aspctu commited on
Commit
3895206
·
verified ·
1 Parent(s): 257fbd1

Update hf_quant_config.json

Browse files
Files changed (1) hide show
  1. hf_quant_config.json +245 -245
hf_quant_config.json CHANGED
@@ -8,251 +8,251 @@
8
  "kv_cache_quant_algo": "FP8",
9
  "group_size": 16,
10
  "exclude_modules": [
11
- "language_model.lm_head",
12
- "language_model.model.layers.0.self_attn.kv_a_proj_with_mqa",
13
- "language_model.model.layers.0.self_attn.kv_b_proj",
14
- "language_model.model.layers.0.self_attn.q_a_proj",
15
- "language_model.model.layers.0.self_attn.q_b_proj",
16
- "language_model.model.layers.1.self_attn.kv_a_proj_with_mqa",
17
- "language_model.model.layers.1.self_attn.kv_b_proj",
18
- "language_model.model.layers.1.self_attn.q_a_proj",
19
- "language_model.model.layers.1.self_attn.q_b_proj",
20
- "language_model.model.layers.10.self_attn.kv_a_proj_with_mqa",
21
- "language_model.model.layers.10.self_attn.kv_b_proj",
22
- "language_model.model.layers.10.self_attn.q_a_proj",
23
- "language_model.model.layers.10.self_attn.q_b_proj",
24
- "language_model.model.layers.11.self_attn.kv_a_proj_with_mqa",
25
- "language_model.model.layers.11.self_attn.kv_b_proj",
26
- "language_model.model.layers.11.self_attn.q_a_proj",
27
- "language_model.model.layers.11.self_attn.q_b_proj",
28
- "language_model.model.layers.12.self_attn.kv_a_proj_with_mqa",
29
- "language_model.model.layers.12.self_attn.kv_b_proj",
30
- "language_model.model.layers.12.self_attn.q_a_proj",
31
- "language_model.model.layers.12.self_attn.q_b_proj",
32
- "language_model.model.layers.13.self_attn.kv_a_proj_with_mqa",
33
- "language_model.model.layers.13.self_attn.kv_b_proj",
34
- "language_model.model.layers.13.self_attn.q_a_proj",
35
- "language_model.model.layers.13.self_attn.q_b_proj",
36
- "language_model.model.layers.14.self_attn.kv_a_proj_with_mqa",
37
- "language_model.model.layers.14.self_attn.kv_b_proj",
38
- "language_model.model.layers.14.self_attn.q_a_proj",
39
- "language_model.model.layers.14.self_attn.q_b_proj",
40
- "language_model.model.layers.15.self_attn.kv_a_proj_with_mqa",
41
- "language_model.model.layers.15.self_attn.kv_b_proj",
42
- "language_model.model.layers.15.self_attn.q_a_proj",
43
- "language_model.model.layers.15.self_attn.q_b_proj",
44
- "language_model.model.layers.16.self_attn.kv_a_proj_with_mqa",
45
- "language_model.model.layers.16.self_attn.kv_b_proj",
46
- "language_model.model.layers.16.self_attn.q_a_proj",
47
- "language_model.model.layers.16.self_attn.q_b_proj",
48
- "language_model.model.layers.17.self_attn.kv_a_proj_with_mqa",
49
- "language_model.model.layers.17.self_attn.kv_b_proj",
50
- "language_model.model.layers.17.self_attn.q_a_proj",
51
- "language_model.model.layers.17.self_attn.q_b_proj",
52
- "language_model.model.layers.18.self_attn.kv_a_proj_with_mqa",
53
- "language_model.model.layers.18.self_attn.kv_b_proj",
54
- "language_model.model.layers.18.self_attn.q_a_proj",
55
- "language_model.model.layers.18.self_attn.q_b_proj",
56
- "language_model.model.layers.19.self_attn.kv_a_proj_with_mqa",
57
- "language_model.model.layers.19.self_attn.kv_b_proj",
58
- "language_model.model.layers.19.self_attn.q_a_proj",
59
- "language_model.model.layers.19.self_attn.q_b_proj",
60
- "language_model.model.layers.2.self_attn.kv_a_proj_with_mqa",
61
- "language_model.model.layers.2.self_attn.kv_b_proj",
