VincentGOURBIN commited on
Commit
d41b74a
·
verified ·
1 Parent(s): 5d4a4d7

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tekken.json filter=lfs diff=lfs merge=lfs -text
config.json ADDED
@@ -0,0 +1,1459 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "VoxtralForConditionalGeneration"
4
+ ],
5
+ "audio_config": {
6
+ "activation_dropout": 0.0,
7
+ "activation_function": "gelu",
8
+ "attention_dropout": 0.0,
9
+ "dropout": 0.0,
10
+ "head_dim": 64,
11
+ "hidden_size": 1280,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 5120,
14
+ "layerdrop": 0.0,
15
+ "max_source_positions": 1500,
16
+ "model_type": "voxtral_encoder",
17
+ "num_attention_heads": 20,
18
+ "num_hidden_layers": 32,
19
+ "num_key_value_heads": 20,
20
+ "num_mel_bins": 128,
21
+ "scale_embedding": false,
22
+ "vocab_size": 51866
23
+ },
24
+ "audio_token_id": 24,
25
+ "hidden_size": 5120,
26
+ "model_type": "voxtral",
27
+ "projector_hidden_act": "gelu",
28
+ "quantization": {
29
+ "group_size": 64,
30
+ "bits": 8,
31
+ "audio_tower.conv1": false,
32
+ "audio_tower.conv2": false,
33
+ "audio_tower.embed_positions": false,
34
+ "audio_tower.layers.0.self_attn.k_proj": true,
35
+ "audio_tower.layers.0.self_attn.v_proj": true,
36
+ "audio_tower.layers.0.self_attn.q_proj": true,
37
+ "audio_tower.layers.0.self_attn.out_proj": true,
38
+ "audio_tower.layers.0.self_attn_layer_norm": false,
39
+ "audio_tower.layers.0.activation_fn": false,
40
+ "audio_tower.layers.0.fc1": true,
41
+ "audio_tower.layers.0.fc2": true,
42
+ "audio_tower.layers.0.final_layer_norm": false,
43
+ "audio_tower.layers.1.self_attn.k_proj": true,
44
+ "audio_tower.layers.1.self_attn.v_proj": true,
45
+ "audio_tower.layers.1.self_attn.q_proj": true,
46
+ "audio_tower.layers.1.self_attn.out_proj": true,
47
+ "audio_tower.layers.1.self_attn_layer_norm": false,
48
+ "audio_tower.layers.1.activation_fn": false,
49
+ "audio_tower.layers.1.fc1": true,
50
+ "audio_tower.layers.1.fc2": true,
51
+ "audio_tower.layers.1.final_layer_norm": false,
52
+ "audio_tower.layers.2.self_attn.k_proj": true,
53
+ "audio_tower.layers.2.self_attn.v_proj": true,
54
+ "audio_tower.layers.2.self_attn.q_proj": true,
55
+ "audio_tower.layers.2.self_attn.out_proj": true,
56
+ "audio_tower.layers.2.self_attn_layer_norm": false,
57
+ "audio_tower.layers.2.activation_fn": false,
58
+ "audio_tower.layers.2.fc1": true,
59
+ "audio_tower.layers.2.fc2": true,
60
+ "audio_tower.layers.2.final_layer_norm": false,
61
+ "audio_tower.layers.3.self_attn.k_proj": true,
62
+ "audio_tower.layers.3.self_attn.v_proj": true,
63
+ "audio_tower.layers.3.self_attn.q_proj": true,
64
+ "audio_tower.layers.3.self_attn.out_proj": true,
65
+ "audio_tower.layers.3.self_attn_layer_norm": false,
66
+ "audio_tower.layers.3.activation_fn": false,
67
+ "audio_tower.layers.3.fc1": true,
68
+ "audio_tower.layers.3.fc2": true,
69
+ "audio_tower.layers.3.final_layer_norm": false,
70
+ "audio_tower.layers.4.self_attn.k_proj": true,
71
+ "audio_tower.layers.4.self_attn.v_proj": true,
72
+ "audio_tower.layers.4.self_attn.q_proj": true,
73
+ "audio_tower.layers.4.self_attn.out_proj": true,
74
+ "audio_tower.layers.4.self_attn_layer_norm": false,
75
+ "audio_tower.layers.4.activation_fn": false,
76
+ "audio_tower.layers.4.fc1": true,
77
+ "audio_tower.layers.4.fc2": true,
78
+ "audio_tower.layers.4.final_layer_norm": false,
79
+ "audio_tower.layers.5.self_attn.k_proj": true,
80
+ "audio_tower.layers.5.self_attn.v_proj": true,
81
+ "audio_tower.layers.5.self_attn.q_proj": true,
82
+ "audio_tower.layers.5.self_attn.out_proj": true,
83
+ "audio_tower.layers.5.self_attn_layer_norm": false,
84
+ "audio_tower.layers.5.activation_fn": false,
85
+ "audio_tower.layers.5.fc1": true,
86
+ "audio_tower.layers.5.fc2": true,
87
+ "audio_tower.layers.5.final_layer_norm": false,
88
+ "audio_tower.layers.6.self_attn.k_proj": true,
89
+ "audio_tower.layers.6.self_attn.v_proj": true,
90
+ "audio_tower.layers.6.self_attn.q_proj": true,
91
+ "audio_tower.layers.6.self_attn.out_proj": true,
92
+ "audio_tower.layers.6.self_attn_layer_norm": false,
93
+ "audio_tower.layers.6.activation_fn": false,
94
+ "audio_tower.layers.6.fc1": true,
95
+ "audio_tower.layers.6.fc2": true,
96
+ "audio_tower.layers.6.final_layer_norm": false,
97
+ "audio_tower.layers.7.self_attn.k_proj": true,
98
+ "audio_tower.layers.7.self_attn.v_proj": true,
99
+ "audio_tower.layers.7.self_attn.q_proj": true,
100
+ "audio_tower.layers.7.self_attn.out_proj": true,
101
+ "audio_tower.layers.7.self_attn_layer_norm": false,
102
+ "audio_tower.layers.7.activation_fn": false,
103
+ "audio_tower.layers.7.fc1": true,
104
+ "audio_tower.layers.7.fc2": true,
105
+ "audio_tower.layers.7.final_layer_norm": false,
106
+ "audio_tower.layers.8.self_attn.k_proj": true,
107
+ "audio_tower.layers.8.self_attn.v_proj": true,
108
+ "audio_tower.layers.8.self_attn.q_proj": true,
109
+ "audio_tower.layers.8.self_attn.out_proj": true,
110
+ "audio_tower.layers.8.self_attn_layer_norm": false,
111
+ "audio_tower.layers.8.activation_fn": false,
112
+ "audio_tower.layers.8.fc1": true,
113
+ "audio_tower.layers.8.fc2": true,
114
+ "audio_tower.layers.8.final_layer_norm": false,
115
+ "audio_tower.layers.9.self_attn.k_proj": true,
116
+ "audio_tower.layers.9.self_attn.v_proj": true,
117
+ "audio_tower.layers.9.self_attn.q_proj": true,
118
+ "audio_tower.layers.9.self_attn.out_proj": true,
119
+ "audio_tower.layers.9.self_attn_layer_norm": false,
120
+ "audio_tower.layers.9.activation_fn": false,
121
+ "audio_tower.layers.9.fc1": true,
122
+ "audio_tower.layers.9.fc2": true,
123
+ "audio_tower.layers.9.final_layer_norm": false,
124
+ "audio_tower.layers.10.self_attn.k_proj": true,
125
+ "audio_tower.layers.10.self_attn.v_proj": true,
126
+ "audio_tower.layers.10.self_attn.q_proj": true,
127
+ "audio_tower.layers.10.self_attn.out_proj": true,
128
+ "audio_tower.layers.10.self_attn_layer_norm": false,
129
+ "audio_tower.layers.10.activation_fn": false,
130
+ "audio_tower.layers.10.fc1": true,
131
+ "audio_tower.layers.10.fc2": true,
132
+ "audio_tower.layers.10.final_layer_norm": false,
133
+ "audio_tower.layers.11.self_attn.k_proj": true,
134
+ "audio_tower.layers.11.self_attn.v_proj": true,
135
+ "audio_tower.layers.11.self_attn.q_proj": true,
136
+ "audio_tower.layers.11.self_attn.out_proj": true,
137
+ "audio_tower.layers.11.self_attn_layer_norm": false,
138
+ "audio_tower.layers.11.activation_fn": false,
139
+ "audio_tower.layers.11.fc1": true,
140
+ "audio_tower.layers.11.fc2": true,
141
+ "audio_tower.layers.11.final_layer_norm": false,
142
+ "audio_tower.layers.12.self_attn.k_proj": true,
143
+ "audio_tower.layers.12.self_attn.v_proj": true,
144
+ "audio_tower.layers.12.self_attn.q_proj": true,
145
+ "audio_tower.layers.12.self_attn.out_proj": true,
146
+ "audio_tower.layers.12.self_attn_layer_norm": false,
147
+ "audio_tower.layers.12.activation_fn": false,
148
+ "audio_tower.layers.12.fc1": true,
149
+ "audio_tower.layers.12.fc2": true,
150
+ "audio_tower.layers.12.final_layer_norm": false,
151
+ "audio_tower.layers.13.self_attn.k_proj": true,
152
+ "audio_tower.layers.13.self_attn.v_proj": true,
153
+ "audio_tower.layers.13.self_attn.q_proj": true,
154
+ "audio_tower.layers.13.self_attn.out_proj": true,
155
+ "audio_tower.layers.13.self_attn_layer_norm": false,
156
+ "audio_tower.layers.13.activation_fn": false,
157
+ "audio_tower.layers.13.fc1": true,
158
+ "audio_tower.layers.13.fc2": true,
159
+ "audio_tower.layers.13.final_layer_norm": false,
160
+ "audio_tower.layers.14.self_attn.k_proj": true,
161
+ "audio_tower.layers.14.self_attn.v_proj": true,
162
+ "audio_tower.layers.14.self_attn.q_proj": true,
163
+ "audio_tower.layers.14.self_attn.out_proj": true,
164
+ "audio_tower.layers.14.self_attn_layer_norm": false,
165
+ "audio_tower.layers.14.activation_fn": false,
166
+ "audio_tower.layers.14.fc1": true,
167
+ "audio_tower.layers.14.fc2": true,
168
+ "audio_tower.layers.14.final_layer_norm": false,
169
+ "audio_tower.layers.15.self_attn.k_proj": true,
170
+ "audio_tower.layers.15.self_attn.v_proj": true,
171
+ "audio_tower.layers.15.self_attn.q_proj": true,
172
+ "audio_tower.layers.15.self_attn.out_proj": true,
173
+ "audio_tower.layers.15.self_attn_layer_norm": false,
174
+ "audio_tower.layers.15.activation_fn": false,
175
+ "audio_tower.layers.15.fc1": true,
176
+ "audio_tower.layers.15.fc2": true,
177
+ "audio_tower.layers.15.final_layer_norm": false,
178
+ "audio_tower.layers.16.self_attn.k_proj": true,
179
+ "audio_tower.layers.16.self_attn.v_proj": true,
180
+ "audio_tower.layers.16.self_attn.q_proj": true,
181
+ "audio_tower.layers.16.self_attn.out_proj": true,
182
+ "audio_tower.layers.16.self_attn_layer_norm": false,
183
+ "audio_tower.layers.16.activation_fn": false,
184
+ "audio_tower.layers.16.fc1": true,
185
+ "audio_tower.layers.16.fc2": true,
186
+ "audio_tower.layers.16.final_layer_norm": false,
187
+ "audio_tower.layers.17.self_attn.k_proj": true,
188
+ "audio_tower.layers.17.self_attn.v_proj": true,
189
+ "audio_tower.layers.17.self_attn.q_proj": true,
190
+ "audio_tower.layers.17.self_attn.out_proj": true,
191
+ "audio_tower.layers.17.self_attn_layer_norm": false,
192
+ "audio_tower.layers.17.activation_fn": false,
193
+ "audio_tower.layers.17.fc1": true,
194
