16.layer.1.DenseReluDense.wo",
95
+ "decoder.block.9.layer.1.EncDecAttention.k",
96
+ "decoder.block.6.layer.0.SelfAttention.q",
97
+ "decoder.block.0.layer.1.EncDecAttention.o",
98
+ "decoder.block.17.layer.2.DenseReluDense.wi_0",
99
+ "decoder.block.16.layer.0.SelfAttention.v",
100
+ "decoder.block.17.layer.0.SelfAttention.o",
101
+ "decoder.block.20.layer.0.SelfAttention.k",
102
+ "encoder.block.23.layer.1.DenseReluDense.wi_0",
103
+ "encoder.block.2.layer.1.DenseReluDense.wo",
104
+ "decoder.block.1.layer.1.EncDecAttention.o",
105
+ "decoder.block.15.layer.1.EncDecAttention.q",
106
+ "decoder.block.2.layer.2.DenseReluDense.wi_1",
107
+ "decoder.block.22.layer.2.DenseReluDense.wi_0",
108
+ "encoder.block.16.layer.0.SelfAttention.k",
109
+ "decoder.block.16.layer.2.DenseReluDense.wi_1",
110
+ "decoder.block.23.layer.1.EncDecAttention.q",
111
+ "decoder.block.1.layer.2.DenseReluDense.wo",
112
+ "decoder.block.23.layer.1.EncDecAttention.o",
113
+ "decoder.block.1.layer.1.EncDecAttention.q",
114
+ "decoder.block.17.layer.2.DenseReluDense.wo",
115
+ "decoder.block.13.layer.0.SelfAttention.q",
116
+ "decoder.block.3.layer.2.DenseReluDense.wi_1",
117
+ "decoder.block.16.layer.2.DenseReluDense.wi_0",
118
+ "decoder.block.21.layer.0.SelfAttention.q",
119
+ "decoder.block.18.layer.0.SelfAttention.o",
120
+ "decoder.block.3.layer.1.EncDecAttention.o",
121
+ "decoder.block.2.layer.1.EncDecAttention.o",
122
+ "encoder.block.6.layer.0.SelfAttention.v",
123
+ "decoder.block.20.layer.2.DenseReluDense.wi_0",
124
+ "decoder.block.3.layer.1.EncDecAttention.v",
125
+ "encoder.block.2.layer.1.DenseReluDense.wi_1",
126
+ "decoder.block.11.layer.0.SelfAttention.o",
127
+ "decoder.block.22.layer.2.DenseReluDense.wi_1",
128
+ "decoder.block.22.layer.1.EncDecAttention.v",
129
+ "encoder.block.22.layer.1.DenseReluDense.wi_0",
130
+ "decoder.block.18.layer.2.DenseReluDense.wi_1",
131
+ "encoder.block.19.layer.0.SelfAttention.q",
132
+ "encoder.block.21.layer.0.SelfAttention.k",
133
+ "decoder.block.10.layer.0.SelfAttention.q",
134
+ "encoder.block.11.layer.0.SelfAttention.o",
135
+ "encoder.block.17.layer.1.DenseReluDense.wi_0",
136
+ "decoder.block.18.layer.1.EncDecAttention.k",
137
+ "decoder.block.4.layer.1.EncDecAttention.q",
138
+ "decoder.block.7.layer.1.EncDecAttention.q",
139
+ "decoder.block.19.layer.0.SelfAttention.q",
140
+ "decoder.block.12.layer.1.EncDecAttention.v",
141
+ "decoder.block.3.layer.0.SelfAttention.v",
142
+ "decoder.block.8.layer.0.SelfAttention.k",
143
+ "encoder.block.2.layer.0.SelfAttention.k",
144
+ "decoder.block.10.layer.0.SelfAttention.v",
145
+ "decoder.block.5.layer.2.DenseReluDense.wo",
146
+ "decoder.block.14.layer.1.EncDecAttention.q",
147
+ "encoder.block.6.layer.0.SelfAttention.k",
148
+ "encoder.block.12.layer.1.DenseReluDense.wo",
149
+ "decoder.block.11.layer.2.DenseReluDense.wi_0",
150
+ "decoder.block.15.layer.1.EncDecAttention.v",
151
+ "decoder.block.14.layer.0.SelfAttention.v",
152
+ "decoder.block.23.layer.2.DenseReluDense.wo",
153
+ "decoder.block.1.layer.2.DenseReluDense.wi_1",
154
+ "decoder.block.6.layer.0.SelfAttention.o",
155