62
- "language_model.model.layers.2.self_attn.q_a_proj",
63
- "language_model.model.layers.2.self_attn.q_b_proj",
64
- "language_model.model.layers.20.self_attn.kv_a_proj_with_mqa",
65
- "language_model.model.layers.20.self_attn.kv_b_proj",
66
- "language_model.model.layers.20.self_attn.q_a_proj",
67
- "language_model.model.layers.20.self_attn.q_b_proj",
68
- "language_model.model.layers.21.self_attn.kv_a_proj_with_mqa",
69
- "language_model.model.layers.21.self_attn.kv_b_proj",
70
- "language_model.model.layers.21.self_attn.q_a_proj",
71
- "language_model.model.layers.21.self_attn.q_b_proj",
72
- "language_model.model.layers.22.self_attn.kv_a_proj_with_mqa",
73
- "language_model.model.layers.22.self_attn.kv_b_proj",
74
- "language_model.model.layers.22.self_attn.q_a_proj",
75
- "language_model.model.layers.22.self_attn.q_b_proj",
76
- "language_model.model.layers.23.self_attn.kv_a_proj_with_mqa",
77
- "language_model.model.layers.23.self_attn.kv_b_proj",
78
- "language_model.model.layers.23.self_attn.q_a_proj",
79
- "language_model.model.layers.23.self_attn.q_b_proj",
80
- "language_model.model.layers.24.self_attn.kv_a_proj_with_mqa",
81
- "language_model.model.layers.24.self_attn.kv_b_proj",
82
- "language_model.model.layers.24.self_attn.q_a_proj",
83
- "language_model.model.layers.24.self_attn.q_b_proj",
84
- "language_model.model.layers.25.self_attn.kv_a_proj_with_mqa",
85
- "language_model.model.layers.25.self_attn.kv_b_proj",
86
- "language_model.model.layers.25.self_attn.q_a_proj",
87
- "language_model.model.layers.25.self_attn.q_b_proj",
88
- "language_model.model.layers.26.self_attn.kv_a_proj_with_mqa",
89
- "language_model.model.layers.26.self_attn.kv_b_proj",
90
- "language_model.model.layers.26.self_attn.q_a_proj",
91
- "language_model.model.layers.26.self_attn.q_b_proj",
92
- "language_model.model.layers.27.self_attn.kv_a_proj_with_mqa",
93
- "language_model.model.layers.27.self_attn.kv_b_proj",
94
- "language_model.model.layers.27.self_attn.q_a_proj",
95
- "language_model.model.layers.27.self_attn.q_b_proj",
96
- "language_model.model.layers.28.self_attn.kv_a_proj_with_mqa",
97
- "language_model.model.layers.28.self_attn.kv_b_proj",
98
- "language_model.model.layers.28.self_attn.q_a_proj",
99
- "language_model.model.layers.28.self_attn.q_b_proj",
100
- "language_model.model.layers.29.self_attn.kv_a_proj_with_mqa",
101
- "language_model.model.layers.29.self_attn.kv_b_proj",
102
- "language_model.model.layers.29.self_attn.q_a_proj",
103
- "language_model.model.layers.29.self_attn.q_b_proj",
104
- "language_model.model.layers.3.self_attn.kv_a_proj_with_mqa",
105
- "language_model.model.layers.3.self_attn.kv_b_proj",
106
- "language_model.model.layers.3.self_attn.q_a_proj",
107
- "language_model.model.layers.3.self_attn.q_b_proj",
108
- "language_model.model.layers.30.self_attn.kv_a_proj_with_mqa",
109
- "language_model.model.layers.30.self_attn.kv_b_proj",
110
- "language_model.model.layers.30.self_attn.q_a_proj",
111
- "language_model.model.layers.30.self_attn.q_b_proj",
112
- "language_model.model.layers.31.self_attn.kv_a_proj_with_mqa",
113
- "language_model.model.layers.31.self_attn.kv_b_proj",
114
- "language_model.model.layers.31.self_attn.q_a_proj",
115
- "language_model.model.layers.31.self_attn.q_b_proj",
116
