Safetensors
MLX
mlx-vlm
gemma4_unified
gemma4
agentic
coding
tool-use
lora-merged
4-bit precision
True2456 commited on
Commit
2d2ff1f
·
verified ·
1 Parent(s): ea47da8

Upload adapter_config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. adapter_config.json +339 -0
adapter_config.json ADDED
@@ -0,0 +1,339 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fine_tune_type": "lora",
3
+ "num_layers": -1,
4
+ "lora_parameters": {
5
+ "rank": 8,
6
+ "dropout": 0.0,
7
+ "scale": 2.0,
8
+ "keys": [
9
+ "language_model.model.layers.47.mlp.up_proj",
10
+ "language_model.model.layers.47.mlp.down_proj",
11
+ "language_model.model.layers.47.mlp.gate_proj",
12
+ "language_model.model.layers.47.self_attn.o_proj",
13
+ "language_model.model.layers.47.self_attn.k_proj",
14
+ "language_model.model.layers.47.self_attn.q_proj",
15
+ "language_model.model.layers.46.mlp.up_proj",
16
+ "language_model.model.layers.46.mlp.down_proj",
17
+ "language_model.model.layers.46.mlp.gate_proj",
18
+ "language_model.model.layers.46.self_attn.o_proj",
19
+ "language_model.model.layers.46.self_attn.v_proj",
20
+ "language_model.model.layers.46.self_attn.k_proj",
21
+ "language_model.model.layers.46.self_attn.q_proj",
22
+ "language_model.model.layers.45.mlp.up_proj",
23
+ "language_model.model.layers.45.mlp.down_proj",
24
+ "language_model.model.layers.45.mlp.gate_proj",
25
+ "language_model.model.layers.45.self_attn.o_proj",
26
+ "language_model.model.layers.45.self_attn.v_proj",
27
+ "language_model.model.layers.45.self_attn.k_proj",
28
+ "language_model.model.layers.45.self_attn.q_proj",
29
+ "language_model.model.layers.44.mlp.up_proj",
30
+ "language_model.model.layers.44.mlp.down_proj",
31
+ "language_model.model.layers.44.mlp.gate_proj",
32
+ "language_model.model.layers.44.self_attn.o_proj",
33
+ "language_model.model.layers.44.self_attn.v_proj",
34
+ "language_model.model.layers.44.self_attn.k_proj",
35
+ "language_model.model.layers.44.self_attn.q_proj",
36
+ "language_model.model.layers.43.mlp.up_proj",
37
+ "language_model.model.layers.43.mlp.down_proj",
38
+ "language_model.model.layers.43.mlp.gate_proj",
39
+ "language_model.model.layers.43.self_attn.o_proj",
40
+ "language_model.model.layers.43.self_attn.v_proj",
41
+ "language_model.model.layers.43.self_attn.k_proj",
42
+ "language_model.model.layers.43.self_attn.q_proj",
43
+ "language_model.model.layers.42.mlp.up_proj",
44
+ "language_model.model.layers.42.mlp.down_proj",
45
+ "language_model.model.layers.42.mlp.gate_proj",
46
+ "language_model.model.layers.42.self_attn.o_proj",
47
+ "language_model.model.layers.42.self_attn.v_proj",
48
+ "language_model.model.layers.42.self_attn.k_proj",
49
+ "language_model.model.layers.42.self_attn.q_proj",
50
+ "language_model.model.layers.41.mlp.up_proj",
51
+ "language_model.model.layers.41.mlp.down_proj",
52
+ "language_model.model.layers.41.mlp.gate_proj",
53
+ "language_model.model.layers.41.self_attn.o_proj",
54
+ "language_model.model.layers.41.self_attn.k_proj",
55
+ "language_model.model.layers.41.self_attn.q_proj",
56
+ "language_model.model.layers.40.mlp.up_proj",
57
+ "language_model.model.layers.40.mlp.down_proj",
58
+ "language_model.model.layers.40.mlp.gate_proj",
59
+ "language_model.model.layers.40.self_attn.o_proj",
60
+ "language_model.model.layers.40.self_attn.v_proj",
61
+ "language_model.model.layers.40.self_attn.k_proj",
62
+ "language_model.model.layers.40.self_attn.q_proj",
63
+ "language_model.model.layers.39.mlp.up_proj",
64
+ "language_model.model.layers.39.mlp.down_proj",
65
+ "language_model.model.layers.39.mlp.gate_proj",
66
+ "language_model.model.layers.39.self_attn.o_proj",
67
+ "language_model.model.layers.39.self_attn.v_proj",