+ "audio_tower.layers.17.fc2": true,
195
+ "audio_tower.layers.17.final_layer_norm": false,
196
+ "audio_tower.layers.18.self_attn.k_proj": true,
197
+ "audio_tower.layers.18.self_attn.v_proj": true,
198
+ "audio_tower.layers.18.self_attn.q_proj": true,
199
+ "audio_tower.layers.18.self_attn.out_proj": true,
200
+ "audio_tower.layers.18.self_attn_layer_norm": false,
201
+ "audio_tower.layers.18.activation_fn": false,
202
+ "audio_tower.layers.18.fc1": true,
203
+ "audio_tower.layers.18.fc2": true,
204
+ "audio_tower.layers.18.final_layer_norm": false,
205
+ "audio_tower.layers.19.self_attn.k_proj": true,
206
+ "audio_tower.layers.19.self_attn.v_proj": true,
207
+ "audio_tower.layers.19.self_attn.q_proj": true,
208
+ "audio_tower.layers.19.self_attn.out_proj": true,
209
+ "audio_tower.layers.19.self_attn_layer_norm": false,
210
+ "audio_tower.layers.19.activation_fn": false,
211
+ "audio_tower.layers.19.fc1": true,
212
+ "audio_tower.layers.19.fc2": true,
213
+ "audio_tower.layers.19.final_layer_norm": false,
214
+ "audio_tower.layers.20.self_attn.k_proj": true,
215
+ "audio_tower.layers.20.self_attn.v_proj": true,
216
+ "audio_tower.layers.20.self_attn.q_proj": true,
217
+ "audio_tower.layers.20.self_attn.out_proj": true,
218
+ "audio_tower.layers.20.self_attn_layer_norm": false,
219
+ "audio_tower.layers.20.activation_fn": false,
220
+ "audio_tower.layers.20.fc1": true,
221
+ "audio_tower.layers.20.fc2": true,
222
+ "audio_tower.layers.20.final_layer_norm": false,
223
+ "audio_tower.layers.21.self_attn.k_proj": true,
224
+ "audio_tower.layers.21.self_attn.v_proj": true,
225
+ "audio_tower.layers.21.self_attn.q_proj": true,
226
+ "audio_tower.layers.21.self_attn.out_proj": true,
227
+ "audio_tower.layers.21.self_attn_layer_norm": false,
228
+ "audio_tower.layers.21.activation_fn": false,
229
+ "audio_tower.layers.21.fc1": true,
230
+ "audio_tower.layers.21.fc2": true,
231
+ "audio_tower.layers.21.final_layer_norm": false,
232
+ "audio_tower.layers.22.self_attn.k_proj": true,
233
+ "audio_tower.layers.22.self_attn.v_proj": true,
234
+ "audio_tower.layers.22.self_attn.q_proj": true,
235
+ "audio_tower.layers.22.self_attn.out_proj": true,
236
+ "audio_tower.layers.22.self_attn_layer_norm": false,
237
+ "audio_tower.layers.22.activation_fn": false,
238
+ "audio_tower.layers.22.fc1": true,
239
+ "audio_tower.layers.22.fc2": true,
240
+ "audio_tower.layers.22.final_layer_norm": false,
241
+ "audio_tower.layers.23.self_attn.k_proj": true,
242
+ "audio_tower.layers.23.self_attn.v_proj": true,
243
+ "audio_tower.layers.23.self_attn.q_proj": true,
244
+ "audio_tower.layers.23.self_attn.out_proj": true,
245
+ "audio_tower.layers.23.self_attn_layer_norm": false,
246
+ "audio_tower.layers.23.activation_fn": false,
247
+ "audio_tower.layers.23.fc1": true,
248
+ "audio_tower.layers.23.fc2": true,
249
+ "audio_tower.layers.23.final_layer_norm": false,
250
+ "audio_tower.layers.24.self_attn.k_proj": true,
251
+ "audio_tower.layers.24.self_attn.v_proj": true,
252
+ "audio_tower.layers.24.self_attn.q_proj": true,
253
+ "audio_tower.layers.24.self_attn.out_proj": true,
254
+ "audio_tower.layers.24.self_attn_layer_norm": false,
255
+ "audio_tower.layers.24.activation_fn": false,
256
+ "audio_tower.layers.24.fc1": true,
257
+ "audio_tower.layers.24.fc2": true,
258
+ "audio_tower.layers.24.final_layer_norm": false,
259
+ "audio_tower.layers.25.self_attn.k_proj": true,
260
+ "audio_tower.layers.25.self_attn.v_proj": true,
261
+ "audio_tower.layers.25.self_attn.q_proj": true,
262
+ "audio_tower.layers.25.self_attn.out_proj": true,
263
+ "audio_tower.layers.25.self_attn_layer_norm": false,
264
+ "audio_tower.layers.25.activation_fn": false,
265
+ "audio_tower.layers.25.fc1": true,
266
+ "audio_tower.layers.25.fc2": true,
267
+ "audio_tower.layers.25.final_layer_norm": false,
268
+ "audio_tower.layers.26.self_attn.k_proj": true,
269
+ "audio_tower.layers.26.self_attn.v_proj": true,
270
+ "audio_tower.layers.26.self_attn.q_proj": true,
271
+ "audio_tower.layers.26.self_attn.out_proj": true,
272
+ "audio_tower.layers.26.self_attn_layer_norm": false,
273
+ "audio_tower.layers.26.activation_fn": false,
274
+ "audio_tower.layers.26.fc1": true,
275
+ "audio_tower.layers.26.fc2": true,
276
+ "audio_tower.layers.26.final_layer_norm": false,
277
+ "audio_tower.layers.27.self_attn.k_proj": true,
278
+ "audio_tower.layers.27.self_attn.v_proj": true,
279
+ "audio_tower.layers.27.self_attn.q_proj": true,
280
+ "audio_tower.layers.27.self_attn.out_proj": true,
281
+ "audio_tower.layers.27.self_attn_layer_norm": false,
282
+ "audio_tower.layers.27.activation_fn": false,
283
+ "audio_tower.layers.27.fc1": true,
284
+ "audio_tower.layers.27.fc2": true,
285
+ "audio_tower.layers.27.final_layer_norm": false,
286
+ "audio_tower.layers.28.self_attn.k_proj": true,
287
+ "audio_tower.layers.28.self_attn.v_proj": true,
288
+ "audio_tower.layers.28.self_attn.q_proj": true,
289
+ "audio_tower.layers.28.self_attn.out_proj": true,
290
+ "audio_tower.layers.28.self_attn_layer_norm": false,
291
+ "audio_tower.layers.28.activation_fn": false,
292
+ "audio_tower.layers.28.fc1": true,
293
+ "audio_tower.layers.28.fc2": true,
294
+ "audio_tower.layers.28.final_layer_norm": false,
295
+ "audio_tower.layers.29.self_attn.k_proj": true,
296
+ "audio_tower.layers.29.self_attn.v_proj": true,
297
+ "audio_tower.layers.29.self_attn.q_proj": true,
298
+ "audio_tower.layers.29.self_attn.out_proj": true,
299
+ "audio_tower.layers.29.self_attn_layer_norm": false,
300
+ "audio_tower.layers.29.activation_fn": false,
301
+ "audio_tower.layers.29.fc1": true,
302
+ "audio_tower.layers.29.fc2": true,
303
+ "audio_tower.layers.29.final_layer_norm": false,
304
+ "audio_tower.layers.30.self_attn.k_proj": true,
305
+ "audio_tower.layers.30.self_attn.v_proj": true,
306
+ "audio_tower.layers.30.self_attn.q_proj": true,
307
+ "audio_tower.layers.30.self_attn.out_proj": true,
308
+ "audio_tower.layers.30.self_attn_layer_norm": false,
309
+ "audio_tower.layers.30.activation_fn": false,
310
+ "audio_tower.layers.30.fc1": true,
311
+ "audio_tower.layers.30.fc2": true,
312
+ "audio_tower.layers.30.final_layer_norm": false,
313
+ "audio_tower.layers.31.self_attn.k_proj": true,
314
+ "audio_tower.layers.31.self_attn.v_proj": true,
315
+ "audio_tower.layers.31.self_attn.q_proj": true,
316
+ "audio_tower.layers.31.self_attn.out_proj": true,
317
+ "audio_tower.layers.31.self_attn_layer_norm": false,
318
+ "audio_tower.layers.31.activation_fn": false,
319
+ "audio_tower.layers.31.fc1": true,
320
+ "audio_tower.layers.31.fc2": true,
321
+ "audio_tower.layers.31.final_layer_norm": false,
322
+ "audio_tower.layer_norm": false,
323
+ "multi_modal_projector.linear_1": true,
324
+ "multi_modal_projector.act": false,
325
+ "multi_modal_projector.linear_2": true,
326
+ "language_model.embed_tokens": true,
327
+ "language_model.layers.0.self_attn.q_proj": true,
328
+ "language_model.layers.0.self_attn.k_proj": true,
329
+ "language_model.layers.0.self_attn.v_proj": true,
330
+ "language_model.layers.0.self_attn.o_proj": true,
331
+ "language_model.layers.0.self_attn.rope": false,
332
+ "language_model.layers.0.mlp.gate_proj": true,
333
+ "language_model.layers.0.mlp.up_proj": true,
334
+ "language_model.layers.0.mlp.down_proj": true,
335
+ "language_model.layers.0.input_layernorm": false,
336
+ "language_model.layers.0.post_attention_layernorm": false,
337
+ "language_model.layers.1.self_attn.q_proj": true,
338
+ "language_model.layers.1.self_attn.k_proj": true,
339
+ "language_model.layers.1.self_attn.v_proj": true,
340
+ "language_model.layers.1.self_attn.o_proj": true,
341
+ "language_model.layers.1.self_attn.rope": false,
342
+ "language_model.layers.1.mlp.gate_proj": true,
343
+ "language_model.layers.1.mlp.up_proj": true,
344
+ "language_model.layers.1.mlp.down_proj": true,
345
+ "language_model.layers.1.input_layernorm": false,
346
+ "language_model.layers.1.post_attention_layernorm": false,
347
+ "language_model.layers.2.self_attn.q_proj": true,
348
+ "language_model.layers.2.self_attn.k_proj": true,
349
+ "language_model.layers.2.self_attn.v_proj": true,
350
+ "language_model.layers.2.self_attn.o_proj": true,
351
+ "language_model.layers.2.self_attn.rope": false,
352
+ "language_model.layers.2.mlp.gate_proj": true,
353
+ "language_model.layers.2.mlp.up_proj": true,
354
+ "language_model.layers.2.mlp.down_proj": true,
355
+ "language_model.layers.2.input_layernorm": false,
356
+ "language_model.layers.2.post_attention_layernorm": false,
357
+ "language_model.layers.3.self_attn.q_proj": true,
358
+ "language_model.layers.3.self_attn.k_proj": true,
359
+ "language_model.layers.3.self_attn.v_proj": true,
360
+ "language_model.layers.3.self_attn.o_proj": true,
361
+ "language_model.layers.3.self_attn.rope": false,
362
+ "language_model.layers.3.mlp.gate_proj": true,
363
+ "language_model.layers.3.mlp.up_proj": true,
364
+ "language_model.layers.3.mlp.down_proj": true,
365
+ "language_model.layers.3.input_layernorm": false,
366
+ "language_model.layers.3.post_attention_layernorm": false,
367
+ "language_model.layers.4.self_attn.q_proj": true,
368
+ "language_model.layers.4.self_attn.k_proj": true,
369
+ "language_model.layers.4.self_attn.v_proj": true,
370
+ "language_model.layers.4.self_attn.o_proj": true,