- "language_model.model.layers.32.self_attn.kv_a_proj_with_mqa",
117
- "language_model.model.layers.32.self_attn.kv_b_proj",
118
- "language_model.model.layers.32.self_attn.q_a_proj",
119
- "language_model.model.layers.32.self_attn.q_b_proj",
120
- "language_model.model.layers.33.self_attn.kv_a_proj_with_mqa",
121
- "language_model.model.layers.33.self_attn.kv_b_proj",
122
- "language_model.model.layers.33.self_attn.q_a_proj",
123
- "language_model.model.layers.33.self_attn.q_b_proj",
124
- "language_model.model.layers.34.self_attn.kv_a_proj_with_mqa",
125
- "language_model.model.layers.34.self_attn.kv_b_proj",
126
- "language_model.model.layers.34.self_attn.q_a_proj",
127
- "language_model.model.layers.34.self_attn.q_b_proj",
128
- "language_model.model.layers.35.self_attn.kv_a_proj_with_mqa",
129
- "language_model.model.layers.35.self_attn.kv_b_proj",
130
- "language_model.model.layers.35.self_attn.q_a_proj",
131
- "language_model.model.layers.35.self_attn.q_b_proj",
132
- "language_model.model.layers.36.self_attn.kv_a_proj_with_mqa",
133
- "language_model.model.layers.36.self_attn.kv_b_proj",
134
- "language_model.model.layers.36.self_attn.q_a_proj",
135
- "language_model.model.layers.36.self_attn.q_b_proj",
136
- "language_model.model.layers.37.self_attn.kv_a_proj_with_mqa",
137
- "language_model.model.layers.37.self_attn.kv_b_proj",
138
- "language_model.model.layers.37.self_attn.q_a_proj",
139
- "language_model.model.layers.37.self_attn.q_b_proj",
140
- "language_model.model.layers.38.self_attn.kv_a_proj_with_mqa",
141
- "language_model.model.layers.38.self_attn.kv_b_proj",
142
- "language_model.model.layers.38.self_attn.q_a_proj",
143
- "language_model.model.layers.38.self_attn.q_b_proj",
144
- "language_model.model.layers.39.self_attn.kv_a_proj_with_mqa",
145
- "language_model.model.layers.39.self_attn.kv_b_proj",
146
- "language_model.model.layers.39.self_attn.q_a_proj",
147
- "language_model.model.layers.39.self_attn.q_b_proj",
148
- "language_model.model.layers.4.self_attn.kv_a_proj_with_mqa",
149
- "language_model.model.layers.4.self_attn.kv_b_proj",
150
- "language_model.model.layers.4.self_attn.q_a_proj",
151
- "language_model.model.layers.4.self_attn.q_b_proj",
152
- "language_model.model.layers.40.self_attn.kv_a_proj_with_mqa",
153
- "language_model.model.layers.40.self_attn.kv_b_proj",
154
- "language_model.model.layers.40.self_attn.q_a_proj",
155
- "language_model.model.layers.40.self_attn.q_b_proj",
156
- "language_model.model.layers.41.self_attn.kv_a_proj_with_mqa",
157
- "language_model.model.layers.41.self_attn.kv_b_proj",
158
- "language_model.model.layers.41.self_attn.q_a_proj",
159
- "language_model.model.layers.41.self_attn.q_b_proj",
160
- "language_model.model.layers.42.self_attn.kv_a_proj_with_mqa",
161
- "language_model.model.layers.42.self_attn.kv_b_proj",
162
- "language_model.model.layers.42.self_attn.q_a_proj",
163
- "language_model.model.layers.42.self_attn.q_b_proj",
164
- "language_model.model.layers.43.self_attn.kv_a_proj_with_mqa",
165
- "language_model.model.layers.43.self_attn.kv_b_proj",
166
- "language_model.model.layers.43.self_attn.q_a_proj",
167
- "language_model.model.layers.43.self_attn.q_b_proj",
168
- "language_model.model.layers.44.self_attn.kv_a_proj_with_mqa",
169
- "language_model.model.layers.44.self_attn.kv_b_proj",
170
- "language_model.model.layers.44.self_attn.q_a_proj",