68
+ "language_model.model.layers.39.self_attn.k_proj",
69
+ "language_model.model.layers.39.self_attn.q_proj",
70
+ "language_model.model.layers.38.mlp.up_proj",
71
+ "language_model.model.layers.38.mlp.down_proj",
72
+ "language_model.model.layers.38.mlp.gate_proj",
73
+ "language_model.model.layers.38.self_attn.o_proj",
74
+ "language_model.model.layers.38.self_attn.v_proj",
75
+ "language_model.model.layers.38.self_attn.k_proj",
76
+ "language_model.model.layers.38.self_attn.q_proj",
77
+ "language_model.model.layers.37.mlp.up_proj",
78
+ "language_model.model.layers.37.mlp.down_proj",
79
+ "language_model.model.layers.37.mlp.gate_proj",
80
+ "language_model.model.layers.37.self_attn.o_proj",
81
+ "language_model.model.layers.37.self_attn.v_proj",
82
+ "language_model.model.layers.37.self_attn.k_proj",
83
+ "language_model.model.layers.37.self_attn.q_proj",
84
+ "language_model.model.layers.36.mlp.up_proj",
85
+ "language_model.model.layers.36.mlp.down_proj",
86
+ "language_model.model.layers.36.mlp.gate_proj",
87
+ "language_model.model.layers.36.self_attn.o_proj",
88
+ "language_model.model.layers.36.self_attn.v_proj",
89
+ "language_model.model.layers.36.self_attn.k_proj",
90
+ "language_model.model.layers.36.self_attn.q_proj",
91
+ "language_model.model.layers.35.mlp.up_proj",
92
+ "language_model.model.layers.35.mlp.down_proj",
93
+ "language_model.model.layers.35.mlp.gate_proj",
94
+ "language_model.model.layers.35.self_attn.o_proj",
95
+ "language_model.model.layers.35.self_attn.k_proj",
96
+ "language_model.model.layers.35.self_attn.q_proj",
97
+ "language_model.model.layers.34.mlp.up_proj",
98
+ "language_model.model.layers.34.mlp.down_proj",
99
+ "language_model.model.layers.34.mlp.gate_proj",
100
+ "language_model.model.layers.34.self_attn.o_proj",
101
+ "language_model.model.layers.34.self_attn.v_proj",
102
+ "language_model.model.layers.34.self_attn.k_proj",
103
+ "language_model.model.layers.34.self_attn.q_proj",
104
+ "language_model.model.layers.33.mlp.up_proj",
105
+ "language_model.model.layers.33.mlp.down_proj",
106
+ "language_model.model.layers.33.mlp.gate_proj",
107
+ "language_model.model.layers.33.self_attn.o_proj",
108
+ "language_model.model.layers.33.self_attn.v_proj",
109
+ "language_model.model.layers.33.self_attn.k_proj",
110
+ "language_model.model.layers.33.self_attn.q_proj",
111
+ "language_model.model.layers.32.mlp.up_proj",
112
+ "language_model.model.layers.32.mlp.down_proj",
113
+ "language_model.model.layers.32.mlp.gate_proj",
114
+ "language_model.model.layers.32.self_attn.o_proj",
115
+ "language_model.model.layers.32.self_attn.v_proj",
116
+ "language_model.model.layers.32.self_attn.k_proj",
117
+ "language_model.model.layers.32.self_attn.q_proj",
118
+ "language_model.model.layers.31.mlp.up_proj",
119
+ "language_model.model.layers.31.mlp.down_proj",
120
+ "language_model.model.layers.31.mlp.gate_proj",
121
+ "language_model.model.layers.31.self_attn.o_proj",
122
+ "language_model.model.layers.31.self_attn.v_proj",
123
+ "language_model.model.layers.31.self_attn.k_proj",
124
+ "language_model.model.layers.31.self_attn.q_proj",
125
+ "language_model.model.layers.30.mlp.up_proj",
126
+ "language_model.model.layers.30.mlp.down_proj",
127
+ "language_model.model.layers.30.mlp.gate_proj",
128
+ "language_model.model.layers.30.self_attn.o_proj",
129
+ "language_model.model.layers.30.self_attn.v_proj",
130
+ "language_model.model.layers.30.self_attn.k_proj",
131
+ "language_model.model.layers.30.self_attn.q_proj",
132
+ "language_model.model.layers.29.mlp.up_proj",
133
+ "language_model.model.layers.29.mlp.down_proj",
134
+ "language_model.model.layers.29.mlp.gate_proj",
135