+ "encoder.block.0.layer.1.DenseReluDense.wo",
156
+ "decoder.block.18.layer.1.EncDecAttention.q",
157
+ "decoder.block.17.layer.1.EncDecAttention.v",
158
+ "decoder.block.10.layer.1.EncDecAttention.o",
159
+ "decoder.block.11.layer.1.EncDecAttention.o",
160
+ "decoder.block.19.layer.0.SelfAttention.v",
161
+ "decoder.block.13.layer.1.EncDecAttention.q",
162
+ "decoder.block.16.layer.0.SelfAttention.k",
163
+ "decoder.block.9.layer.1.EncDecAttention.q",
164
+ "encoder.block.9.layer.0.SelfAttention.k",
165
+ "decoder.block.5.layer.0.SelfAttention.v",
166
+ "decoder.block.4.layer.0.SelfAttention.o",
167
+ "decoder.block.10.layer.0.SelfAttention.k",
168
+ "encoder.block.3.layer.0.SelfAttention.k",
169
+ "encoder.block.13.layer.0.SelfAttention.k",
170
+ "decoder.block.14.layer.0.SelfAttention.q",
171
+ "encoder.block.17.layer.1.DenseReluDense.wi_1",
172
+ "decoder.block.0.layer.2.DenseReluDense.wo",
173
+ "encoder.block.16.layer.0.SelfAttention.o",
174
+ "decoder.block.0.layer.2.DenseReluDense.wi_0",
175
+ "encoder.block.17.layer.0.SelfAttention.v",
176
+ "encoder.block.19.layer.0.SelfAttention.o",
177
+ "decoder.block.8.layer.2.DenseReluDense.wo",
178
+ "decoder.block.22.layer.0.SelfAttention.k",
179
+ "encoder.block.13.layer.1.DenseReluDense.wi_1",
180
+ "encoder.block.16.layer.1.DenseReluDense.wi_1",
181
+ "decoder.block.9.layer.2.DenseReluDense.wi_1",
182
+ "decoder.block.2.layer.0.SelfAttention.k",
183
+ "encoder.block.17.layer.0.SelfAttention.q",
184
+ "decoder.block.16.layer.2.DenseReluDense.wo",
185
+ "decoder.block.6.layer.2.DenseReluDense.wi_0",
186
+ "decoder.block.4.layer.2.DenseReluDense.wi_0",
187
+ "decoder.block.11.layer.1.EncDecAttention.k",
188
+ "decoder.block.2.layer.1.EncDecAttention.q",
189
+ "encoder.block.5.layer.1.DenseReluDense.wo",
190
+ "decoder.block.2.layer.2.DenseReluDense.wo",
191
+ "encoder.block.10.layer.1.DenseReluDense.wi_1",
192
+ "decoder.block.18.layer.2.DenseReluDense.wi_0",
193
+ "encoder.block.10.layer.0.SelfAttention.k",
194
+ "encoder.block.21.layer.1.DenseReluDense.wo",
195
+ "decoder.block.8.layer.0.SelfAttention.o",
196
+ "decoder.block.10.layer.2.DenseReluDense.wo",
197
+ "encoder.block.11.layer.0.SelfAttention.v",
198
+ "decoder.block.5.layer.0.SelfAttention.q",
199
+ "decoder.block.19.layer.2.DenseReluDense.wi_0",
200
+ "decoder.block.7.layer.0.SelfAttention.k",
201
+ "decoder.block.8.layer.1.EncDecAttention.o",
202
+ "encoder.block.12.layer.0.SelfAttention.o",
203
+ "encoder.block.0.layer.1.DenseReluDense.wi_1",
204
+ "encoder.block.3.layer.0.SelfAttention.q",
205
+ "decoder.block.1.layer.0.SelfAttention.v",
206
+ "encoder.block.18.layer.0.SelfAttention.v",
207
+ "encoder.block.19.layer.1.DenseReluDense.wo",
208
+ "decoder.block.23.layer.1.EncDecAttention.v",
209
+ "encoder.block.14.layer.0.SelfAttention.k",
210
+ "decoder.block.12.layer.2.DenseReluDense.wi_0",
211
+ "decoder.block.20.layer.0.SelfAttention.v",
212
+ "decoder.block.14.layer.1.EncDecAttention.o",
213
+ "decoder.block.11.layer.0.SelfAttention.q",
214
+ "encoder.block.15.layer.0.SelfAttention.k",
215
+ "decoder.block.23.layer.2.DenseReluDense.wi_0",
216
+ "decoder.block.2.layer.0.SelfAttention.o",