171
- "language_model.model.layers.44.self_attn.q_b_proj",
172
- "language_model.model.layers.45.self_attn.kv_a_proj_with_mqa",
173
- "language_model.model.layers.45.self_attn.kv_b_proj",
174
- "language_model.model.layers.45.self_attn.q_a_proj",
175
- "language_model.model.layers.45.self_attn.q_b_proj",
176
- "language_model.model.layers.46.self_attn.kv_a_proj_with_mqa",
177
- "language_model.model.layers.46.self_attn.kv_b_proj",
178
- "language_model.model.layers.46.self_attn.q_a_proj",
179
- "language_model.model.layers.46.self_attn.q_b_proj",
180
- "language_model.model.layers.47.self_attn.kv_a_proj_with_mqa",
181
- "language_model.model.layers.47.self_attn.kv_b_proj",
182
- "language_model.model.layers.47.self_attn.q_a_proj",
183
- "language_model.model.layers.47.self_attn.q_b_proj",
184
- "language_model.model.layers.48.self_attn.kv_a_proj_with_mqa",
185
- "language_model.model.layers.48.self_attn.kv_b_proj",
186
- "language_model.model.layers.48.self_attn.q_a_proj",
187
- "language_model.model.layers.48.self_attn.q_b_proj",
188
- "language_model.model.layers.49.self_attn.kv_a_proj_with_mqa",
189
- "language_model.model.layers.49.self_attn.kv_b_proj",
190
- "language_model.model.layers.49.self_attn.q_a_proj",
191
- "language_model.model.layers.49.self_attn.q_b_proj",
192
- "language_model.model.layers.5.self_attn.kv_a_proj_with_mqa",
193
- "language_model.model.layers.5.self_attn.kv_b_proj",
194
- "language_model.model.layers.5.self_attn.q_a_proj",
195
- "language_model.model.layers.5.self_attn.q_b_proj",
196
- "language_model.model.layers.50.self_attn.kv_a_proj_with_mqa",
197
- "language_model.model.layers.50.self_attn.kv_b_proj",
198
- "language_model.model.layers.50.self_attn.q_a_proj",
199
- "language_model.model.layers.50.self_attn.q_b_proj",
200
- "language_model.model.layers.51.self_attn.kv_a_proj_with_mqa",
201
- "language_model.model.layers.51.self_attn.kv_b_proj",
202
- "language_model.model.layers.51.self_attn.q_a_proj",
203
- "language_model.model.layers.51.self_attn.q_b_proj",
204
- "language_model.model.layers.52.self_attn.kv_a_proj_with_mqa",
205
- "language_model.model.layers.52.self_attn.kv_b_proj",
206
- "language_model.model.layers.52.self_attn.q_a_proj",
207
- "language_model.model.layers.52.self_attn.q_b_proj",
208
- "language_model.model.layers.53.self_attn.kv_a_proj_with_mqa",
209
- "language_model.model.layers.53.self_attn.kv_b_proj",
210
- "language_model.model.layers.53.self_attn.q_a_proj",
211
- "language_model.model.layers.53.self_attn.q_b_proj",
212
- "language_model.model.layers.54.self_attn.kv_a_proj_with_mqa",
213
- "language_model.model.layers.54.self_attn.kv_b_proj",
214
- "language_model.model.layers.54.self_attn.q_a_proj",
215
- "language_model.model.layers.54.self_attn.q_b_proj",
216
- "language_model.model.layers.55.self_attn.kv_a_proj_with_mqa",
217
- "language_model.model.layers.55.self_attn.kv_b_proj",
218
- "language_model.model.layers.55.self_attn.q_a_proj",
219
- "language_model.model.layers.55.self_attn.q_b_proj",
220
- "language_model.model.layers.56.self_attn.kv_a_proj_with_mqa",
221
- "language_model.model.layers.56.self_attn.kv_b_proj",
222
- "language_model.model.layers.56.self_attn.q_a_proj",
223
- "language_model.model.layers.56.self_attn.q_b_proj",
224
- "language_model.model.layers.57.self_attn.kv_a_proj_with_mqa",