+ "language_model.model.layers.29.self_attn.o_proj",
136
+ "language_model.model.layers.29.self_attn.k_proj",
137
+ "language_model.model.layers.29.self_attn.q_proj",
138
+ "language_model.model.layers.28.mlp.up_proj",
139
+ "language_model.model.layers.28.mlp.down_proj",
140
+ "language_model.model.layers.28.mlp.gate_proj",
141
+ "language_model.model.layers.28.self_attn.o_proj",
142
+ "language_model.model.layers.28.self_attn.v_proj",
143
+ "language_model.model.layers.28.self_attn.k_proj",
144
+ "language_model.model.layers.28.self_attn.q_proj",
145
+ "language_model.model.layers.27.mlp.up_proj",
146
+ "language_model.model.layers.27.mlp.down_proj",
147
+ "language_model.model.layers.27.mlp.gate_proj",
148
+ "language_model.model.layers.27.self_attn.o_proj",
149
+ "language_model.model.layers.27.self_attn.v_proj",
150
+ "language_model.model.layers.27.self_attn.k_proj",
151
+ "language_model.model.layers.27.self_attn.q_proj",
152
+ "language_model.model.layers.26.mlp.up_proj",
153
+ "language_model.model.layers.26.mlp.down_proj",
154
+ "language_model.model.layers.26.mlp.gate_proj",
155
+ "language_model.model.layers.26.self_attn.o_proj",
156
+ "language_model.model.layers.26.self_attn.v_proj",
157
+ "language_model.model.layers.26.self_attn.k_proj",
158
+ "language_model.model.layers.26.self_attn.q_proj",
159
+ "language_model.model.layers.25.mlp.up_proj",
160
+ "language_model.model.layers.25.mlp.down_proj",
161
+ "language_model.model.layers.25.mlp.gate_proj",
162
+ "language_model.model.layers.25.self_attn.o_proj",
163
+ "language_model.model.layers.25.self_attn.v_proj",
164
+ "language_model.model.layers.25.self_attn.k_proj",
165
+ "language_model.model.layers.25.self_attn.q_proj",
166
+ "language_model.model.layers.24.mlp.up_proj",
167
+ "language_model.model.layers.24.mlp.down_proj",
168
+ "language_model.model.layers.24.mlp.gate_proj",
169
+ "language_model.model.layers.24.self_attn.o_proj",
170
+ "language_model.model.layers.24.self_attn.v_proj",
171
+ "language_model.model.layers.24.self_attn.k_proj",
172
+ "language_model.model.layers.24.self_attn.q_proj",
173
+ "language_model.model.layers.23.mlp.up_proj",
174
+ "language_model.model.layers.23.mlp.down_proj",
175
+ "language_model.model.layers.23.mlp.gate_proj",
176
+ "language_model.model.layers.23.self_attn.o_proj",
177
+ "language_model.model.layers.23.self_attn.k_proj",
178
+ "language_model.model.layers.23.self_attn.q_proj",
179
+ "language_model.model.layers.22.mlp.up_proj",
180
+ "language_model.model.layers.22.mlp.down_proj",
181
+ "language_model.model.layers.22.mlp.gate_proj",
182
+ "language_model.model.layers.22.self_attn.o_proj",
183
+ "language_model.model.layers.22.self_attn.v_proj",
184
+ "language_model.model.layers.22.self_attn.k_proj",
185
+ "language_model.model.layers.22.self_attn.q_proj",
186
+ "language_model.model.layers.21.mlp.up_proj",
187
+ "language_model.model.layers.21.mlp.down_proj",
188
+ "language_model.model.layers.21.mlp.gate_proj",
189
+ "language_model.model.layers.21.self_attn.o_proj",
190
+ "language_model.model.layers.21.self_attn.v_proj",
191
+ "language_model.model.layers.21.self_attn.k_proj",
192
+ "language_model.model.layers.21.self_attn.q_proj",
193
+ "language_model.model.layers.20.mlp.up_proj",
194
+ "language_model.model.layers.20.mlp.down_proj",
195
+ "language_model.model.layers.20.mlp.gate_proj",
196
+ "language_model.model.layers.20.self_attn.o_proj",
197
+ "language_model.model.layers.20.self_attn.v_proj",
198
+ "language_model.model.layers.20.self_attn.k_proj",
199
+ "language_model.model.layers.20.self_attn.q_proj",
200
+ "language_model.model.layers.19.mlp.up_proj",
201
+ "language_model.model.layers.19.mlp.down_proj",
202