217
+ "decoder.block.0.layer.2.DenseReluDense.wi_1",
218
+ "encoder.block.14.layer.0.SelfAttention.q",
219
+ "encoder.block.12.layer.0.SelfAttention.v",
220
+ "encoder.block.20.layer.1.DenseReluDense.wi_0",
221
+ "encoder.block.19.layer.1.DenseReluDense.wi_1",
222
+ "decoder.block.6.layer.2.DenseReluDense.wi_1",
223
+ "decoder.block.7.layer.0.SelfAttention.v",
224
+ "encoder.block.7.layer.1.DenseReluDense.wo",
225
+ "decoder.block.4.layer.1.EncDecAttention.k",
226
+ "decoder.block.1.layer.1.EncDecAttention.k",
227
+ "decoder.block.17.layer.0.SelfAttention.q",
228
+ "encoder.block.9.layer.0.SelfAttention.v",
229
+ "encoder.block.19.layer.0.SelfAttention.k",
230
+ "encoder.block.1.layer.0.SelfAttention.k",
231
+ "decoder.block.16.layer.1.EncDecAttention.q",
232
+ "decoder.block.18.layer.1.EncDecAttention.o",
233
+ "encoder.block.20.layer.0.SelfAttention.q",
234
+ "decoder.block.14.layer.2.DenseReluDense.wi_0",
235
+ "encoder.block.23.layer.0.SelfAttention.k",
236
+ "decoder.block.21.layer.1.EncDecAttention.k",
237
+ "encoder.block.22.layer.1.DenseReluDense.wo",
238
+ "decoder.block.13.layer.1.EncDecAttention.k",
239
+ "decoder.block.10.layer.1.EncDecAttention.k",
240
+ "encoder.block.20.layer.1.DenseReluDense.wo",
241
+ "decoder.block.2.layer.1.EncDecAttention.v",
242
+ "encoder.block.23.layer.0.SelfAttention.v",
243
+ "decoder.block.14.layer.2.DenseReluDense.wo",
244
+ "encoder.block.5.layer.0.SelfAttention.v",
245
+ "decoder.block.10.layer.1.EncDecAttention.v",
246
+ "encoder.block.14.layer.1.DenseReluDense.wi_0",
247
+ "encoder.block.13.layer.1.DenseReluDense.wo",
248
+ "encoder.block.8.layer.1.DenseReluDense.wi_0",
249
+ "encoder.block.3.layer.1.DenseReluDense.wo",
250
+ "encoder.block.5.layer.1.DenseReluDense.wi_1",
251
+ "decoder.block.6.layer.1.EncDecAttention.q",
252
+ "encoder.block.14.layer.0.SelfAttention.o",
253
+ "decoder.block.23.layer.0.SelfAttention.v",
254
+ "encoder.block.18.layer.1.DenseReluDense.wi_1",
255
+ "decoder.block.7.layer.2.DenseReluDense.wi_0",
256
+ "decoder.block.17.layer.1.EncDecAttention.q",
257
+ "decoder.block.20.layer.0.SelfAttention.q",
258
+ "encoder.block.12.layer.0.SelfAttention.q",
259
+ "encoder.block.8.layer.0.SelfAttention.q",
260
+ "encoder.block.6.layer.1.DenseReluDense.wi_1",
261
+ "decoder.block.4.layer.2.DenseReluDense.wi_1",
262
+ "encoder.block.18.layer.0.SelfAttention.k",
263
+ "decoder.block.15.layer.1.EncDecAttention.k",
264
+ "decoder.block.14.layer.0.SelfAttention.k",
265
+ "decoder.block.2.layer.0.SelfAttention.q",
266
+ "decoder.block.15.layer.2.DenseReluDense.wo",
267
+ "encoder.block.14.layer.0.SelfAttention.v",
268
+ "decoder.block.13.layer.1.EncDecAttention.v",
269
+ "encoder.block.20.layer.0.SelfAttention.v",
270
+ "decoder.block.12.layer.0.SelfAttention.k",
271
+ "encoder.block.19.layer.0.SelfAttention.v",
272
+ "decoder.block.11.layer.2.DenseReluDense.wo",
273
+ "encoder.block.1.layer.1.DenseReluDense.wi_0",
274
+ "encoder.block.16.layer.0.SelfAttention.v",
275
+ "encoder.block.1.layer.0.SelfAttention.v",
276
+ "decoder.block.19.layer.1.EncDecAttention.v",
277
+ "decoder.block.3.layer.0.SelfAttention.q",