371
+ "language_model.layers.4.self_attn.rope": false,
372
+ "language_model.layers.4.mlp.gate_proj": true,
373
+ "language_model.layers.4.mlp.up_proj": true,
374
+ "language_model.layers.4.mlp.down_proj": true,
375
+ "language_model.layers.4.input_layernorm": false,
376
+ "language_model.layers.4.post_attention_layernorm": false,
377
+ "language_model.layers.5.self_attn.q_proj": true,
378
+ "language_model.layers.5.self_attn.k_proj": true,
379
+ "language_model.layers.5.self_attn.v_proj": true,
380
+ "language_model.layers.5.self_attn.o_proj": true,
381
+ "language_model.layers.5.self_attn.rope": false,
382
+ "language_model.layers.5.mlp.gate_proj": true,
383
+ "language_model.layers.5.mlp.up_proj": true,
384
+ "language_model.layers.5.mlp.down_proj": true,
385
+ "language_model.layers.5.input_layernorm": false,
386
+ "language_model.layers.5.post_attention_layernorm": false,
387
+ "language_model.layers.6.self_attn.q_proj": true,
388
+ "language_model.layers.6.self_attn.k_proj": true,
389
+ "language_model.layers.6.self_attn.v_proj": true,
390
+ "language_model.layers.6.self_attn.o_proj": true,
391
+ "language_model.layers.6.self_attn.rope": false,
392
+ "language_model.layers.6.mlp.gate_proj": true,
393
+ "language_model.layers.6.mlp.up_proj": true,
394
+ "language_model.layers.6.mlp.down_proj": true,
395
+ "language_model.layers.6.input_layernorm": false,
396
+ "language_model.layers.6.post_attention_layernorm": false,
397
+ "language_model.layers.7.self_attn.q_proj": true,
398
+ "language_model.layers.7.self_attn.k_proj": true,
399
+ "language_model.layers.7.self_attn.v_proj": true,
400
+ "language_model.layers.7.self_attn.o_proj": true,
401
+ "language_model.layers.7.self_attn.rope": false,
402
+ "language_model.layers.7.mlp.gate_proj": true,
403
+ "language_model.layers.7.mlp.up_proj": true,
404
+ "language_model.layers.7.mlp.down_proj": true,
405
+ "language_model.layers.7.input_layernorm": false,
406
+ "language_model.layers.7.post_attention_layernorm": false,
407
+ "language_model.layers.8.self_attn.q_proj": true,
408
+ "language_model.layers.8.self_attn.k_proj": true,
409
+ "language_model.layers.8.self_attn.v_proj": true,
410
+ "language_model.layers.8.self_attn.o_proj": true,
411
+ "language_model.layers.8.self_attn.rope": false,
412
+ "language_model.layers.8.mlp.gate_proj": true,
413
+ "language_model.layers.8.mlp.up_proj": true,
414
+ "language_model.layers.8.mlp.down_proj": true,
415
+ "language_model.layers.8.input_layernorm": false,
416
+ "language_model.layers.8.post_attention_layernorm": false,
417
+ "language_model.layers.9.self_attn.q_proj": true,
418
+ "language_model.layers.9.self_attn.k_proj": true,
419
+ "language_model.layers.9.self_attn.v_proj": true,
420
+ "language_model.layers.9.self_attn.o_proj": true,
421
+ "language_model.layers.9.self_attn.rope": false,
422
+ "language_model.layers.9.mlp.gate_proj": true,
423
+ "language_model.layers.9.mlp.up_proj": true,
424
+ "language_model.layers.9.mlp.down_proj": true,
425
+ "language_model.layers.9.input_layernorm": false,
426
+ "language_model.layers.9.post_attention_layernorm": false,
427
+ "language_model.layers.10.self_attn.q_proj": true,
428
+ "language_model.layers.10.self_attn.k_proj": true,
429
+ "language_model.layers.10.self_attn.v_proj": true,
430
+ "language_model.layers.10.self_attn.o_proj": true,
431
+ "language_model.layers.10.self_attn.rope": false,
432
+ "language_model.layers.10.mlp.gate_proj": true,
433
+ "language_model.layers.10.mlp.up_proj": true,
434
+ "language_model.layers.10.mlp.down_proj": true,
435
+ "language_model.layers.10.input_layernorm": false,
436
+ "language_model.layers.10.post_attention_layernorm": false,
437
+ "language_model.layers.11.self_attn.q_proj": true,
438
+ "language_model.layers.11.self_attn.k_proj": true,
439
+ "language_model.layers.11.self_attn.v_proj": true,
440
+ "language_model.layers.11.self_attn.o_proj": true,
441
+ "language_model.layers.11.self_attn.rope": false,
442
+ "language_model.layers.11.mlp.gate_proj": true,
443
+ "language_model.layers.11.mlp.up_proj": true,
444
+ "language_model.layers.11.mlp.down_proj": true,
445
+ "language_model.layers.11.input_layernorm": false,
446
+ "language_model.layers.11.post_attention_layernorm": false,
447
+ "language_model.layers.12.self_attn.q_proj": true,
448
+ "language_model.layers.12.self_attn.k_proj": true,
449
+ "language_model.layers.12.self_attn.v_proj": true,
450
+ "language_model.layers.12.self_attn.o_proj": true,
451
+ "language_model.layers.12.self_attn.rope": false,
452
+ "language_model.layers.12.mlp.gate_proj": true,
453
+ "language_model.layers.12.mlp.up_proj": true,
454
+ "language_model.layers.12.mlp.down_proj": true,
455
+ "language_model.layers.12.input_layernorm": false,
456
+ "language_model.layers.12.post_attention_layernorm": false,
457
+ "language_model.layers.13.self_attn.q_proj": true,
458
+ "language_model.layers.13.self_attn.k_proj": true,
459
+ "language_model.layers.13.self_attn.v_proj": true,
460
+ "language_model.layers.13.self_attn.o_proj": true,
461
+ "language_model.layers.13.self_attn.rope": false,
462
+ "language_model.layers.13.mlp.gate_proj": true,
463
+ "language_model.layers.13.mlp.up_proj": true,
464
+ "language_model.layers.13.mlp.down_proj": true,
465
+ "language_model.layers.13.input_layernorm": false,
466
+ "language_model.layers.13.post_attention_layernorm": false,
467
+ "language_model.layers.14.self_attn.q_proj": true,
468
+ "language_model.layers.14.self_attn.k_proj": true,
469
+ "language_model.layers.14.self_attn.v_proj": true,
470
+ "language_model.layers.14.self_attn.o_proj": true,
471
+ "language_model.layers.14.self_attn.rope": false,
472
+ "language_model.layers.14.mlp.gate_proj": true,
473
+ "language_model.layers.14.mlp.up_proj": true,
474
+ "language_model.layers.14.mlp.down_proj": true,
475
+ "language_model.layers.14.input_layernorm": false,
476
+ "language_model.layers.14.post_attention_layernorm": false,
477
+ "language_model.layers.15.self_attn.q_proj": true,
478
+ "language_model.layers.15.self_attn.k_proj": true,
479
+ "language_model.layers.15.self_attn.v_proj": true,
480
+ "language_model.layers.15.self_attn.o_proj": true,
481
+ "language_model.layers.15.self_attn.rope": false,
482
+ "language_model.layers.15.mlp.gate_proj": true,
483
+ "language_model.layers.15.mlp.up_proj": true,
484
+ "language_model.layers.15.mlp.down_proj": true,
485
+ "language_model.layers.15.input_layernorm": false,
486
+ "language_model.layers.15.post_attention_layernorm": false,
487
+ "language_model.layers.16.self_attn.q_proj": true,
488
+ "language_model.layers.16.self_attn.k_proj": true,
489
+ "language_model.layers.16.self_attn.v_proj": true,
490
+ "language_model.layers.16.self_attn.o_proj": true,
491
+ "language_model.layers.16.self_attn.rope": false,
492
+ "language_model.layers.16.mlp.gate_proj": true,
493
+ "language_model.layers.16.mlp.up_proj": true,
494
+ "language_model.layers.16.mlp.down_proj": true,
495
+ "language_model.layers.16.input_layernorm": false,
496
+ "language_model.layers.16.post_attention_layernorm": false,
497
+ "language_model.layers.17.self_attn.q_proj": true,
498
+ "language_model.layers.17.self_attn.k_proj": true,
499
+ "language_model.layers.17.self_attn.v_proj": true,
500
+ "language_model.layers.17.self_attn.o_proj": true,
501
+ "language_model.layers.17.self_attn.rope": false,
502
+ "language_model.layers.17.mlp.gate_proj": true,
503
+ "language_model.layers.17.mlp.up_proj": true,
504
+ "language_model.layers.17.mlp.down_proj": true,
505
+ "language_model.layers.17.input_layernorm": false,
506
+ "language_model.layers.17.post_attention_layernorm": false,
507
+ "language_model.layers.18.self_attn.q_proj": true,
508
+ "language_model.layers.18.self_attn.k_proj": true,
509
+ "language_model.layers.18.self_attn.v_proj": true,
510
+ "language_model.layers.18.self_attn.o_proj": true,
511
+ "language_model.layers.18.self_attn.rope": false,
512
+ "language_model.layers.18.mlp.gate_proj": true,
513
+ "language_model.layers.18.mlp.up_proj": true,
514
+ "language_model.layers.18.mlp.down_proj": true,
515
+ "language_model.layers.18.input_layernorm": false,
516
+ "language_model.layers.18.post_attention_layernorm": false,
517
+ "language_model.layers.19.self_attn.q_proj": true,
518
+ "language_model.layers.19.self_attn.k_proj": true,
519
+ "language_model.layers.19.self_attn.v_proj": true,
520
+ "language_model.layers.19.self_attn.o_proj": true,
521
+ "language_model.layers.19.self_attn.rope": false,
522
+ "language_model.layers.19.mlp.gate_proj": true,
523
+ "language_model.layers.19.mlp.up_proj": true,
524
+ "language_model.layers.19.mlp.down_proj": true,
525
+ "language_model.layers.19.input_layernorm": false,
526
+ "language_model.layers.19.post_attention_layernorm": false,
527
+ "language_model.layers.20.self_attn.q_proj": true,
528
+ "language_model.layers.20.self_attn.k_proj": true,
529
+ "language_model.layers.20.self_attn.v_proj": true,
530
+ "language_model.layers.20.self_attn.o_proj": true,
531
+ "language_model.layers.20.self_attn.rope": false,
532
+ "language_model.layers.20.mlp.gate_proj": true,
533
+ "language_model.layers.20.mlp.up_proj": true,
534
+ "language_model.layers.20.mlp.down_proj": true,
535
