hidude561 commited on
Commit
359f4c4
·
verified ·
1 Parent(s): ad35b38

Remove edit-LoRA files (moved to justice-edit); restore v86-only

Browse files
Files changed (1) hide show
  1. adapter_config.json +0 -335
adapter_config.json DELETED
@@ -1,335 +0,0 @@
1
- {
2
- "alora_invocation_tokens": null,
3
- "alpha_pattern": {},
4
- "arrow_config": null,
5
- "auto_mapping": null,
6
- "base_model_name_or_path": "google/gemma-4-E4B-it",
7
- "bias": "none",
8
- "corda_config": null,
9
- "ensure_weight_tying": false,
10
- "eva_config": null,
11
- "exclude_modules": null,
12
- "fan_in_fan_out": false,
13
- "inference_mode": true,
14
- "init_lora_weights": true,
15
- "layer_replication": null,
16
- "layers_pattern": null,
17
- "layers_to_transform": null,
18
- "loftq_config": {},
19
- "lora_alpha": 32,
20
- "lora_bias": false,
21
- "lora_dropout": 0.05,
22
- "lora_ga_config": null,
23
- "megatron_config": null,
24
- "megatron_core": "megatron.core",
25
- "modules_to_save": null,
26
- "peft_type": "LORA",
27
- "peft_version": "0.19.1",
28
- "qalora_group_size": 16,
29
- "r": 16,
30
- "rank_pattern": {},
31
- "revision": null,
32
- "target_modules": [
33
- "model.language_model.layers.11.self_attn.v_proj",
34
- "model.language_model.layers.35.self_attn.o_proj",
35
- "model.audio_tower.layers.5.self_attn.k_proj.linear",
36
- "model.language_model.layers.9.self_attn.v_proj",
37
- "model.language_model.layers.39.mlp.gate_proj",
38
- "model.language_model.layers.7.self_attn.o_proj",
39
- "model.language_model.layers.13.self_attn.v_proj",
40
- "model.language_model.layers.3.self_attn.v_proj",
41
- "model.language_model.layers.17.mlp.down_proj",
42
- "model.language_model.layers.3.mlp.gate_proj",
43
- "model.language_model.layers.20.mlp.up_proj",
44
- "model.language_model.layers.25.self_attn.q_proj",
45
- "model.language_model.layers.11.self_attn.o_proj",
46
- "model.language_model.layers.28.self_attn.o_proj",
47
- "model.language_model.layers.4.mlp.gate_proj",
48
- "model.language_model.layers.3.self_attn.k_proj",
49
- "model.language_model.layers.1.self_attn.o_proj",
50
- "model.language_model.layers.23.self_attn.q_proj",
51
- "model.language_model.layers.0.self_attn.k_proj",
52
- "model.language_model.layers.1.self_attn.v_proj",
53
- "model.language_model.layers.2.self_attn.q_proj",
54
- "model.language_model.layers.31.self_attn.q_proj",
55
- "model.audio_tower.layers.0.self_attn.v_proj.linear",
56
- "model.language_model.layers.36.self_attn.q_proj",
57
- "model.audio_tower.layers.6.self_attn.k_proj.linear",
58
- "model.language_model.layers.22.mlp.gate_proj",
59
- "model.language_model.layers.22.mlp.down_proj",
60
- "model.audio_tower.layers.9.self_attn.v_proj.linear",
61
- "model.language_model.layers.5.self_attn.k_proj",
62
- "model.language_model.layers.15.self_attn.v_proj",
63
- "model.language_model.layers.17.mlp.up_proj",
64
- "model.language_model.layers.20.mlp.down_proj",
65
- "model.language_model.layers.40.self_attn.o_proj",
66
- "model.language_model.layers.20.self_attn.k_proj",
67
- "model.language_model.layers.34.self_attn.q_proj",
68
- "model.audio_tower.layers.8.self_attn.k_proj.linear",
69
- "model.audio_tower.layers.11.self_attn.q_proj.linear",
70
- "model.language_model.layers.14.mlp.down_proj",
71
- "model.language_model.layers.5.mlp.down_proj",
72
- "model.language_model.layers.40.self_attn.q_proj",
73
- "model.language_model.layers.18.self_attn.k_proj",
74
- "model.audio_tower.layers.3.self_attn.v_proj.linear",
75
- "model.language_model.layers.34.mlp.down_proj",