278
+ "decoder.block.7.layer.2.DenseReluDense.wo",
279
+ "decoder.block.22.layer.0.SelfAttention.q",
280
+ "decoder.block.8.layer.2.DenseReluDense.wi_1",
281
+ "encoder.block.1.layer.1.DenseReluDense.wo",
282
+ "decoder.block.3.layer.0.SelfAttention.o",
283
+ "encoder.block.13.layer.1.DenseReluDense.wi_0",
284
+ "decoder.block.9.layer.0.SelfAttention.k",
285
+ "decoder.block.7.layer.0.SelfAttention.o",
286
+ "decoder.block.2.layer.0.SelfAttention.v",
287
+ "decoder.block.13.layer.0.SelfAttention.v",
288
+ "encoder.block.22.layer.0.SelfAttention.q",
289
+ "decoder.block.9.layer.1.EncDecAttention.v",
290
+ "encoder.block.13.layer.0.SelfAttention.v",
291
+ "decoder.block.13.layer.2.DenseReluDense.wo",
292
+ "encoder.block.11.layer.1.DenseReluDense.wi_0",
293
+ "encoder.block.7.layer.0.SelfAttention.k",
294
+ "decoder.block.5.layer.1.EncDecAttention.o",
295
+ "decoder.block.9.layer.2.DenseReluDense.wi_0",
296
+ "encoder.block.13.layer.0.SelfAttention.o",
297
+ "decoder.block.18.layer.0.SelfAttention.v",
298
+ "encoder.block.18.layer.1.DenseReluDense.wi_0",
299
+ "encoder.block.11.layer.0.SelfAttention.q",
300
+ "encoder.block.2.layer.0.SelfAttention.o",
301
+ "decoder.block.8.layer.1.EncDecAttention.v",
302
+ "encoder.block.0.layer.1.DenseReluDense.wi_0",
303
+ "decoder.block.12.layer.0.SelfAttention.v",
304
+ "decoder.block.9.layer.0.SelfAttention.o",
305
+ "encoder.block.1.layer.1.DenseReluDense.wi_1",
306
+ "decoder.block.21.layer.1.EncDecAttention.v",
307
+ "decoder.block.20.layer.2.DenseReluDense.wo",
308
+ "encoder.block.5.layer.1.DenseReluDense.wi_0",
309
+ "encoder.block.18.layer.1.DenseReluDense.wo",
310
+ "encoder.block.8.layer.0.SelfAttention.o",
311
+ "decoder.block.17.layer.1.EncDecAttention.k",
312
+ "encoder.block.1.layer.0.SelfAttention.q",
313
+ "encoder.block.7.layer.1.DenseReluDense.wi_1",
314
+ "decoder.block.23.layer.1.EncDecAttention.k",
315
+ "decoder.block.16.layer.0.SelfAttention.q",
316
+ "decoder.block.7.layer.2.DenseReluDense.wi_1",
317
+ "encoder.block.9.layer.1.DenseReluDense.wo",
318
+ "encoder.block.23.layer.1.DenseReluDense.wi_1",
319
+ "decoder.block.0.layer.0.SelfAttention.k",
320
+ "decoder.block.7.layer.0.SelfAttention.q",
321
+ "encoder.block.11.layer.1.DenseReluDense.wo",
322
+ "decoder.block.13.layer.0.SelfAttention.o",
323
+ "decoder.block.9.layer.0.SelfAttention.q",
324
+ "encoder.block.9.layer.1.DenseReluDense.wi_0",
325
+ "decoder.block.22.layer.2.DenseReluDense.wo",
326
+ "encoder.block.1.layer.0.SelfAttention.o",
327
+ "encoder.block.0.layer.0.SelfAttention.k",
328
+ "encoder.block.5.layer.0.SelfAttention.k",
329
+ "decoder.block.12.layer.2.DenseReluDense.wi_1",
330
+ "encoder.block.21.layer.0.SelfAttention.v",
331
+ "decoder.block.21.layer.2.DenseReluDense.wo",
332
+ "decoder.block.17.layer.0.SelfAttention.k",
333
+ "decoder.block.0.layer.1.EncDecAttention.k",
334
+ "encoder.block.0.layer.0.SelfAttention.q",
335
+ "encoder.block.9.layer.0.SelfAttention.q",
336
+ "encoder.block.21.layer.0.SelfAttention.q",
337
+ "decoder.block.2.layer.2.DenseReluDense.wi_0",
338
+ "encoder.block.4.layer.0.SelfAttention.k",
339