+ "language_model.model.layers.19.mlp.gate_proj",
203
+ "language_model.model.layers.19.self_attn.o_proj",
204
+ "language_model.model.layers.19.self_attn.v_proj",
205
+ "language_model.model.layers.19.self_attn.k_proj",
206
+ "language_model.model.layers.19.self_attn.q_proj",
207
+ "language_model.model.layers.18.mlp.up_proj",
208
+ "language_model.model.layers.18.mlp.down_proj",
209
+ "language_model.model.layers.18.mlp.gate_proj",
210
+ "language_model.model.layers.18.self_attn.o_proj",
211
+ "language_model.model.layers.18.self_attn.v_proj",
212
+ "language_model.model.layers.18.self_attn.k_proj",
213
+ "language_model.model.layers.18.self_attn.q_proj",
214
+ "language_model.model.layers.17.mlp.up_proj",
215
+ "language_model.model.layers.17.mlp.down_proj",
216
+ "language_model.model.layers.17.mlp.gate_proj",
217
+ "language_model.model.layers.17.self_attn.o_proj",
218
+ "language_model.model.layers.17.self_attn.k_proj",
219
+ "language_model.model.layers.17.self_attn.q_proj",
220
+ "language_model.model.layers.16.mlp.up_proj",
221
+ "language_model.model.layers.16.mlp.down_proj",
222
+ "language_model.model.layers.16.mlp.gate_proj",
223
+ "language_model.model.layers.16.self_attn.o_proj",
224
+ "language_model.model.layers.16.self_attn.v_proj",
225
+ "language_model.model.layers.16.self_attn.k_proj",
226
+ "language_model.model.layers.16.self_attn.q_proj",
227
+ "language_model.model.layers.15.mlp.up_proj",
228
+ "language_model.model.layers.15.mlp.down_proj",
229
+ "language_model.model.layers.15.mlp.gate_proj",
230
+ "language_model.model.layers.15.self_attn.o_proj",
231
+ "language_model.model.layers.15.self_attn.v_proj",
232
+ "language_model.model.layers.15.self_attn.k_proj",
233
+ "language_model.model.layers.15.self_attn.q_proj",
234
+ "language_model.model.layers.14.mlp.up_proj",
235
+ "language_model.model.layers.14.mlp.down_proj",
236
+ "language_model.model.layers.14.mlp.gate_proj",
237
+ "language_model.model.layers.14.self_attn.o_proj",
238
+ "language_model.model.layers.14.self_attn.v_proj",
239
+ "language_model.model.layers.14.self_attn.k_proj",
240
+ "language_model.model.layers.14.self_attn.q_proj",
241
+ "language_model.model.layers.13.mlp.up_proj",
242
+ "language_model.model.layers.13.mlp.down_proj",
243
+ "language_model.model.layers.13.mlp.gate_proj",
244
+ "language_model.model.layers.13.self_attn.o_proj",
245
+ "language_model.model.layers.13.self_attn.v_proj",
246
+ "language_model.model.layers.13.self_attn.k_proj",
247
+ "language_model.model.layers.13.self_attn.q_proj",
248
+ "language_model.model.layers.12.mlp.up_proj",
249
+ "language_model.model.layers.12.mlp.down_proj",
250
+ "language_model.model.layers.12.mlp.gate_proj",
251
+ "language_model.model.layers.12.self_attn.o_proj",
252
+ "language_model.model.layers.12.self_attn.v_proj",
253
+ "language_model.model.layers.12.self_attn.k_proj",
254
+ "language_model.model.layers.12.self_attn.q_proj",
255
+ "language_model.model.layers.11.mlp.up_proj",
256
+ "language_model.model.layers.11.mlp.down_proj",
257
+ "language_model.model.layers.11.mlp.gate_proj",
258
+ "language_model.model.layers.11.self_attn.o_proj",
259
+ "language_model.model.layers.11.self_attn.k_proj",
260
+ "language_model.model.layers.11.self_attn.q_proj",
261
+ "language_model.model.layers.10.mlp.up_proj",
262
+ "language_model.model.layers.10.mlp.down_proj",
263
+ "language_model.model.layers.10.mlp.gate_proj",
264
+ "language_model.model.layers.10.self_attn.o_proj",
265
+ "language_model.model.layers.10.self_attn.v_proj",
266
+ "language_model.model.layers.10.self_attn.k_proj",
267
+ "language_model.model.layers.10.self_attn.q_proj",
268
+ "language_model.model.layers.9.mlp.up_proj",
269