76
- "model.language_model.layers.5.mlp.gate_proj",
77
- "model.language_model.layers.37.mlp.up_proj",
78
- "model.audio_tower.layers.4.self_attn.v_proj.linear",
79
- "model.language_model.layers.24.self_attn.q_proj",
80
- "model.language_model.layers.20.self_attn.q_proj",
81
- "model.audio_tower.layers.2.self_attn.k_proj.linear",
82
- "model.language_model.layers.0.mlp.down_proj",
83
- "model.language_model.layers.18.mlp.up_proj",
84
- "model.language_model.layers.6.self_attn.k_proj",
85
- "model.audio_tower.layers.10.self_attn.v_proj.linear",
86
- "model.language_model.layers.2.mlp.up_proj",
87
- "model.language_model.layers.8.mlp.gate_proj",
88
- "model.audio_tower.layers.7.self_attn.q_proj.linear",
89
- "model.language_model.layers.14.self_attn.k_proj",
90
- "model.language_model.layers.10.self_attn.o_proj",
91
- "model.language_model.layers.22.self_attn.k_proj",
92
- "model.language_model.layers.30.self_attn.o_proj",
93
- "model.language_model.layers.32.mlp.gate_proj",
94
- "model.language_model.layers.8.self_attn.q_proj",
95
- "model.language_model.layers.37.mlp.down_proj",
96
- "model.language_model.layers.17.self_attn.k_proj",
97
- "model.language_model.layers.16.mlp.down_proj",
98
- "model.language_model.layers.0.self_attn.v_proj",
99
- "model.language_model.layers.20.self_attn.o_proj",
100
- "model.audio_tower.layers.1.self_attn.v_proj.linear",
101
- "model.language_model.layers.17.mlp.gate_proj",
102
- "model.language_model.layers.26.mlp.up_proj",
103
- "model.language_model.layers.33.mlp.gate_proj",
104
- "model.language_model.layers.6.self_attn.q_proj",
105
- "model.language_model.layers.2.self_attn.k_proj",
106
- "model.language_model.layers.6.mlp.down_proj",
107
- "model.language_model.layers.28.mlp.up_proj",
108
- "model.audio_tower.layers.5.self_attn.q_proj.linear",
109
- "model.language_model.layers.24.mlp.up_proj",
110
- "model.language_model.layers.32.mlp.down_proj",
111
- "model.language_model.layers.1.mlp.down_proj",
112
- "model.audio_tower.layers.2.self_attn.v_proj.linear",
113
- "model.audio_tower.layers.11.self_attn.v_proj.linear",
114
- "model.language_model.layers.0.mlp.up_proj",
115
- "model.language_model.layers.1.self_attn.k_proj",
116
- "model.language_model.layers.21.self_attn.v_proj",
117
- "model.language_model.layers.30.self_attn.q_proj",
118
- "model.language_model.layers.11.self_attn.q_proj",
119
- "model.language_model.layers.6.mlp.gate_proj",
120
- "model.language_model.layers.40.mlp.gate_proj",
121
- "model.language_model.layers.7.self_attn.q_proj",
122
- "model.language_model.layers.14.self_attn.q_proj",
123
- "model.language_model.layers.24.mlp.down_proj",
124
- "model.language_model.layers.24.mlp.gate_proj",
125
- "model.language_model.layers.38.mlp.gate_proj",
126
- "model.language_model.layers.5.self_attn.q_proj",
127
- "model.language_model.layers.9.self_attn.o_proj",
128
- "model.language_model.layers.12.mlp.gate_proj",
129
- "model.language_model.layers.15.mlp.up_proj",
130
- "model.language_model.layers.27.mlp.gate_proj",
131
- "model.language_model.layers.33.mlp.up_proj",
132
- "model.language_model.layers.31.self_attn.o_proj",
133
- "model.language_model.layers.11.mlp.down_proj",
134
- "model.language_model.layers.9.self_attn.q_proj",
135
- "model.language_model.layers.17.self_attn.q_proj",
136
- "model.language_model.layers.34.mlp.up_proj",
137
- "model.language_model.layers.5.self_attn.v_proj",
138
- "model.language_model.layers.24.self_attn.o_proj",
139
- "model.language_model.layers.13.self_attn.k_proj",
140