+ "language_model.layers.20.input_layernorm": false,
536
+ "language_model.layers.20.post_attention_layernorm": false,
537
+ "language_model.layers.21.self_attn.q_proj": true,
538
+ "language_model.layers.21.self_attn.k_proj": true,
539
+ "language_model.layers.21.self_attn.v_proj": true,
540
+ "language_model.layers.21.self_attn.o_proj": true,
541
+ "language_model.layers.21.self_attn.rope": false,
542
+ "language_model.layers.21.mlp.gate_proj": true,
543
+ "language_model.layers.21.mlp.up_proj": true,
544
+ "language_model.layers.21.mlp.down_proj": true,
545
+ "language_model.layers.21.input_layernorm": false,
546
+ "language_model.layers.21.post_attention_layernorm": false,
547
+ "language_model.layers.22.self_attn.q_proj": true,
548
+ "language_model.layers.22.self_attn.k_proj": true,
549
+ "language_model.layers.22.self_attn.v_proj": true,
550
+ "language_model.layers.22.self_attn.o_proj": true,
551
+ "language_model.layers.22.self_attn.rope": false,
552
+ "language_model.layers.22.mlp.gate_proj": true,
553
+ "language_model.layers.22.mlp.up_proj": true,
554
+ "language_model.layers.22.mlp.down_proj": true,
555
+ "language_model.layers.22.input_layernorm": false,
556
+ "language_model.layers.22.post_attention_layernorm": false,
557
+ "language_model.layers.23.self_attn.q_proj": true,
558
+ "language_model.layers.23.self_attn.k_proj": true,
559
+ "language_model.layers.23.self_attn.v_proj": true,
560
+ "language_model.layers.23.self_attn.o_proj": true,
561
+ "language_model.layers.23.self_attn.rope": false,
562
+ "language_model.layers.23.mlp.gate_proj": true,
563
+ "language_model.layers.23.mlp.up_proj": true,
564
+ "language_model.layers.23.mlp.down_proj": true,
565
+ "language_model.layers.23.input_layernorm": false,
566
+ "language_model.layers.23.post_attention_layernorm": false,
567
+ "language_model.layers.24.self_attn.q_proj": true,
568
+ "language_model.layers.24.self_attn.k_proj": true,
569
+ "language_model.layers.24.self_attn.v_proj": true,
570
+ "language_model.layers.24.self_attn.o_proj": true,
571
+ "language_model.layers.24.self_attn.rope": false,
572
+ "language_model.layers.24.mlp.gate_proj": true,
573
+ "language_model.layers.24.mlp.up_proj": true,
574
+ "language_model.layers.24.mlp.down_proj": true,
575
+ "language_model.layers.24.input_layernorm": false,
576
+ "language_model.layers.24.post_attention_layernorm": false,
577
+ "language_model.layers.25.self_attn.q_proj": true,
578
+ "language_model.layers.25.self_attn.k_proj": true,
579
+ "language_model.layers.25.self_attn.v_proj": true,
580
+ "language_model.layers.25.self_attn.o_proj": true,
581
+ "language_model.layers.25.self_attn.rope": false,
582
+ "language_model.layers.25.mlp.gate_proj": true,
583
+ "language_model.layers.25.mlp.up_proj": true,
584
+ "language_model.layers.25.mlp.down_proj": true,
585
+ "language_model.layers.25.input_layernorm": false,
586
+ "language_model.layers.25.post_attention_layernorm": false,
587
+ "language_model.layers.26.self_attn.q_proj": true,
588
+ "language_model.layers.26.self_attn.k_proj": true,
589
+ "language_model.layers.26.self_attn.v_proj": true,
590
+ "language_model.layers.26.self_attn.o_proj": true,
591
+ "language_model.layers.26.self_attn.rope": false,
592
+ "language_model.layers.26.mlp.gate_proj": true,
593
+ "language_model.layers.26.mlp.up_proj": true,
594
+ "language_model.layers.26.mlp.down_proj": true,
595
+ "language_model.layers.26.input_layernorm": false,
596
+ "language_model.layers.26.post_attention_layernorm": false,
597
+ "language_model.layers.27.self_attn.q_proj": true,
598
+ "language_model.layers.27.self_attn.k_proj": true,
599
+ "language_model.layers.27.self_attn.v_proj": true,
600
+ "language_model.layers.27.self_attn.o_proj": true,
601
+ "language_model.layers.27.self_attn.rope": false,
602
+ "language_model.layers.27.mlp.gate_proj": true,
603
+ "language_model.layers.27.mlp.up_proj": true,
604
+ "language_model.layers.27.mlp.down_proj": true,
605
+ "language_model.layers.27.input_layernorm": false,
606
+ "language_model.layers.27.post_attention_layernorm": false,
607
+ "language_model.layers.28.self_attn.q_proj": true,
608
+ "language_model.layers.28.self_attn.k_proj": true,
609
+ "language_model.layers.28.self_attn.v_proj": true,
610
+ "language_model.layers.28.self_attn.o_proj": true,
611
+ "language_model.layers.28.self_attn.rope": false,
612
+ "language_model.layers.28.mlp.gate_proj": true,
613
+ "language_model.layers.28.mlp.up_proj": true,
614
+ "language_model.layers.28.mlp.down_proj": true,
615
+ "language_model.layers.28.input_layernorm": false,
616
+ "language_model.layers.28.post_attention_layernorm": false,
617
+ "language_model.layers.29.self_attn.q_proj": true,
618
+ "language_model.layers.29.self_attn.k_proj": true,
619
+ "language_model.layers.29.self_attn.v_proj": true,
620
+ "language_model.layers.29.self_attn.o_proj": true,
621
+ "language_model.layers.29.self_attn.rope": false,
622
+ "language_model.layers.29.mlp.gate_proj": true,
623
+ "language_model.layers.29.mlp.up_proj": true,
624
+ "language_model.layers.29.mlp.down_proj": true,
625
+ "language_model.layers.29.input_layernorm": false,
626
+ "language_model.layers.29.post_attention_layernorm": false,
627
+ "language_model.layers.30.self_attn.q_proj": true,
628
+ "language_model.layers.30.self_attn.k_proj": true,
629
+ "language_model.layers.30.self_attn.v_proj": true,
630
+ "language_model.layers.30.self_attn.o_proj": true,
631
+ "language_model.layers.30.self_attn.rope": false,
632
+ "language_model.layers.30.mlp.gate_proj": true,
633
+ "language_model.layers.30.mlp.up_proj": true,
634
+ "language_model.layers.30.mlp.down_proj": true,
635
+ "language_model.layers.30.input_layernorm": false,
636
+ "language_model.layers.30.post_attention_layernorm": false,
637
+ "language_model.layers.31.self_attn.q_proj": true,
638
+ "language_model.layers.31.self_attn.k_proj": true,
639
+ "language_model.layers.31.self_attn.v_proj": true,
640
+ "language_model.layers.31.self_attn.o_proj": true,
641
+ "language_model.layers.31.self_attn.rope": false,
642
+ "language_model.layers.31.mlp.gate_proj": true,
643
+ "language_model.layers.31.mlp.up_proj": true,
644
+ "language_model.layers.31.mlp.down_proj": true,
645
+ "language_model.layers.31.input_layernorm": false,
646
+ "language_model.layers.31.post_attention_layernorm": false,
647
+ "language_model.layers.32.self_attn.q_proj": true,
648
+ "language_model.layers.32.self_attn.k_proj": true,
649
+ "language_model.layers.32.self_attn.v_proj": true,
650
+ "language_model.layers.32.self_attn.o_proj": true,
651
+ "language_model.layers.32.self_attn.rope": false,
652
+ "language_model.layers.32.mlp.gate_proj": true,
653
+ "language_model.layers.32.mlp.up_proj": true,
654
+ "language_model.layers.32.mlp.down_proj": true,
655
+ "language_model.layers.32.input_layernorm": false,
656
+ "language_model.layers.32.post_attention_layernorm": false,
657
+ "language_model.layers.33.self_attn.q_proj": true,
658
+ "language_model.layers.33.self_attn.k_proj": true,
659
+ "language_model.layers.33.self_attn.v_proj": true,
660
+ "language_model.layers.33.self_attn.o_proj": true,
661
+ "language_model.layers.33.self_attn.rope": false,
662
+ "language_model.layers.33.mlp.gate_proj": true,
663
+ "language_model.layers.33.mlp.up_proj": true,
664
+ "language_model.layers.33.mlp.down_proj": true,
665
+ "language_model.layers.33.input_layernorm": false,
666
+ "language_model.layers.33.post_attention_layernorm": false,
667
+ "language_model.layers.34.self_attn.q_proj": true,
668
+ "language_model.layers.34.self_attn.k_proj": true,
669
+ "language_model.layers.34.self_attn.v_proj": true,
670
+ "language_model.layers.34.self_attn.o_proj": true,
671
+ "language_model.layers.34.self_attn.rope": false,
672
+ "language_model.layers.34.mlp.gate_proj": true,
673
+ "language_model.layers.34.mlp.up_proj": true,
674
+ "language_model.layers.34.mlp.down_proj": true,
675
+ "language_model.layers.34.input_layernorm": false,
676
+ "language_model.layers.34.post_attention_layernorm": false,
677
+ "language_model.layers.35.self_attn.q_proj": true,
678
+ "language_model.layers.35.self_attn.k_proj": true,
679
+ "language_model.layers.35.self_attn.v_proj": true,
680
+ "language_model.layers.35.self_attn.o_proj": true,
681
+ "language_model.layers.35.self_attn.rope": false,
682
+ "language_model.layers.35.mlp.gate_proj": true,
683
+ "language_model.layers.35.mlp.up_proj": true,
684
+ "language_model.layers.35.mlp.down_proj": true,
685
+ "language_model.layers.35.input_layernorm": false,
686
+ "language_model.layers.35.post_attention_layernorm": false,
687
+ "language_model.layers.36.self_attn.q_proj": true,
688
+ "language_model.layers.36.self_attn.k_proj": true,
689
+ "language_model.layers.36.self_attn.v_proj": true,
690
+ "language_model.layers.36.self_attn.o_proj": true,
691
+ "language_model.layers.36.self_attn.rope": false,
692
+ "language_model.layers.36.mlp.gate_proj": true,
693
+ "language_model.layers.36.mlp.up_proj": true,
694
+ "language_model.layers.36.mlp.down_proj": true,
695
+ "language_model.layers.36.input_layernorm": false,
696
+ "language_model.layers.36.post_attention_layernorm": false,
697
+ "language_model.layers.37.self_attn.q_proj": true,
698