225
- "language_model.model.layers.57.self_attn.kv_b_proj",
226
- "language_model.model.layers.57.self_attn.q_a_proj",
227
- "language_model.model.layers.57.self_attn.q_b_proj",
228
- "language_model.model.layers.58.self_attn.kv_a_proj_with_mqa",
229
- "language_model.model.layers.58.self_attn.kv_b_proj",
230
- "language_model.model.layers.58.self_attn.q_a_proj",
231
- "language_model.model.layers.58.self_attn.q_b_proj",
232
- "language_model.model.layers.59.self_attn.kv_a_proj_with_mqa",
233
- "language_model.model.layers.59.self_attn.kv_b_proj",
234
- "language_model.model.layers.59.self_attn.q_a_proj",
235
- "language_model.model.layers.59.self_attn.q_b_proj",
236
- "language_model.model.layers.6.self_attn.kv_a_proj_with_mqa",
237
- "language_model.model.layers.6.self_attn.kv_b_proj",
238
- "language_model.model.layers.6.self_attn.q_a_proj",
239
- "language_model.model.layers.6.self_attn.q_b_proj",
240
- "language_model.model.layers.60.self_attn.kv_a_proj_with_mqa",
241
- "language_model.model.layers.60.self_attn.kv_b_proj",
242
- "language_model.model.layers.60.self_attn.q_a_proj",
243
- "language_model.model.layers.60.self_attn.q_b_proj",
244
- "language_model.model.layers.7.self_attn.kv_a_proj_with_mqa",
245
- "language_model.model.layers.7.self_attn.kv_b_proj",
246
- "language_model.model.layers.7.self_attn.q_a_proj",
247
- "language_model.model.layers.7.self_attn.q_b_proj",
248
- "language_model.model.layers.8.self_attn.kv_a_proj_with_mqa",
249
- "language_model.model.layers.8.self_attn.kv_b_proj",
250
- "language_model.model.layers.8.self_attn.q_a_proj",
251
- "language_model.model.layers.8.self_attn.q_b_proj",
252
- "language_model.model.layers.9.self_attn.kv_a_proj_with_mqa",
253
- "language_model.model.layers.9.self_attn.kv_b_proj",
254
- "language_model.model.layers.9.self_attn.q_a_proj",
255
- "language_model.model.layers.9.self_attn.q_b_proj",
256
  "mm_projector*",
257
  "vision_tower*"
258
  ]
 
8
  "kv_cache_quant_algo": "FP8",
9
  "group_size": 16,
10
  "exclude_modules": [
11
+ "lm_head",
12
+ "model.layers.0.self_attn.kv_a_proj_with_mqa",
13
+ "model.layers.0.self_attn.kv_b_proj",
14
+ "model.layers.0.self_attn.q_a_proj",
15
+ "model.layers.0.self_attn.q_b_proj",
16
+ "model.layers.1.self_attn.kv_a_proj_with_mqa",
17
+ "model.layers.1.self_attn.kv_b_proj",
18
+ "model.layers.1.self_attn.q_a_proj",
19
+ "model.layers.1.self_attn.q_b_proj",
20
+ "model.layers.10.self_attn.kv_a_proj_with_mqa",
21
+ "model.layers.10.self_attn.kv_b_proj",
22
+ "model.layers.10.self_attn.q_a_proj",
23
+ "model.layers.10.self_attn.q_b_proj",
24
+ "model.layers.11.self_attn.kv_a_proj_with_mqa",
25
+ "model.layers.11.self_attn.kv_b_proj",
26
+ "model.layers.11.self_attn.q_a_proj",
27
+ "model.layers.11.self_attn.q_b_proj",
28
+ "model.layers.12.self_attn.kv_a_proj_with_mqa",
29
+ "model.layers.12.self_attn.kv_b_proj",
30
+ "model.layers.12.self_attn.q_a_proj",
31
+ "model.layers.12.self_attn.q_b_proj",
32
+ "model.layers.13.self_attn.kv_a_proj_with_mqa",
33
+ "model.layers.13.self_attn.kv_b_proj",
34
+ "model.layers.13.self_attn.q_a_proj",
35
+ "model.layers.13.self_attn.q_b_proj",
36
+ "model.layers.14.self_attn.kv_a_proj_with_mqa",
37
+ "model.layers.14.self_attn.kv_b_proj",
38
+ "model.layers.14.self_attn.q_a_proj",
39
+ "model.layers.14.self_attn.q_b_proj",
40