- "model.language_model.layers.35.mlp.gate_proj",
141
- "model.language_model.layers.15.mlp.gate_proj",
142
- "model.audio_tower.layers.10.self_attn.k_proj.linear",
143
- "model.language_model.layers.19.self_attn.k_proj",
144
- "model.language_model.layers.32.self_attn.o_proj",
145
- "model.audio_tower.layers.6.self_attn.v_proj.linear",
146
- "model.language_model.layers.6.mlp.up_proj",
147
- "model.language_model.layers.10.mlp.up_proj",
148
- "model.language_model.layers.41.self_attn.q_proj",
149
- "model.language_model.layers.28.self_attn.q_proj",
150
- "model.language_model.layers.13.self_attn.o_proj",
151
- "model.audio_tower.layers.5.self_attn.v_proj.linear",
152
- "model.language_model.layers.16.self_attn.q_proj",
153
- "model.language_model.layers.17.self_attn.o_proj",
154
- "model.audio_tower.layers.10.self_attn.q_proj.linear",
155
- "model.language_model.layers.1.self_attn.q_proj",
156
- "model.language_model.layers.8.self_attn.o_proj",
157
- "model.language_model.layers.12.mlp.down_proj",
158
- "model.language_model.layers.8.mlp.down_proj",
159
- "model.language_model.layers.10.self_attn.k_proj",
160
- "model.language_model.layers.38.mlp.down_proj",
161
- "model.language_model.layers.10.mlp.gate_proj",
162
- "model.language_model.layers.16.self_attn.o_proj",
163
- "model.language_model.layers.18.self_attn.v_proj",
164
- "model.language_model.layers.4.self_attn.v_proj",
165
- "model.language_model.layers.15.mlp.down_proj",
166
- "model.language_model.layers.4.self_attn.o_proj",
167
- "model.language_model.layers.17.self_attn.v_proj",
168
- "model.language_model.layers.21.self_attn.o_proj",
169
- "model.language_model.layers.20.self_attn.v_proj",
170
- "model.language_model.layers.21.self_attn.k_proj",
171
- "model.language_model.layers.23.mlp.gate_proj",
172
- "model.language_model.layers.27.mlp.down_proj",
173
- "model.language_model.layers.3.self_attn.o_proj",
174
- "model.language_model.layers.29.mlp.gate_proj",
175
- "model.language_model.layers.30.mlp.down_proj",
176
- "model.language_model.layers.27.mlp.up_proj",
177
- "model.language_model.layers.16.self_attn.v_proj",
178
- "model.language_model.layers.13.mlp.down_proj",
179
- "model.language_model.layers.25.mlp.down_proj",
180
- "model.language_model.layers.4.self_attn.k_proj",
181
- "model.language_model.layers.21.mlp.up_proj",
182
- "model.language_model.layers.33.self_attn.q_proj",
183
- "model.language_model.layers.36.mlp.up_proj",
184
- "model.language_model.layers.36.mlp.down_proj",
185
- "model.language_model.layers.33.mlp.down_proj",
186
- "model.audio_tower.layers.8.self_attn.q_proj.linear",
187
- "model.language_model.layers.18.self_attn.q_proj",
188
- "model.language_model.layers.12.self_attn.q_proj",
189
- "model.language_model.layers.1.mlp.up_proj",
190
- "model.language_model.layers.8.self_attn.v_proj",
191
- "model.language_model.layers.8.mlp.up_proj",
192
- "model.language_model.layers.9.mlp.gate_proj",
193
- "model.language_model.layers.10.self_attn.v_proj",
194
- "model.language_model.layers.40.mlp.down_proj",
195
- "model.audio_tower.layers.6.self_attn.q_proj.linear",
196
- "model.language_model.layers.30.mlp.gate_proj",
197
- "model.language_model.layers.41.mlp.gate_proj",
198
- "model.language_model.layers.23.self_attn.o_proj",
199
- "model.language_model.layers.39.mlp.up_proj",
200
- "model.language_model.layers.13.mlp.up_proj",
201
- "model.language_model.layers.29.self_attn.o_proj",
202
- "model.language_model.layers.25.mlp.up_proj",
203
- "model.language_model.layers.19.self_attn.o_proj",