+ "language_model.model.layers.9.mlp.down_proj",
270
+ "language_model.model.layers.9.mlp.gate_proj",
271
+ "language_model.model.layers.9.self_attn.o_proj",
272
+ "language_model.model.layers.9.self_attn.v_proj",
273
+ "language_model.model.layers.9.self_attn.k_proj",
274
+ "language_model.model.layers.9.self_attn.q_proj",
275
+ "language_model.model.layers.8.mlp.up_proj",
276
+ "language_model.model.layers.8.mlp.down_proj",
277
+ "language_model.model.layers.8.mlp.gate_proj",
278
+ "language_model.model.layers.8.self_attn.o_proj",
279
+ "language_model.model.layers.8.self_attn.v_proj",
280
+ "language_model.model.layers.8.self_attn.k_proj",
281
+ "language_model.model.layers.8.self_attn.q_proj",
282
+ "language_model.model.layers.7.mlp.up_proj",
283
+ "language_model.model.layers.7.mlp.down_proj",
284
+ "language_model.model.layers.7.mlp.gate_proj",
285
+ "language_model.model.layers.7.self_attn.o_proj",
286
+ "language_model.model.layers.7.self_attn.v_proj",
287
+ "language_model.model.layers.7.self_attn.k_proj",
288
+ "language_model.model.layers.7.self_attn.q_proj",
289
+ "language_model.model.layers.6.mlp.up_proj",
290
+ "language_model.model.layers.6.mlp.down_proj",
291
+ "language_model.model.layers.6.mlp.gate_proj",
292
+ "language_model.model.layers.6.self_attn.o_proj",
293
+ "language_model.model.layers.6.self_attn.v_proj",
294
+ "language_model.model.layers.6.self_attn.k_proj",
295
+ "language_model.model.layers.6.self_attn.q_proj",
296
+ "language_model.model.layers.5.mlp.up_proj",
297
+ "language_model.model.layers.5.mlp.down_proj",
298
+ "language_model.model.layers.5.mlp.gate_proj",
299
+ "language_model.model.layers.5.self_attn.o_proj",
300
+ "language_model.model.layers.5.self_attn.k_proj",
301
+ "language_model.model.layers.5.self_attn.q_proj",
302
+ "language_model.model.layers.4.mlp.up_proj",
303
+ "language_model.model.layers.4.mlp.down_proj",
304
+ "language_model.model.layers.4.mlp.gate_proj",
305
+ "language_model.model.layers.4.self_attn.o_proj",
306
+ "language_model.model.layers.4.self_attn.v_proj",
307
+ "language_model.model.layers.4.self_attn.k_proj",
308
+ "language_model.model.layers.4.self_attn.q_proj",
309
+ "language_model.model.layers.3.mlp.up_proj",
310
+ "language_model.model.layers.3.mlp.down_proj",
311
+ "language_model.model.layers.3.mlp.gate_proj",
312
+ "language_model.model.layers.3.self_attn.o_proj",
313
+ "language_model.model.layers.3.self_attn.v_proj",
314
+ "language_model.model.layers.3.self_attn.k_proj",
315
+ "language_model.model.layers.3.self_attn.q_proj",
316
+ "language_model.model.layers.2.mlp.up_proj",
317
+ "language_model.model.layers.2.mlp.down_proj",
318
+ "language_model.model.layers.2.mlp.gate_proj",
319
+ "language_model.model.layers.2.self_attn.o_proj",
320
+ "language_model.model.layers.2.self_attn.v_proj",
321
+ "language_model.model.layers.2.self_attn.k_proj",
322
+ "language_model.model.layers.2.self_attn.q_proj",
323
+ "language_model.model.layers.1.mlp.up_proj",
324
+ "language_model.model.layers.1.mlp.down_proj",
325
+ "language_model.model.layers.1.mlp.gate_proj",
326
+ "language_model.model.layers.1.self_attn.o_proj",
327
+ "language_model.model.layers.1.self_attn.v_proj",
328
+ "language_model.model.layers.1.self_attn.k_proj",
329
+ "language_model.model.layers.1.self_attn.q_proj",
330
+ "language_model.model.layers.0.mlp.up_proj",
331
+ "language_model.model.layers.0.mlp.down_proj",
332
+ "language_model.model.layers.0.mlp.gate_proj",
333
+ "language_model.model.layers.0.self_attn.o_proj",
334
+ "language_model.model.layers.0.self_attn.v_proj",
335
+ "language_model.model.layers.0.self_attn.k_proj",
336
+ "language_model.model.layers.0.self_attn.q_proj"
337
+ ]
338
+ }
339
+ }