+ "encoder.block.10.layer.0.SelfAttention.o",
340
+ "encoder.block.14.layer.1.DenseReluDense.wi_1",
341
+ "decoder.block.8.layer.1.EncDecAttention.q",
342
+ "decoder.block.6.layer.1.EncDecAttention.k",
343
+ "decoder.block.20.layer.2.DenseReluDense.wi_1",
344
+ "encoder.block.19.layer.1.DenseReluDense.wi_0",
345
+ "decoder.block.6.layer.0.SelfAttention.v",
346
+ "decoder.block.8.layer.2.DenseReluDense.wi_0",
347
+ "decoder.block.14.layer.2.DenseReluDense.wi_1",
348
+ "encoder.block.15.layer.1.DenseReluDense.wo",
349
+ "decoder.block.20.layer.1.EncDecAttention.q",
350
+ "decoder.block.11.layer.2.DenseReluDense.wi_1",
351
+ "encoder.block.4.layer.1.DenseReluDense.wo",
352
+ "encoder.block.10.layer.0.SelfAttention.v",
353
+ "decoder.block.18.layer.2.DenseReluDense.wo",
354
+ "encoder.block.8.layer.1.DenseReluDense.wi_1",
355
+ "encoder.block.10.layer.0.SelfAttention.q",
356
+ "encoder.block.2.layer.0.SelfAttention.v",
357
+ "encoder.block.0.layer.0.SelfAttention.v",
358
+ "decoder.block.4.layer.2.DenseReluDense.wo",
359
+ "encoder.block.17.layer.1.DenseReluDense.wo",
360
+ "decoder.block.7.layer.1.EncDecAttention.v",
361
+ "encoder.block.15.layer.1.DenseReluDense.wi_1",
362
+ "encoder.block.10.layer.1.DenseReluDense.wo",
363
+ "decoder.block.12.layer.1.EncDecAttention.k",
364
+ "encoder.block.4.layer.1.DenseReluDense.wi_1",
365
+ "decoder.block.8.layer.0.SelfAttention.q",
366
+ "decoder.block.20.layer.0.SelfAttention.o",
367
+ "decoder.block.22.layer.0.SelfAttention.v",
368
+ "decoder.block.1.layer.0.SelfAttention.k",
369
+ "encoder.block.15.layer.1.DenseReluDense.wi_0",
370
+ "decoder.block.4.layer.0.SelfAttention.q",
371
+ "encoder.block.2.layer.0.SelfAttention.q",
372
+ "decoder.block.10.layer.1.EncDecAttention.q",
373
+ "decoder.block.6.layer.1.EncDecAttention.o",
374
+ "decoder.block.11.layer.1.EncDecAttention.q",
375
+ "decoder.block.23.layer.0.SelfAttention.o",
376
+ "encoder.block.7.layer.0.SelfAttention.q",
377
+ "decoder.block.12.layer.1.EncDecAttention.o",
378
+ "encoder.block.12.layer.0.SelfAttention.k",
379
+ "encoder.block.22.layer.0.SelfAttention.o",
380
+ "encoder.block.21.layer.1.DenseReluDense.wi_1",
381
+ "decoder.block.8.layer.1.EncDecAttention.k",
382
+ "decoder.block.5.layer.0.SelfAttention.k",
383
+ "decoder.block.18.layer.0.SelfAttention.q",
384
+ "decoder.block.10.layer.0.SelfAttention.o",
385
+ "decoder.block.15.layer.0.SelfAttention.k",
386
+ "decoder.block.23.layer.0.SelfAttention.k",
387
+ "encoder.block.4.layer.0.SelfAttention.v",
388
+ "decoder.block.9.layer.0.SelfAttention.v",
389
+ "decoder.block.3.layer.2.DenseReluDense.wi_0",
390
+ "decoder.block.11.layer.0.SelfAttention.k",
391
+ "decoder.block.21.layer.0.SelfAttention.o",
392
+ "encoder.block.22.layer.0.SelfAttention.k",
393
+ "decoder.block.15.layer.1.EncDecAttention.o",
394
+ "decoder.block.4.layer.1.EncDecAttention.o",
395
+ "encoder.block.20.layer.0.SelfAttention.k",
396
+ "decoder.block.14.layer.1.EncDecAttention.v",
397
+ "decoder.block.0.layer.0.SelfAttention.o",
398
+ "encoder.block.21.layer.1.DenseReluDense.wi_0",
399
+ "encoder.block.8.layer.0.SelfAttention.k",
400
+ "decoder.block.16.layer.1.EncDecAttention.k",