204
- "model.language_model.layers.36.self_attn.o_proj",
205
- "model.language_model.layers.8.self_attn.k_proj",
206
- "model.language_model.layers.9.self_attn.k_proj",
207
- "model.language_model.layers.15.self_attn.k_proj",
208
- "model.language_model.layers.19.self_attn.v_proj",
209
- "model.language_model.layers.28.mlp.gate_proj",
210
- "model.language_model.layers.0.mlp.gate_proj",
211
- "model.language_model.layers.29.self_attn.q_proj",
212
- "model.language_model.layers.25.self_attn.o_proj",
213
- "model.language_model.layers.40.mlp.up_proj",
214
- "model.language_model.layers.7.mlp.down_proj",
215
- "model.language_model.layers.14.mlp.gate_proj",
216
- "model.language_model.layers.7.mlp.up_proj",
217
- "model.language_model.layers.22.self_attn.o_proj",
218
- "model.audio_tower.layers.4.self_attn.k_proj.linear",
219
- "model.language_model.layers.3.mlp.down_proj",
220
- "model.language_model.layers.26.mlp.gate_proj",
221
- "model.language_model.layers.16.self_attn.k_proj",
222
- "model.language_model.layers.32.mlp.up_proj",
223
- "model.language_model.layers.5.mlp.up_proj",
224
- "model.language_model.layers.36.mlp.gate_proj",
225
- "model.language_model.layers.0.self_attn.q_proj",
226
- "model.language_model.layers.39.self_attn.o_proj",
227
- "model.audio_tower.layers.3.self_attn.k_proj.linear",
228
- "model.language_model.layers.41.mlp.down_proj",
229
- "model.language_model.layers.2.mlp.down_proj",
230
- "model.language_model.layers.12.mlp.up_proj",
231
- "model.language_model.layers.19.mlp.gate_proj",
232
- "model.language_model.layers.2.mlp.gate_proj",
233
- "model.language_model.layers.29.mlp.up_proj",
234
- "model.language_model.layers.21.mlp.gate_proj",
235
- "model.language_model.layers.4.mlp.down_proj",
236
- "model.audio_tower.layers.1.self_attn.k_proj.linear",
237
- "model.language_model.layers.38.self_attn.q_proj",
238
- "model.language_model.layers.22.self_attn.q_proj",
239
- "model.language_model.layers.14.mlp.up_proj",
240
- "model.language_model.layers.19.mlp.up_proj",
241
- "model.language_model.layers.12.self_attn.o_proj",
242
- "model.language_model.layers.27.self_attn.q_proj",
243
- "model.language_model.layers.20.mlp.gate_proj",
244
- "model.language_model.layers.4.self_attn.q_proj",
245
- "model.language_model.layers.30.mlp.up_proj",
246
- "model.audio_tower.layers.7.self_attn.v_proj.linear",
247
- "model.language_model.layers.9.mlp.down_proj",
248
- "model.language_model.layers.9.mlp.up_proj",
249
- "model.language_model.layers.14.self_attn.v_proj",
250
- "model.audio_tower.layers.1.self_attn.q_proj.linear",
251
- "model.language_model.layers.7.self_attn.v_proj",
252
- "model.language_model.layers.18.self_attn.o_proj",
253
- "model.language_model.layers.28.mlp.down_proj",
254
- "model.language_model.layers.37.self_attn.o_proj",
255
- "model.audio_tower.layers.0.self_attn.k_proj.linear",
256
- "model.language_model.layers.4.mlp.up_proj",
257
- "model.language_model.layers.1.mlp.gate_proj",
258
- "model.language_model.layers.18.mlp.down_proj",
259
- "model.language_model.layers.23.self_attn.k_proj",
260
- "model.language_model.layers.31.mlp.down_proj",
261
- "model.language_model.layers.10.self_attn.q_proj",
262
- "model.language_model.layers.34.mlp.gate_proj",
263
- "model.audio_tower.layers.2.self_attn.q_proj.linear",
264
- "model.language_model.layers.22.mlp.up_proj",
265
- "model.audio_tower.layers.4.self_attn.q_proj.linear",
266
- "model.language_model.layers.15.self_attn.q_proj",
267
- "model.language_model.layers.13.self_attn.q_proj",