+ "language_model.layers.37.self_attn.k_proj": true,
699
+ "language_model.layers.37.self_attn.v_proj": true,
700
+ "language_model.layers.37.self_attn.o_proj": true,
701
+ "language_model.layers.37.self_attn.rope": false,
702
+ "language_model.layers.37.mlp.gate_proj": true,
703
+ "language_model.layers.37.mlp.up_proj": true,
704
+ "language_model.layers.37.mlp.down_proj": true,
705
+ "language_model.layers.37.input_layernorm": false,
706
+ "language_model.layers.37.post_attention_layernorm": false,
707
+ "language_model.layers.38.self_attn.q_proj": true,
708
+ "language_model.layers.38.self_attn.k_proj": true,
709
+ "language_model.layers.38.self_attn.v_proj": true,
710
+ "language_model.layers.38.self_attn.o_proj": true,
711
+ "language_model.layers.38.self_attn.rope": false,
712
+ "language_model.layers.38.mlp.gate_proj": true,
713
+ "language_model.layers.38.mlp.up_proj": true,
714
+ "language_model.layers.38.mlp.down_proj": true,
715
+ "language_model.layers.38.input_layernorm": false,
716
+ "language_model.layers.38.post_attention_layernorm": false,
717
+ "language_model.layers.39.self_attn.q_proj": true,
718
+ "language_model.layers.39.self_attn.k_proj": true,
719
+ "language_model.layers.39.self_attn.v_proj": true,
720
+ "language_model.layers.39.self_attn.o_proj": true,
721
+ "language_model.layers.39.self_attn.rope": false,
722
+ "language_model.layers.39.mlp.gate_proj": true,
723
+ "language_model.layers.39.mlp.up_proj": true,
724
+ "language_model.layers.39.mlp.down_proj": true,
725
+ "language_model.layers.39.input_layernorm": false,
726
+ "language_model.layers.39.post_attention_layernorm": false,
727
+ "language_model.norm": false,
728
+ "embed_tokens": true,
729
+ "lm_head": true
730
+ },
731
+ "quantization_config": {
732
+ "group_size": 64,
733
+ "bits": 8,
734
+ "audio_tower.conv1": false,
735
+ "audio_tower.conv2": false,
736
+ "audio_tower.embed_positions": false,
737
+ "audio_tower.layers.0.self_attn.k_proj": true,
738
+ "audio_tower.layers.0.self_attn.v_proj": true,
739
+ "audio_tower.layers.0.self_attn.q_proj": true,
740
+ "audio_tower.layers.0.self_attn.out_proj": true,
741
+ "audio_tower.layers.0.self_attn_layer_norm": false,
742
+ "audio_tower.layers.0.activation_fn": false,
743
+ "audio_tower.layers.0.fc1": true,
744
+ "audio_tower.layers.0.fc2": true,
745
+ "audio_tower.layers.0.final_layer_norm": false,
746
+ "audio_tower.layers.1.self_attn.k_proj": true,
747
+ "audio_tower.layers.1.self_attn.v_proj": true,
748
+ "audio_tower.layers.1.self_attn.q_proj": true,
749
+ "audio_tower.layers.1.self_attn.out_proj": true,
750
+ "audio_tower.layers.1.self_attn_layer_norm": false,
751
+ "audio_tower.layers.1.activation_fn": false,
752
+ "audio_tower.layers.1.fc1": true,
753
+ "audio_tower.layers.1.fc2": true,
754
+ "audio_tower.layers.1.final_layer_norm": false,
755
+ "audio_tower.layers.2.self_attn.k_proj": true,
756
+ "audio_tower.layers.2.self_attn.v_proj": true,
757
+ "audio_tower.layers.2.self_attn.q_proj": true,
758
+ "audio_tower.layers.2.self_attn.out_proj": true,
759
+ "audio_tower.layers.2.self_attn_layer_norm": false,
760
+ "audio_tower.layers.2.activation_fn": false,
761
+ "audio_tower.layers.2.fc1": true,
762
+ "audio_tower.layers.2.fc2": true,
763
+ "audio_tower.layers.2.final_layer_norm": false,
764
+ "audio_tower.layers.3.self_attn.k_proj": true,
765
+ "audio_tower.layers.3.self_attn.v_proj": true,
766
+ "audio_tower.layers.3.self_attn.q_proj": true,
767
+ "audio_tower.layers.3.self_attn.out_proj": true,
768
+ "audio_tower.layers.3.self_attn_layer_norm": false,
769
+ "audio_tower.layers.3.activation_fn": false,
770
+ "audio_tower.layers.3.fc1": true,
771
+ "audio_tower.layers.3.fc2": true,
772
+ "audio_tower.layers.3.final_layer_norm": false,
773
+ "audio_tower.layers.4.self_attn.k_proj": true,
774
+ "audio_tower.layers.4.self_attn.v_proj": true,
775
+ "audio_tower.layers.4.self_attn.q_proj": true,
776
+ "audio_tower.layers.4.self_attn.out_proj": true,
777
+ "audio_tower.layers.4.self_attn_layer_norm": false,
778
+ "audio_tower.layers.4.activation_fn": false,
779
+ "audio_tower.layers.4.fc1": true,
780
+ "audio_tower.layers.4.fc2": true,
781
+ "audio_tower.layers.4.final_layer_norm": false,
782
+ "audio_tower.layers.5.self_attn.k_proj": true,
783
+ "audio_tower.layers.5.self_attn.v_proj": true,
784
+ "audio_tower.layers.5.self_attn.q_proj": true,
785
+ "audio_tower.layers.5.self_attn.out_proj": true,
786
+ "audio_tower.layers.5.self_attn_layer_norm": false,
787
+ "audio_tower.layers.5.activation_fn": false,
788
+ "audio_tower.layers.5.fc1": true,
789
+ "audio_tower.layers.5.fc2": true,
790
+ "audio_tower.layers.5.final_layer_norm": false,
791
+ "audio_tower.layers.6.self_attn.k_proj": true,
792
+ "audio_tower.layers.6.self_attn.v_proj": true,
793
+ "audio_tower.layers.6.self_attn.q_proj": true,
794
+ "audio_tower.layers.6.self_attn.out_proj": true,
795
+ "audio_tower.layers.6.self_attn_layer_norm": false,
796
+ "audio_tower.layers.6.activation_fn": false,
797
+ "audio_tower.layers.6.fc1": true,
798
+ "audio_tower.layers.6.fc2": true,
799
+ "audio_tower.layers.6.final_layer_norm": false,
800
+ "audio_tower.layers.7.self_attn.k_proj": true,
801
+ "audio_tower.layers.7.self_attn.v_proj": true,
802
+ "audio_tower.layers.7.self_attn.q_proj": true,
803
+ "audio_tower.layers.7.self_attn.out_proj": true,
804
+ "audio_tower.layers.7.self_attn_layer_norm": false,
805
+ "audio_tower.layers.7.activation_fn": false,
806
+ "audio_tower.layers.7.fc1": true,
807
+ "audio_tower.layers.7.fc2": true,
808
+ "audio_tower.layers.7.final_layer_norm": false,
809
+ "audio_tower.layers.8.self_attn.k_proj": true,
810
+ "audio_tower.layers.8.self_attn.v_proj": true,
811
+ "audio_tower.layers.8.self_attn.q_proj": true,
812
+ "audio_tower.layers.8.self_attn.out_proj": true,
813
+ "audio_tower.layers.8.self_attn_layer_norm": false,
814
+ "audio_tower.layers.8.activation_fn": false,
815
+ "audio_tower.layers.8.fc1": true,
816
+ "audio_tower.layers.8.fc2": true,
817
+ "audio_tower.layers.8.final_layer_norm": false,
818
+ "audio_tower.layers.9.self_attn.k_proj": true,
819
+ "audio_tower.layers.9.self_attn.v_proj": true,
820
+ "audio_tower.layers.9.self_attn.q_proj": true,
821
+ "audio_tower.layers.9.self_attn.out_proj": true,
822
+ "audio_tower.layers.9.self_attn_layer_norm": false,
823
+ "audio_tower.layers.9.activation_fn": false,
824
+ "audio_tower.layers.9.fc1": true,
825
+ "audio_tower.layers.9.fc2": true,
826
+ "audio_tower.layers.9.final_layer_norm": false,
827
+ "audio_tower.layers.10.self_attn.k_proj": true,
828
+ "audio_tower.layers.10.self_attn.v_proj": true,
829
+ "audio_tower.layers.10.self_attn.q_proj": true,
830
+ "audio_tower.layers.10.self_attn.out_proj": true,
831
+ "audio_tower.layers.10.self_attn_layer_norm": false,
832
+ "audio_tower.layers.10.activation_fn": false,
833
+ "audio_tower.layers.10.fc1": true,
834
+ "audio_tower.layers.10.fc2": true,
835
+ "audio_tower.layers.10.final_layer_norm": false,
836
+ "audio_tower.layers.11.self_attn.k_proj": true,
837
+ "audio_tower.layers.11.self_attn.v_proj": true,
838
+ "audio_tower.layers.11.self_attn.q_proj": true,
839
+ "audio_tower.layers.11.self_attn.out_proj": true,
840
+ "audio_tower.layers.11.self_attn_layer_norm": false,
841
+ "audio_tower.layers.11.activation_fn": false,
842
+ "audio_tower.layers.11.fc1": true,
843
+ "audio_tower.layers.11.fc2": true,
844
+ "audio_tower.layers.11.final_layer_norm": false,
845
+ "audio_tower.layers.12.self_attn.k_proj": true,
846
+ "audio_tower.layers.12.self_attn.v_proj": true,
847
+ "audio_tower.layers.12.self_attn.q_proj": true,
848
+ "audio_tower.layers.12.self_attn.out_proj": true,
849
+ "audio_tower.layers.12.self_attn_layer_norm": false,
850
+ "audio_tower.layers.12.activation_fn": false,
851
+ "audio_tower.layers.12.fc1": true,
852
+ "audio_tower.layers.12.fc2": true,
853
+ "audio_tower.layers.12.final_layer_norm": false,
854
+ "audio_tower.layers.13.self_attn.k_proj": true,
855
+ "audio_tower.layers.13.self_attn.v_proj": true,
856
+ "audio_tower.layers.13.self_attn.q_proj": true,
857
+ "audio_tower.layers.13.self_attn.out_proj": true,
858
+ "audio_tower.layers.13.self_attn_layer_norm": false,
859
+ "audio_tower.layers.13.activation_fn": false,
860
+ "audio_tower.layers.13.fc1": true,
861
+ "audio_tower.layers.13.fc2": true,
862
+ "audio_tower.layers.13.final_layer_norm": false,
863
+ "audio_tower.layers.14.self_attn.k_proj": true,
864
+ "audio_tower.layers.14.self_attn.v_proj": true,
865
+ "audio_tower.layers.14.self_attn.q_proj": true,
866
+ "audio_tower.layers.14.self_attn.out_proj": true,
867
+ "audio_tower.layers.14.self_attn_layer_norm": false,
868
+ "audio_tower.layers.14.activation_fn": false,
869
+ "audio_tower.layers.14.fc1": true,
870
+ "audio_tower.layers.14.fc2": true,
871
+ "audio_tower.layers.14.final_layer_norm": false,
872
+ "audio_tower.layers.15.self_attn.k_proj": true,
873
+ "audio_tower.layers.15.self_attn.v_proj": true,
874
+ "audio_tower.layers.15.self_attn.q_proj": true,
875
+ "audio_tower.layers.15.self_attn.out_proj": true,
876
+ "audio_tower.layers.15.self_attn_layer_norm": false,
877
+ "audio_tower.layers.15.activation_fn": false,
878
+ "audio_tower.layers.15.fc1": true,
879
+ "audio_tower.layers.15.fc2": true,
880
+ "audio_tower.layers.15.final_layer_norm": false,