401
+ "decoder.block.21.layer.0.SelfAttention.v",
402
+ "decoder.block.1.layer.0.SelfAttention.o",
403
+ "encoder.block.14.layer.1.DenseReluDense.wo",
404
+ "decoder.block.2.layer.1.EncDecAttention.k",
405
+ "decoder.block.18.layer.1.EncDecAttention.v",
406
+ "decoder.block.5.layer.2.DenseReluDense.wi_1",
407
+ "decoder.block.13.layer.1.EncDecAttention.o",
408
+ "encoder.block.6.layer.0.SelfAttention.q",
409
+ "decoder.block.23.layer.2.DenseReluDense.wi_1",
410
+ "encoder.block.23.layer.0.SelfAttention.q",
411
+ "encoder.block.12.layer.1.DenseReluDense.wi_0",
412
+ "decoder.block.5.layer.1.EncDecAttention.k",
413
+ "decoder.block.13.layer.2.DenseReluDense.wi_1",
414
+ "encoder.block.7.layer.0.SelfAttention.o",
415
+ "decoder.block.3.layer.2.DenseReluDense.wo",
416
+ "decoder.block.13.layer.0.SelfAttention.k",
417
+ "decoder.block.1.layer.1.EncDecAttention.v",
418
+ "encoder.block.20.layer.1.DenseReluDense.wi_1",
419
+ "lm_head",
420
+ "decoder.block.1.layer.2.DenseReluDense.wi_0",
421
+ "decoder.block.14.layer.1.EncDecAttention.k",
422
+ "encoder.block.21.layer.0.SelfAttention.o",
423
+ "encoder.block.4.layer.0.SelfAttention.q",
424
+ "decoder.block.5.layer.1.EncDecAttention.v",
425
+ "decoder.block.21.layer.2.DenseReluDense.wi_0",
426
+ "decoder.block.5.layer.2.DenseReluDense.wi_0",
427
+ "decoder.block.16.layer.1.EncDecAttention.o",
428
+ "decoder.block.18.layer.0.SelfAttention.k",
429
+ "decoder.block.5.layer.1.EncDecAttention.q",
430
+ "encoder.block.13.layer.0.SelfAttention.q",
431
+ "decoder.block.12.layer.1.EncDecAttention.q",
432
+ "decoder.block.19.layer.2.DenseReluDense.wo",
433
+ "encoder.block.4.layer.0.SelfAttention.o",
434
+ "decoder.block.8.layer.0.SelfAttention.v",
435
+ "encoder.block.10.layer.1.DenseReluDense.wi_0",
436
+ "decoder.block.0.layer.1.EncDecAttention.v",
437
+ "decoder.block.19.layer.1.EncDecAttention.k",
438
+ "encoder.block.23.layer.0.SelfAttention.o",
439
+ "decoder.block.6.layer.0.SelfAttention.k",
440
+ "encoder.block.15.layer.0.SelfAttention.v",
441
+ "decoder.block.7.layer.1.EncDecAttention.o",
442
+ "encoder.block.8.layer.1.DenseReluDense.wo",
443
+ "decoder.block.15.layer.2.DenseReluDense.wi_0",
444
+ "encoder.block.16.layer.1.DenseReluDense.wi_0",
445
+ "decoder.block.21.layer.1.EncDecAttention.o",
446
+ "encoder.block.7.layer.1.DenseReluDense.wi_0",
447
+ "decoder.block.12.layer.2.DenseReluDense.wo",
448
+ "decoder.block.4.layer.1.EncDecAttention.v",
449
+ "decoder.block.17.layer.1.EncDecAttention.o",
450
+ "encoder.block.0.layer.0.SelfAttention.o",
451
+ "decoder.block.1.layer.0.SelfAttention.q",
452
+ "encoder.block.23.layer.1.DenseReluDense.wo",
453
+ "encoder.block.17.layer.0.SelfAttention.o",
454
+ "encoder.block.6.layer.0.SelfAttention.o"
455
+ ],
456
+ "task_type": "SEQ_2_SEQ_LM",
457
+ "use_rslora": false
458
+ }
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa692765069ef762ec1dbfc30513891e2eae102a632429fb63b3ac461d813587
3
+ size 550808560
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "decoder_start_token_id": 0,
3
+ "eos_token_id": 1,
4
+ "pad_token_id": 0,
5
+ "transformers_version": "4.36.2"
6
+ }