268
- "model.language_model.layers.29.mlp.down_proj",
269
- "model.audio_tower.layers.8.self_attn.v_proj.linear",
270
- "model.language_model.layers.41.mlp.up_proj",
271
- "model.language_model.layers.21.mlp.down_proj",
272
- "model.language_model.layers.3.self_attn.q_proj",
273
- "model.language_model.layers.10.mlp.down_proj",
274
- "model.language_model.layers.26.mlp.down_proj",
275
- "model.language_model.layers.37.self_attn.q_proj",
276
- "model.language_model.layers.35.mlp.up_proj",
277
- "model.language_model.layers.19.self_attn.q_proj",
278
- "model.language_model.layers.35.mlp.down_proj",
279
- "model.language_model.layers.39.mlp.down_proj",
280
- "model.language_model.layers.41.self_attn.o_proj",
281
- "model.language_model.layers.2.self_attn.v_proj",
282
- "model.audio_tower.layers.9.self_attn.k_proj.linear",
283
- "model.language_model.layers.26.self_attn.o_proj",
284
- "model.audio_tower.layers.3.self_attn.q_proj.linear",
285
- "model.language_model.layers.38.mlp.up_proj",
286
- "model.language_model.layers.18.mlp.gate_proj",
287
- "model.language_model.layers.11.self_attn.k_proj",
288
- "model.language_model.layers.31.mlp.gate_proj",
289
- "model.language_model.layers.37.mlp.gate_proj",
290
- "model.language_model.layers.6.self_attn.o_proj",
291
- "model.language_model.layers.31.mlp.up_proj",
292
- "model.language_model.layers.7.self_attn.k_proj",
293
- "model.language_model.layers.26.self_attn.q_proj",
294
- "model.language_model.layers.12.self_attn.v_proj",
295
- "model.language_model.layers.11.mlp.up_proj",
296
- "model.audio_tower.layers.11.self_attn.k_proj.linear",
297
- "model.language_model.layers.6.self_attn.v_proj",
298
- "model.language_model.layers.21.self_attn.q_proj",
299
- "model.language_model.layers.27.self_attn.o_proj",
300
- "model.language_model.layers.32.self_attn.q_proj",
301
- "model.language_model.layers.23.mlp.up_proj",
302
- "model.language_model.layers.5.self_attn.o_proj",
303
- "model.language_model.layers.39.self_attn.q_proj",
304
- "model.language_model.layers.2.self_attn.o_proj",
305
- "model.language_model.layers.14.self_attn.o_proj",
306
- "model.language_model.layers.19.mlp.down_proj",
307
- "model.language_model.layers.0.self_attn.o_proj",
308
- "model.language_model.layers.7.mlp.gate_proj",
309
- "model.language_model.layers.12.self_attn.k_proj",
310
- "model.audio_tower.layers.9.self_attn.q_proj.linear",
311
- "model.language_model.layers.3.mlp.up_proj",
312
- "model.language_model.layers.35.self_attn.q_proj",
313
- "model.language_model.layers.15.self_attn.o_proj",
314
- "model.language_model.layers.22.self_attn.v_proj",
315
- "model.language_model.layers.13.mlp.gate_proj",
316
- "model.language_model.layers.16.mlp.gate_proj",
317
- "model.language_model.layers.16.mlp.up_proj",
318
- "model.language_model.layers.23.mlp.down_proj",
319
- "model.language_model.layers.25.mlp.gate_proj",
320
- "model.language_model.layers.38.self_attn.o_proj",
321
- "model.audio_tower.layers.0.self_attn.q_proj.linear",
322
- "model.language_model.layers.23.self_attn.v_proj",
323
- "model.audio_tower.layers.7.self_attn.k_proj.linear",
324
- "model.language_model.layers.33.self_attn.o_proj",
325
- "model.language_model.layers.11.mlp.gate_proj",
326
- "model.language_model.layers.34.self_attn.o_proj"
327
- ],
328
- "target_parameters": null,
329
- "task_type": "CAUSAL_LM",
330
- "trainable_token_indices": null,
331
- "use_bdlora": null,
332
- "use_dora": false,
333
- "use_qalora": false,
334
- "use_rslora": false
335
- }