881
+ "audio_tower.layers.16.self_attn.k_proj": true,
882
+ "audio_tower.layers.16.self_attn.v_proj": true,
883
+ "audio_tower.layers.16.self_attn.q_proj": true,
884
+ "audio_tower.layers.16.self_attn.out_proj": true,
885
+ "audio_tower.layers.16.self_attn_layer_norm": false,
886
+ "audio_tower.layers.16.activation_fn": false,
887
+ "audio_tower.layers.16.fc1": true,
888
+ "audio_tower.layers.16.fc2": true,
889
+ "audio_tower.layers.16.final_layer_norm": false,
890
+ "audio_tower.layers.17.self_attn.k_proj": true,
891
+ "audio_tower.layers.17.self_attn.v_proj": true,
892
+ "audio_tower.layers.17.self_attn.q_proj": true,
893
+ "audio_tower.layers.17.self_attn.out_proj": true,
894
+ "audio_tower.layers.17.self_attn_layer_norm": false,
895
+ "audio_tower.layers.17.activation_fn": false,
896
+ "audio_tower.layers.17.fc1": true,
897
+ "audio_tower.layers.17.fc2": true,
898
+ "audio_tower.layers.17.final_layer_norm": false,
899
+ "audio_tower.layers.18.self_attn.k_proj": true,
900
+ "audio_tower.layers.18.self_attn.v_proj": true,
901
+ "audio_tower.layers.18.self_attn.q_proj": true,
902
+ "audio_tower.layers.18.self_attn.out_proj": true,
903
+ "audio_tower.layers.18.self_attn_layer_norm": false,
904
+ "audio_tower.layers.18.activation_fn": false,
905
+ "audio_tower.layers.18.fc1": true,
906
+ "audio_tower.layers.18.fc2": true,
907
+ "audio_tower.layers.18.final_layer_norm": false,
908
+ "audio_tower.layers.19.self_attn.k_proj": true,
909
+ "audio_tower.layers.19.self_attn.v_proj": true,
910
+ "audio_tower.layers.19.self_attn.q_proj": true,
911
+ "audio_tower.layers.19.self_attn.out_proj": true,
912
+ "audio_tower.layers.19.self_attn_layer_norm": false,
913
+ "audio_tower.layers.19.activation_fn": false,
914
+ "audio_tower.layers.19.fc1": true,
915
+ "audio_tower.layers.19.fc2": true,
916
+ "audio_tower.layers.19.final_layer_norm": false,
917
+ "audio_tower.layers.20.self_attn.k_proj": true,
918
+ "audio_tower.layers.20.self_attn.v_proj": true,
919
+ "audio_tower.layers.20.self_attn.q_proj": true,
920
+ "audio_tower.layers.20.self_attn.out_proj": true,
921
+ "audio_tower.layers.20.self_attn_layer_norm": false,
922
+ "audio_tower.layers.20.activation_fn": false,
923
+ "audio_tower.layers.20.fc1": true,
924
+ "audio_tower.layers.20.fc2": true,
925
+ "audio_tower.layers.20.final_layer_norm": false,
926
+ "audio_tower.layers.21.self_attn.k_proj": true,
927
+ "audio_tower.layers.21.self_attn.v_proj": true,
928
+ "audio_tower.layers.21.self_attn.q_proj": true,
929
+ "audio_tower.layers.21.self_attn.out_proj": true,
930
+ "audio_tower.layers.21.self_attn_layer_norm": false,
931
+ "audio_tower.layers.21.activation_fn": false,
932
+ "audio_tower.layers.21.fc1": true,
933
+ "audio_tower.layers.21.fc2": true,
934
+ "audio_tower.layers.21.final_layer_norm": false,
935
+ "audio_tower.layers.22.self_attn.k_proj": true,
936
+ "audio_tower.layers.22.self_attn.v_proj": true,
937
+ "audio_tower.layers.22.self_attn.q_proj": true,
938
+ "audio_tower.layers.22.self_attn.out_proj": true,
939
+ "audio_tower.layers.22.self_attn_layer_norm": false,
940
+ "audio_tower.layers.22.activation_fn": false,
941
+ "audio_tower.layers.22.fc1": true,
942
+ "audio_tower.layers.22.fc2": true,
943
+ "audio_tower.layers.22.final_layer_norm": false,
944
+ "audio_tower.layers.23.self_attn.k_proj": true,
945
+ "audio_tower.layers.23.self_attn.v_proj": true,
946
+ "audio_tower.layers.23.self_attn.q_proj": true,
947
+ "audio_tower.layers.23.self_attn.out_proj": true,
948
+ "audio_tower.layers.23.self_attn_layer_norm": false,
949
+ "audio_tower.layers.23.activation_fn": false,
950
+ "audio_tower.layers.23.fc1": true,
951
+ "audio_tower.layers.23.fc2": true,
952
+ "audio_tower.layers.23.final_layer_norm": false,
953
+ "audio_tower.layers.24.self_attn.k_proj": true,
954
+ "audio_tower.layers.24.self_attn.v_proj": true,
955
+ "audio_tower.layers.24.self_attn.q_proj": true,
956
+ "audio_tower.layers.24.self_attn.out_proj": true,
957
+ "audio_tower.layers.24.self_attn_layer_norm": false,
958
+ "audio_tower.layers.24.activation_fn": false,
959
+ "audio_tower.layers.24.fc1": true,
960
+ "audio_tower.layers.24.fc2": true,
961
+ "audio_tower.layers.24.final_layer_norm": false,
962
+ "audio_tower.layers.25.self_attn.k_proj": true,
963
+ "audio_tower.layers.25.self_attn.v_proj": true,
964
+ "audio_tower.layers.25.self_attn.q_proj": true,
965
+ "audio_tower.layers.25.self_attn.out_proj": true,
966
+ "audio_tower.layers.25.self_attn_layer_norm": false,
967
+ "audio_tower.layers.25.activation_fn": false,
968
+ "audio_tower.layers.25.fc1": true,
969
+ "audio_tower.layers.25.fc2": true,
970
+ "audio_tower.layers.25.final_layer_norm": false,
971
+ "audio_tower.layers.26.self_attn.k_proj": true,
972
+ "audio_tower.layers.26.self_attn.v_proj": true,
973
+ "audio_tower.layers.26.self_attn.q_proj": true,
974
+ "audio_tower.layers.26.self_attn.out_proj": true,
975
+ "audio_tower.layers.26.self_attn_layer_norm": false,
976
+ "audio_tower.layers.26.activation_fn": false,
977
+ "audio_tower.layers.26.fc1": true,
978
+ "audio_tower.layers.26.fc2": true,
979
+ "audio_tower.layers.26.final_layer_norm": false,
980
+ "audio_tower.layers.27.self_attn.k_proj": true,
981
+ "audio_tower.layers.27.self_attn.v_proj": true,
982
+ "audio_tower.layers.27.self_attn.q_proj": true,
983
+ "audio_tower.layers.27.self_attn.out_proj": true,
984
+ "audio_tower.layers.27.self_attn_layer_norm": false,
985
+ "audio_tower.layers.27.activation_fn": false,
986
+ "audio_tower.layers.27.fc1": true,
987
+ "audio_tower.layers.27.fc2": true,
988
+ "audio_tower.layers.27.final_layer_norm": false,
989
+ "audio_tower.layers.28.self_attn.k_proj": true,
990
+ "audio_tower.layers.28.self_attn.v_proj": true,
991
+ "audio_tower.layers.28.self_attn.q_proj": true,
992
+ "audio_tower.layers.28.self_attn.out_proj": true,
993
+ "audio_tower.layers.28.self_attn_layer_norm": false,
994
+ "audio_tower.layers.28.activation_fn": false,
995
+ "audio_tower.layers.28.fc1": true,
996
+ "audio_tower.layers.28.fc2": true,
997
+ "audio_tower.layers.28.final_layer_norm": false,
998
+ "audio_tower.layers.29.self_attn.k_proj": true,
999
+ "audio_tower.layers.29.self_attn.v_proj": true,
1000
+ "audio_tower.layers.29.self_attn.q_proj": true,
1001
+ "audio_tower.layers.29.self_attn.out_proj": true,
1002
+ "audio_tower.layers.29.self_attn_layer_norm": false,
1003
+ "audio_tower.layers.29.activation_fn": false,
1004
+ "audio_tower.layers.29.fc1": true,
1005
+ "audio_tower.layers.29.fc2": true,
1006
+ "audio_tower.layers.29.final_layer_norm": false,
1007
+ "audio_tower.layers.30.self_attn.k_proj": true,
1008
+ "audio_tower.layers.30.self_attn.v_proj": true,
1009
+ "audio_tower.layers.30.self_attn.q_proj": true,
1010
+ "audio_tower.layers.30.self_attn.out_proj": true,
1011
+ "audio_tower.layers.30.self_attn_layer_norm": false,
1012
+ "audio_tower.layers.30.activation_fn": false,
1013
+ "audio_tower.layers.30.fc1": true,
1014
+ "audio_tower.layers.30.fc2": true,
1015
+ "audio_tower.layers.30.final_layer_norm": false,
1016
+ "audio_tower.layers.31.self_attn.k_proj": true,
1017
+ "audio_tower.layers.31.self_attn.v_proj": true,
1018
+ "audio_tower.layers.31.self_attn.q_proj": true,
1019
+ "audio_tower.layers.31.self_attn.out_proj": true,
1020
+ "audio_tower.layers.31.self_attn_layer_norm": false,
1021
+ "audio_tower.layers.31.activation_fn": false,
1022
+ "audio_tower.layers.31.fc1": true,
1023
+ "audio_tower.layers.31.fc2": true,
1024
+ "audio_tower.layers.31.final_layer_norm": false,
1025
+ "audio_tower.layer_norm": false,
1026
+ "multi_modal_projector.linear_1": true,
1027
+ "multi_modal_projector.act": false,
1028
+ "multi_modal_projector.linear_2": true,
1029
+ "language_model.embed_tokens": true,
1030
+ "language_model.layers.0.self_attn.q_proj": true,
1031
+ "language_model.layers.0.self_attn.k_proj": true,
1032
+ "language_model.layers.0.self_attn.v_proj": true,
1033
+ "language_model.layers.0.self_attn.o_proj": true,
1034
+ "language_model.layers.0.self_attn.rope": false,
1035
+ "language_model.layers.0.mlp.gate_proj": true,
1036
+ "language_model.layers.0.mlp.up_proj": true,
1037
+ "language_model.layers.0.mlp.down_proj": true,
1038
+ "language_model.layers.0.input_layernorm": false,
1039
+ "language_model.layers.0.post_attention_layernorm": false,
1040
+ "language_model.layers.1.self_attn.q_proj": true,
1041
+ "language_model.layers.1.self_attn.k_proj": true,
1042
+ "language_model.layers.1.self_attn.v_proj": true,
1043
+ "language_model.layers.1.self_attn.o_proj": true,
1044
+ "language_model.layers.1.self_attn.rope": false,
1045
+ "language_model.layers.1.mlp.gate_proj": true,
1046
+ "language_model.layers.1.mlp.up_proj": true,
1047
+ "language_model.layers.1.mlp.down_proj": true,
1048
+ "language_model.layers.1.input_layernorm": false,
1049
+ "language_model.layers.1.post_attention_layernorm": false,
1050
+ "language_model.layers.2.self_attn.q_proj": true,
1051
+ "language_model.layers.2.self_attn.k_proj": true,
1052
+ "language_model.layers.2.self_attn.v_proj": true,
1053
+ "language_model.layers.2.self_attn.o_proj": true,
1054
+ "language_model.layers.2.self_attn.rope": false,
1055
+ "language_model.layers.2.mlp.gate_proj": true,
1056
+ "language_model.layers.2.mlp.up_proj": true,
1057