+ "model.layers.15.self_attn.kv_a_proj_with_mqa",
41
+ "model.layers.15.self_attn.kv_b_proj",
42
+ "model.layers.15.self_attn.q_a_proj",
43
+ "model.layers.15.self_attn.q_b_proj",
44
+ "model.layers.16.self_attn.kv_a_proj_with_mqa",
45
+ "model.layers.16.self_attn.kv_b_proj",
46
+ "model.layers.16.self_attn.q_a_proj",
47
+ "model.layers.16.self_attn.q_b_proj",
48
+ "model.layers.17.self_attn.kv_a_proj_with_mqa",
49
+ "model.layers.17.self_attn.kv_b_proj",
50
+ "model.layers.17.self_attn.q_a_proj",
51
+ "model.layers.17.self_attn.q_b_proj",
52
+ "model.layers.18.self_attn.kv_a_proj_with_mqa",
53
+ "model.layers.18.self_attn.kv_b_proj",
54
+ "model.layers.18.self_attn.q_a_proj",
55
+ "model.layers.18.self_attn.q_b_proj",
56
+ "model.layers.19.self_attn.kv_a_proj_with_mqa",
57
+ "model.layers.19.self_attn.kv_b_proj",
58
+ "model.layers.19.self_attn.q_a_proj",
59
+ "model.layers.19.self_attn.q_b_proj",
60
+ "model.layers.2.self_attn.kv_a_proj_with_mqa",
61
+ "model.layers.2.self_attn.kv_b_proj",
62
+ "model.layers.2.self_attn.q_a_proj",
63
+ "model.layers.2.self_attn.q_b_proj",
64
+ "model.layers.20.self_attn.kv_a_proj_with_mqa",
65
+ "model.layers.20.self_attn.kv_b_proj",
66
+ "model.layers.20.self_attn.q_a_proj",
67
+ "model.layers.20.self_attn.q_b_proj",
68
+ "model.layers.21.self_attn.kv_a_proj_with_mqa",
69
+ "model.layers.21.self_attn.kv_b_proj",
70
+ "model.layers.21.self_attn.q_a_proj",
71
+ "model.layers.21.self_attn.q_b_proj",
72
+ "model.layers.22.self_attn.kv_a_proj_with_mqa",
73
+ "model.layers.22.self_attn.kv_b_proj",
74
+ "model.layers.22.self_attn.q_a_proj",
75
+ "model.layers.22.self_attn.q_b_proj",
76
+ "model.layers.23.self_attn.kv_a_proj_with_mqa",
77
+ "model.layers.23.self_attn.kv_b_proj",
78
+ "model.layers.23.self_attn.q_a_proj",
79
+ "model.layers.23.self_attn.q_b_proj",
80
+ "model.layers.24.self_attn.kv_a_proj_with_mqa",
81
+ "model.layers.24.self_attn.kv_b_proj",
82
+ "model.layers.24.self_attn.q_a_proj",
83
+ "model.layers.24.self_attn.q_b_proj",
84
+ "model.layers.25.self_attn.kv_a_proj_with_mqa",
85
+ "model.layers.25.self_attn.kv_b_proj",
86
+ "model.layers.25.self_attn.q_a_proj",
87
+ "model.layers.25.self_attn.q_b_proj",
88
+ "model.layers.26.self_attn.kv_a_proj_with_mqa",
89
+ "model.layers.26.self_attn.kv_b_proj",
90
+ "model.layers.26.self_attn.q_a_proj",
91
+ "model.layers.26.self_attn.q_b_proj",
92
+ "model.layers.27.self_attn.kv_a_proj_with_mqa",
93
+ "model.layers.27.self_attn.kv_b_proj",
94
+ "model.layers.27.self_attn.q_a_proj",
95
+ "model.layers.27.self_attn.q_b_proj",
96
+ "model.layers.28.self_attn.kv_a_proj_with_mqa",
97
+ "model.layers.28.self_attn.kv_b_proj",
98
+ "model.layers.28.self_attn.q_a_proj",
99
+ "model.layers.28.self_attn.q_b_proj",
100
+ "model.layers.29.self_attn.kv_a_proj_with_mqa",
101
+ "model.layers.29.self_attn.kv_b_proj",
102
+ "model.layers.29.self_attn.q_a_proj",
103
+ "model.layers.29.self_attn.q_b_proj",
104
+ "model.layers.3.self_attn.kv_a_proj_with_mqa",
105
+ "model.layers.3.self_attn.kv_b_proj",
106
+ "model.layers.3.self_attn.q_a_proj",
107
+ "model.layers.3.self_attn.q_b_proj",
108
+ "model.layers.30.self_attn.kv_a_proj_with_mqa",
109
+ "model.layers.30.self_attn.kv_b_proj",
110
+ "model.layers.30.self_attn.q_a_proj",
111
+ "model.layers.30.self_attn.q_b_proj",
112
+ "model.layers.31.self_attn.kv_a_proj_with_mqa",
113
+ "model.layers.31.self_attn.kv_b_proj",
114
+ "model.layers.31.self_attn.q_a_proj",
115
+ "model.layers.31.self_attn.q_b_proj",
116
+ "model.layers.32.self_attn.kv_a_proj_with_mqa",
117
+ "model.layers.32.self_attn.kv_b_proj",
118
+ "model.layers.32.self_attn.q_a_proj",
119
+ "model.layers.32.self_attn.q_b_proj",
120
+ "model.layers.33.self_attn.kv_a_proj_with_mqa",
121
+ "model.layers.33.self_attn.kv_b_proj",
122
+ "model.layers.33.self_attn.q_a_proj",
123
+ "model.layers.33.self_attn.q_b_proj",
124
+ "model.layers.34.self_attn.kv_a_proj_with_mqa",
125
+ "model.layers.34.self_attn.kv_b_proj",
126
+ "model.layers.34.self_attn.q_a_proj",
127
+ "model.layers.34.self_attn.q_b_proj",
128
+ "model.layers.35.self_attn.kv_a_proj_with_mqa",
129
+ "model.layers.35.self_attn.kv_b_proj",
130
+ "model.layers.35.self_attn.q_a_proj",
131
+ "model.layers.35.self_attn.q_b_proj",
132
+ "model.layers.36.self_attn.kv_a_proj_with_mqa",
133
+ "model.layers.36.self_attn.kv_b_proj",
134
+ "model.layers.36.self_attn.q_a_proj",
135
+ "model.layers.36.self_attn.q_b_proj",
136
+ "model.layers.37.self_attn.kv_a_proj_with_mqa",
137
+ "model.layers.37.self_attn.kv_b_proj",
138
+ "model.layers.37.self_attn.q_a_proj",
139
+ "model.layers.37.self_attn.q_b_proj",
140
+ "model.layers.38.self_attn.kv_a_proj_with_mqa",
141
+ "model.layers.38.self_attn.kv_b_proj",
142
+ "model.layers.38.self_attn.q_a_proj",
143
+ "model.layers.38.self_attn.q_b_proj",
144
+ "model.layers.39.self_attn.kv_a_proj_with_mqa",
145
+ "model.layers.39.self_attn.kv_b_proj",
146
+ "model.layers.39.self_attn.q_a_proj",
147
+ "model.layers.39.self_attn.q_b_proj",
148
+ "model.layers.4.self_attn.kv_a_proj_with_mqa",
149
+ "model.layers.4.self_attn.kv_b_proj",
150
+ "model.layers.4.self_attn.q_a_proj",
151
+ "model.layers.4.self_attn.q_b_proj",
152
+ "model.layers.40.self_attn.kv_a_proj_with_mqa",
153
+ "model.layers.40.self_attn.kv_b_proj",
154
+ "model.layers.40.self_attn.q_a_proj",
155
+ "model.layers.40.self_attn.q_b_proj",
156
+ "model.layers.41.self_attn.kv_a_proj_with_mqa",
157
+ "model.layers.41.self_attn.kv_b_proj",
158
+ "model.layers.41.self_attn.q_a_proj",
159
+ "model.layers.41.self_attn.q_b_proj",
160
+ "model.layers.42.self_attn.kv_a_proj_with_mqa",
161
+ "model.layers.42.self_attn.kv_b_proj",
162
+ "model.layers.42.self_attn.q_a_proj",
163
+ "model.layers.42.self_attn.q_b_proj",
164
+ "model.layers.43.self_attn.kv_a_proj_with_mqa",
165
+ "model.layers.43.self_attn.kv_b_proj",
166
+ "model.layers.43.self_attn.q_a_proj",
167
+ "model.layers.43.self_attn.q_b_proj",
168
+ "model.layers.44.self_attn.kv_a_proj_with_mqa",
169
+ "model.layers.44.self_attn.kv_b_proj",
170
+ "model.layers.44.self_attn.q_a_proj",
171
+ "model.layers.44.self_attn.q_b_proj",
172
+ "model.layers.45.self_attn.kv_a_proj_with_mqa",
173
+ "model.layers.45.self_attn.kv_b_proj",
174
+ "model.layers.45.self_attn.q_a_proj",
175
+ "model.layers.45.self_attn.q_b_proj",
176
+ "model.layers.46.self_attn.kv_a_proj_with_mqa",
177
+ "model.layers.46.self_attn.kv_b_proj",
178
+ "model.layers.46.self_attn.q_a_proj",
179
+ "model.layers.46.self_attn.q_b_proj",
180
+ "model.layers.47.self_attn.kv_a_proj_with_mqa",
181
+ "model.layers.47.self_attn.kv_b_proj",
182
+ "model.layers.47.self_attn.q_a_proj",
183
+ "model.layers.47.self_attn.q_b_proj",
184
+ "model.layers.48.self_attn.kv_a_proj_with_mqa",