+ "language_model.layers.2.mlp.down_proj": true,
1058
+ "language_model.layers.2.input_layernorm": false,
1059
+ "language_model.layers.2.post_attention_layernorm": false,
1060
+ "language_model.layers.3.self_attn.q_proj": true,
1061
+ "language_model.layers.3.self_attn.k_proj": true,
1062
+ "language_model.layers.3.self_attn.v_proj": true,
1063
+ "language_model.layers.3.self_attn.o_proj": true,
1064
+ "language_model.layers.3.self_attn.rope": false,
1065
+ "language_model.layers.3.mlp.gate_proj": true,
1066
+ "language_model.layers.3.mlp.up_proj": true,
1067
+ "language_model.layers.3.mlp.down_proj": true,
1068
+ "language_model.layers.3.input_layernorm": false,
1069
+ "language_model.layers.3.post_attention_layernorm": false,
1070
+ "language_model.layers.4.self_attn.q_proj": true,
1071
+ "language_model.layers.4.self_attn.k_proj": true,
1072
+ "language_model.layers.4.self_attn.v_proj": true,
1073
+ "language_model.layers.4.self_attn.o_proj": true,
1074
+ "language_model.layers.4.self_attn.rope": false,
1075
+ "language_model.layers.4.mlp.gate_proj": true,
1076
+ "language_model.layers.4.mlp.up_proj": true,
1077
+ "language_model.layers.4.mlp.down_proj": true,
1078
+ "language_model.layers.4.input_layernorm": false,
1079
+ "language_model.layers.4.post_attention_layernorm": false,
1080
+ "language_model.layers.5.self_attn.q_proj": true,
1081
+ "language_model.layers.5.self_attn.k_proj": true,
1082
+ "language_model.layers.5.self_attn.v_proj": true,
1083
+ "language_model.layers.5.self_attn.o_proj": true,
1084
+ "language_model.layers.5.self_attn.rope": false,
1085
+ "language_model.layers.5.mlp.gate_proj": true,
1086
+ "language_model.layers.5.mlp.up_proj": true,
1087
+ "language_model.layers.5.mlp.down_proj": true,
1088
+ "language_model.layers.5.input_layernorm": false,
1089
+ "language_model.layers.5.post_attention_layernorm": false,
1090
+ "language_model.layers.6.self_attn.q_proj": true,
1091
+ "language_model.layers.6.self_attn.k_proj": true,
1092
+ "language_model.layers.6.self_attn.v_proj": true,
1093
+ "language_model.layers.6.self_attn.o_proj": true,
1094
+ "language_model.layers.6.self_attn.rope": false,
1095
+ "language_model.layers.6.mlp.gate_proj": true,
1096
+ "language_model.layers.6.mlp.up_proj": true,
1097
+ "language_model.layers.6.mlp.down_proj": true,
1098
+ "language_model.layers.6.input_layernorm": false,
1099
+ "language_model.layers.6.post_attention_layernorm": false,
1100
+ "language_model.layers.7.self_attn.q_proj": true,
1101
+ "language_model.layers.7.self_attn.k_proj": true,
1102
+ "language_model.layers.7.self_attn.v_proj": true,
1103
+ "language_model.layers.7.self_attn.o_proj": true,
1104
+ "language_model.layers.7.self_attn.rope": false,
1105
+ "language_model.layers.7.mlp.gate_proj": true,
1106
+ "language_model.layers.7.mlp.up_proj": true,
1107
+ "language_model.layers.7.mlp.down_proj": true,
1108
+ "language_model.layers.7.input_layernorm": false,
1109
+ "language_model.layers.7.post_attention_layernorm": false,
1110
+ "language_model.layers.8.self_attn.q_proj": true,
1111
+ "language_model.layers.8.self_attn.k_proj": true,
1112
+ "language_model.layers.8.self_attn.v_proj": true,
1113
+ "language_model.layers.8.self_attn.o_proj": true,
1114
+ "language_model.layers.8.self_attn.rope": false,
1115
+ "language_model.layers.8.mlp.gate_proj": true,
1116
+ "language_model.layers.8.mlp.up_proj": true,
1117
+ "language_model.layers.8.mlp.down_proj": true,
1118
+ "language_model.layers.8.input_layernorm": false,
1119
+ "language_model.layers.8.post_attention_layernorm": false,
1120
+ "language_model.layers.9.self_attn.q_proj": true,
1121
+ "language_model.layers.9.self_attn.k_proj": true,
1122
+ "language_model.layers.9.self_attn.v_proj": true,
1123
+ "language_model.layers.9.self_attn.o_proj": true,
1124
+ "language_model.layers.9.self_attn.rope": false,
1125
+ "language_model.layers.9.mlp.gate_proj": true,
1126
+ "language_model.layers.9.mlp.up_proj": true,
1127
+ "language_model.layers.9.mlp.down_proj": true,
1128
+ "language_model.layers.9.input_layernorm": false,
1129
+ "language_model.layers.9.post_attention_layernorm": false,
1130
+ "language_model.layers.10.self_attn.q_proj": true,
1131
+ "language_model.layers.10.self_attn.k_proj": true,
1132
+ "language_model.layers.10.self_attn.v_proj": true,
1133
+ "language_model.layers.10.self_attn.o_proj": true,
1134
+ "language_model.layers.10.self_attn.rope": false,
1135
+ "language_model.layers.10.mlp.gate_proj": true,
1136
+ "language_model.layers.10.mlp.up_proj": true,
1137
+ "language_model.layers.10.mlp.down_proj": true,
1138
+ "language_model.layers.10.input_layernorm": false,
1139
+ "language_model.layers.10.post_attention_layernorm": false,
1140
+ "language_model.layers.11.self_attn.q_proj": true,
1141
+ "language_model.layers.11.self_attn.k_proj": true,
1142
+ "language_model.layers.11.self_attn.v_proj": true,
1143
+ "language_model.layers.11.self_attn.o_proj": true,
1144
+ "language_model.layers.11.self_attn.rope": false,
1145
+ "language_model.layers.11.mlp.gate_proj": true,
1146
+ "language_model.layers.11.mlp.up_proj": true,
1147
+ "language_model.layers.11.mlp.down_proj": true,
1148
+ "language_model.layers.11.input_layernorm": false,
1149
+ "language_model.layers.11.post_attention_layernorm": false,
1150
+ "language_model.layers.12.self_attn.q_proj": true,
1151
+ "language_model.layers.12.self_attn.k_proj": true,
1152
+ "language_model.layers.12.self_attn.v_proj": true,
1153
+ "language_model.layers.12.self_attn.o_proj": true,
1154
+ "language_model.layers.12.self_attn.rope": false,
1155
+ "language_model.layers.12.mlp.gate_proj": true,
1156
+ "language_model.layers.12.mlp.up_proj": true,
1157
+ "language_model.layers.12.mlp.down_proj": true,
1158
+ "language_model.layers.12.input_layernorm": false,
1159
+ "language_model.layers.12.post_attention_layernorm": false,
1160
+ "language_model.layers.13.self_attn.q_proj": true,
1161
+ "language_model.layers.13.self_attn.k_proj": true,
1162
+ "language_model.layers.13.self_attn.v_proj": true,
1163
+ "language_model.layers.13.self_attn.o_proj": true,
1164
+ "language_model.layers.13.self_attn.rope": false,
1165
+ "language_model.layers.13.mlp.gate_proj": true,
1166
+ "language_model.layers.13.mlp.up_proj": true,
1167
+ "language_model.layers.13.mlp.down_proj": true,
1168
+ "language_model.layers.13.input_layernorm": false,
1169
+ "language_model.layers.13.post_attention_layernorm": false,
1170
+ "language_model.layers.14.self_attn.q_proj": true,
1171
+ "language_model.layers.14.self_attn.k_proj": true,
1172
+ "language_model.layers.14.self_attn.v_proj": true,
1173
+ "language_model.layers.14.self_attn.o_proj": true,
1174
+ "language_model.layers.14.self_attn.rope": false,
1175
+ "language_model.layers.14.mlp.gate_proj": true,
1176
+ "language_model.layers.14.mlp.up_proj": true,
1177
+ "language_model.layers.14.mlp.down_proj": true,
1178
+ "language_model.layers.14.input_layernorm": false,
1179
+ "language_model.layers.14.post_attention_layernorm": false,
1180
+ "language_model.layers.15.self_attn.q_proj": true,
1181
+ "language_model.layers.15.self_attn.k_proj": true,
1182
+ "language_model.layers.15.self_attn.v_proj": true,
1183
+ "language_model.layers.15.self_attn.o_proj": true,
1184
+ "language_model.layers.15.self_attn.rope": false,
1185
+ "language_model.layers.15.mlp.gate_proj": true,
1186
+ "language_model.layers.15.mlp.up_proj": true,
1187
+ "language_model.layers.15.mlp.down_proj": true,
1188
+ "language_model.layers.15.input_layernorm": false,
1189
+ "language_model.layers.15.post_attention_layernorm": false,
1190
+ "language_model.layers.16.self_attn.q_proj": true,
1191
+ "language_model.layers.16.self_attn.k_proj": true,
1192
+ "language_model.layers.16.self_attn.v_proj": true,
1193
+ "language_model.layers.16.self_attn.o_proj": true,
1194
+ "language_model.layers.16.self_attn.rope": false,
1195
+ "language_model.layers.16.mlp.gate_proj": true,
1196
+ "language_model.layers.16.mlp.up_proj": true,
1197
+ "language_model.layers.16.mlp.down_proj": true,
1198
+ "language_model.layers.16.input_layernorm": false,
1199
+ "language_model.layers.16.post_attention_layernorm": false,
1200
+ "language_model.layers.17.self_attn.q_proj": true,
1201
+ "language_model.layers.17.self_attn.k_proj": true,
1202
+ "language_model.layers.17.self_attn.v_proj": true,
1203
+ "language_model.layers.17.self_attn.o_proj": true,
1204
+ "language_model.layers.17.self_attn.rope": false,
1205
+ "language_model.layers.17.mlp.gate_proj": true,
1206
+ "language_model.layers.17.mlp.up_proj": true,
1207
+ "language_model.layers.17.mlp.down_proj": true,
1208
+ "language_model.layers.17.input_layernorm": false,
1209
+ "language_model.layers.17.post_attention_layernorm": false,
1210
+ "language_model.layers.18.self_attn.q_proj": true,
1211
+ "language_model.layers.18.self_attn.k_proj": true,
1212
+ "language_model.layers.18.self_attn.v_proj": true,
1213
+ "language_model.layers.18.self_attn.o_proj": true,
1214
+ "language_model.layers.18.self_attn.rope": false,
1215
+ "language_model.layers.18.mlp.gate_proj": true,
1216
+ "language_model.layers.18.mlp.up_proj": true,
1217
+ "language_model.layers.18.mlp.down_proj": true,
1218
+ "language_model.layers.18.input_layernorm": false,