185
+ "model.layers.48.self_attn.kv_b_proj",
186
+ "model.layers.48.self_attn.q_a_proj",
187
+ "model.layers.48.self_attn.q_b_proj",
188
+ "model.layers.49.self_attn.kv_a_proj_with_mqa",
189
+ "model.layers.49.self_attn.kv_b_proj",
190
+ "model.layers.49.self_attn.q_a_proj",
191
+ "model.layers.49.self_attn.q_b_proj",
192
+ "model.layers.5.self_attn.kv_a_proj_with_mqa",
193
+ "model.layers.5.self_attn.kv_b_proj",
194
+ "model.layers.5.self_attn.q_a_proj",
195
+ "model.layers.5.self_attn.q_b_proj",
196
+ "model.layers.50.self_attn.kv_a_proj_with_mqa",
197
+ "model.layers.50.self_attn.kv_b_proj",
198
+ "model.layers.50.self_attn.q_a_proj",
199
+ "model.layers.50.self_attn.q_b_proj",
200
+ "model.layers.51.self_attn.kv_a_proj_with_mqa",
201
+ "model.layers.51.self_attn.kv_b_proj",
202
+ "model.layers.51.self_attn.q_a_proj",
203
+ "model.layers.51.self_attn.q_b_proj",
204
+ "model.layers.52.self_attn.kv_a_proj_with_mqa",
205
+ "model.layers.52.self_attn.kv_b_proj",
206
+ "model.layers.52.self_attn.q_a_proj",
207
+ "model.layers.52.self_attn.q_b_proj",
208
+ "model.layers.53.self_attn.kv_a_proj_with_mqa",
209
+ "model.layers.53.self_attn.kv_b_proj",
210
+ "model.layers.53.self_attn.q_a_proj",
211
+ "model.layers.53.self_attn.q_b_proj",
212
+ "model.layers.54.self_attn.kv_a_proj_with_mqa",
213
+ "model.layers.54.self_attn.kv_b_proj",
214
+ "model.layers.54.self_attn.q_a_proj",
215
+ "model.layers.54.self_attn.q_b_proj",
216
+ "model.layers.55.self_attn.kv_a_proj_with_mqa",
217
+ "model.layers.55.self_attn.kv_b_proj",
218
+ "model.layers.55.self_attn.q_a_proj",
219
+ "model.layers.55.self_attn.q_b_proj",
220
+ "model.layers.56.self_attn.kv_a_proj_with_mqa",
221
+ "model.layers.56.self_attn.kv_b_proj",
222
+ "model.layers.56.self_attn.q_a_proj",
223
+ "model.layers.56.self_attn.q_b_proj",
224
+ "model.layers.57.self_attn.kv_a_proj_with_mqa",
225
+ "model.layers.57.self_attn.kv_b_proj",
226
+ "model.layers.57.self_attn.q_a_proj",
227
+ "model.layers.57.self_attn.q_b_proj",
228
+ "model.layers.58.self_attn.kv_a_proj_with_mqa",
229
+ "model.layers.58.self_attn.kv_b_proj",
230
+ "model.layers.58.self_attn.q_a_proj",
231
+ "model.layers.58.self_attn.q_b_proj",
232
+ "model.layers.59.self_attn.kv_a_proj_with_mqa",
233
+ "model.layers.59.self_attn.kv_b_proj",
234
+ "model.layers.59.self_attn.q_a_proj",
235
+ "model.layers.59.self_attn.q_b_proj",
236
+ "model.layers.6.self_attn.kv_a_proj_with_mqa",
237
+ "model.layers.6.self_attn.kv_b_proj",
238
+ "model.layers.6.self_attn.q_a_proj",
239
+ "model.layers.6.self_attn.q_b_proj",
240
+ "model.layers.60.self_attn.kv_a_proj_with_mqa",
241
+ "model.layers.60.self_attn.kv_b_proj",
242
+ "model.layers.60.self_attn.q_a_proj",
243
+ "model.layers.60.self_attn.q_b_proj",
244
+ "model.layers.7.self_attn.kv_a_proj_with_mqa",
245
+ "model.layers.7.self_attn.kv_b_proj",
246
+ "model.layers.7.self_attn.q_a_proj",
247
+ "model.layers.7.self_attn.q_b_proj",
248
+ "model.layers.8.self_attn.kv_a_proj_with_mqa",
249
+ "model.layers.8.self_attn.kv_b_proj",
250
+ "model.layers.8.self_attn.q_a_proj",
251
+ "model.layers.8.self_attn.q_b_proj",
252
+ "model.layers.9.self_attn.kv_a_proj_with_mqa",
253
+ "model.layers.9.self_attn.kv_b_proj",
254
+ "model.layers.9.self_attn.q_a_proj",
255
+ "model.layers.9.self_attn.q_b_proj",
256
  "mm_projector*",
257
  "vision_tower*"
258
  ]