1219
+ "language_model.layers.18.post_attention_layernorm": false,
1220
+ "language_model.layers.19.self_attn.q_proj": true,
1221
+ "language_model.layers.19.self_attn.k_proj": true,
1222
+ "language_model.layers.19.self_attn.v_proj": true,
1223
+ "language_model.layers.19.self_attn.o_proj": true,
1224
+ "language_model.layers.19.self_attn.rope": false,
1225
+ "language_model.layers.19.mlp.gate_proj": true,
1226
+ "language_model.layers.19.mlp.up_proj": true,
1227
+ "language_model.layers.19.mlp.down_proj": true,
1228
+ "language_model.layers.19.input_layernorm": false,
1229
+ "language_model.layers.19.post_attention_layernorm": false,
1230
+ "language_model.layers.20.self_attn.q_proj": true,
1231
+ "language_model.layers.20.self_attn.k_proj": true,
1232
+ "language_model.layers.20.self_attn.v_proj": true,
1233
+ "language_model.layers.20.self_attn.o_proj": true,
1234
+ "language_model.layers.20.self_attn.rope": false,
1235
+ "language_model.layers.20.mlp.gate_proj": true,
1236
+ "language_model.layers.20.mlp.up_proj": true,
1237
+ "language_model.layers.20.mlp.down_proj": true,
1238
+ "language_model.layers.20.input_layernorm": false,
1239
+ "language_model.layers.20.post_attention_layernorm": false,
1240
+ "language_model.layers.21.self_attn.q_proj": true,
1241
+ "language_model.layers.21.self_attn.k_proj": true,
1242
+ "language_model.layers.21.self_attn.v_proj": true,
1243
+ "language_model.layers.21.self_attn.o_proj": true,
1244
+ "language_model.layers.21.self_attn.rope": false,
1245
+ "language_model.layers.21.mlp.gate_proj": true,
1246
+ "language_model.layers.21.mlp.up_proj": true,
1247
+ "language_model.layers.21.mlp.down_proj": true,
1248
+ "language_model.layers.21.input_layernorm": false,
1249
+ "language_model.layers.21.post_attention_layernorm": false,
1250
+ "language_model.layers.22.self_attn.q_proj": true,
1251
+ "language_model.layers.22.self_attn.k_proj": true,
1252
+ "language_model.layers.22.self_attn.v_proj": true,
1253
+ "language_model.layers.22.self_attn.o_proj": true,
1254
+ "language_model.layers.22.self_attn.rope": false,
1255
+ "language_model.layers.22.mlp.gate_proj": true,
1256
+ "language_model.layers.22.mlp.up_proj": true,
1257
+ "language_model.layers.22.mlp.down_proj": true,
1258
+ "language_model.layers.22.input_layernorm": false,
1259
+ "language_model.layers.22.post_attention_layernorm": false,
1260
+ "language_model.layers.23.self_attn.q_proj": true,
1261
+ "language_model.layers.23.self_attn.k_proj": true,
1262
+ "language_model.layers.23.self_attn.v_proj": true,
1263
+ "language_model.layers.23.self_attn.o_proj": true,
1264
+ "language_model.layers.23.self_attn.rope": false,
1265
+ "language_model.layers.23.mlp.gate_proj": true,
1266
+ "language_model.layers.23.mlp.up_proj": true,
1267
+ "language_model.layers.23.mlp.down_proj": true,
1268
+ "language_model.layers.23.input_layernorm": false,
1269
+ "language_model.layers.23.post_attention_layernorm": false,
1270
+ "language_model.layers.24.self_attn.q_proj": true,
1271
+ "language_model.layers.24.self_attn.k_proj": true,
1272
+ "language_model.layers.24.self_attn.v_proj": true,
1273
+ "language_model.layers.24.self_attn.o_proj": true,
1274
+ "language_model.layers.24.self_attn.rope": false,
1275
+ "language_model.layers.24.mlp.gate_proj": true,
1276
+ "language_model.layers.24.mlp.up_proj": true,
1277
+ "language_model.layers.24.mlp.down_proj": true,
1278
+ "language_model.layers.24.input_layernorm": false,
1279
+ "language_model.layers.24.post_attention_layernorm": false,
1280
+ "language_model.layers.25.self_attn.q_proj": true,
1281
+ "language_model.layers.25.self_attn.k_proj": true,
1282
+ "language_model.layers.25.self_attn.v_proj": true,
1283
+ "language_model.layers.25.self_attn.o_proj": true,
1284
+ "language_model.layers.25.self_attn.rope": false,
1285
+ "language_model.layers.25.mlp.gate_proj": true,
1286
+ "language_model.layers.25.mlp.up_proj": true,
1287
+ "language_model.layers.25.mlp.down_proj": true,
1288
+ "language_model.layers.25.input_layernorm": false,
1289
+ "language_model.layers.25.post_attention_layernorm": false,
1290
+ "language_model.layers.26.self_attn.q_proj": true,
1291
+ "language_model.layers.26.self_attn.k_proj": true,
1292
+ "language_model.layers.26.self_attn.v_proj": true,
1293
+ "language_model.layers.26.self_attn.o_proj": true,
1294
+ "language_model.layers.26.self_attn.rope": false,
1295
+ "language_model.layers.26.mlp.gate_proj": true,
1296
+ "language_model.layers.26.mlp.up_proj": true,
1297
+ "language_model.layers.26.mlp.down_proj": true,
1298
+ "language_model.layers.26.input_layernorm": false,
1299
+ "language_model.layers.26.post_attention_layernorm": false,
1300
+ "language_model.layers.27.self_attn.q_proj": true,
1301
+ "language_model.layers.27.self_attn.k_proj": true,
1302
+ "language_model.layers.27.self_attn.v_proj": true,
1303
+ "language_model.layers.27.self_attn.o_proj": true,
1304
+ "language_model.layers.27.self_attn.rope": false,
1305
+ "language_model.layers.27.mlp.gate_proj": true,
1306
+ "language_model.layers.27.mlp.up_proj": true,
1307
+ "language_model.layers.27.mlp.down_proj": true,
1308
+ "language_model.layers.27.input_layernorm": false,
1309
+ "language_model.layers.27.post_attention_layernorm": false,
1310
+ "language_model.layers.28.self_attn.q_proj": true,
1311
+ "language_model.layers.28.self_attn.k_proj": true,
1312
+ "language_model.layers.28.self_attn.v_proj": true,
1313
+ "language_model.layers.28.self_attn.o_proj": true,
1314
+ "language_model.layers.28.self_attn.rope": false,
1315
+ "language_model.layers.28.mlp.gate_proj": true,
1316
+ "language_model.layers.28.mlp.up_proj": true,
1317
+ "language_model.layers.28.mlp.down_proj": true,
1318
+ "language_model.layers.28.input_layernorm": false,
1319
+ "language_model.layers.28.post_attention_layernorm": false,
1320
+ "language_model.layers.29.self_attn.q_proj": true,
1321
+ "language_model.layers.29.self_attn.k_proj": true,
1322
+ "language_model.layers.29.self_attn.v_proj": true,
1323
+ "language_model.layers.29.self_attn.o_proj": true,
1324
+ "language_model.layers.29.self_attn.rope": false,
1325
+ "language_model.layers.29.mlp.gate_proj": true,
1326
+ "language_model.layers.29.mlp.up_proj": true,
1327
+ "language_model.layers.29.mlp.down_proj": true,
1328
+ "language_model.layers.29.input_layernorm": false,
1329
+ "language_model.layers.29.post_attention_layernorm": false,
1330
+ "language_model.layers.30.self_attn.q_proj": true,
1331
+ "language_model.layers.30.self_attn.k_proj": true,
1332
+ "language_model.layers.30.self_attn.v_proj": true,
1333
+ "language_model.layers.30.self_attn.o_proj": true,
1334
+ "language_model.layers.30.self_attn.rope": false,
1335
+ "language_model.layers.30.mlp.gate_proj": true,
1336
+ "language_model.layers.30.mlp.up_proj": true,
1337
+ "language_model.layers.30.mlp.down_proj": true,
1338
+ "language_model.layers.30.input_layernorm": false,
1339
+ "language_model.layers.30.post_attention_layernorm": false,
1340
+ "language_model.layers.31.self_attn.q_proj": true,
1341
+ "language_model.layers.31.self_attn.k_proj": true,
1342
+ "language_model.layers.31.self_attn.v_proj": true,
1343
+ "language_model.layers.31.self_attn.o_proj": true,
1344
+ "language_model.layers.31.self_attn.rope": false,
1345
+ "language_model.layers.31.mlp.gate_proj": true,
1346
+ "language_model.layers.31.mlp.up_proj": true,
1347
+ "language_model.layers.31.mlp.down_proj": true,
1348
+ "language_model.layers.31.input_layernorm": false,
1349
+ "language_model.layers.31.post_attention_layernorm": false,
1350
+ "language_model.layers.32.self_attn.q_proj": true,
1351
+ "language_model.layers.32.self_attn.k_proj": true,
1352
+ "language_model.layers.32.self_attn.v_proj": true,
1353
+ "language_model.layers.32.self_attn.o_proj": true,
1354
+ "language_model.layers.32.self_attn.rope": false,
1355
+ "language_model.layers.32.mlp.gate_proj": true,
1356
+ "language_model.layers.32.mlp.up_proj": true,
1357
+ "language_model.layers.32.mlp.down_proj": true,
1358
+ "language_model.layers.32.input_layernorm": false,
1359
+ "language_model.layers.32.post_attention_layernorm": false,
1360
+ "language_model.layers.33.self_attn.q_proj": true,
1361
+ "language_model.layers.33.self_attn.k_proj": true,
1362
+ "language_model.layers.33.self_attn.v_proj": true,
1363
+ "language_model.layers.33.self_attn.o_proj": true,
1364
+ "language_model.layers.33.self_attn.rope": false,
1365
+ "language_model.layers.33.mlp.gate_proj": true,
1366
+ "language_model.layers.33.mlp.up_proj": true,
1367
+ "language_model.layers.33.mlp.down_proj": true,
1368
+ "language_model.layers.33.input_layernorm": false,
1369
+ "language_model.layers.33.post_attention_layernorm": false,
1370
+ "language_model.layers.34.self_attn.q_proj": true,
1371
+ "language_model.layers.34.self_attn.k_proj": true,
1372
+ "language_model.layers.34.self_attn.v_proj": true,
1373
+ "language_model.layers.34.self_attn.o_proj": true,
1374
+ "language_model.layers.34.self_attn.rope": false,
1375
+ "language_model.layers.34.mlp.gate_proj": true,
1376
+ "language_model.layers.34.mlp.up_proj": true,
1377
+ "language_model.layers.34.mlp.down_proj": true,
1378
+ "language_model.layers.34.input_layernorm": false,
1379
+ "language_model.layers.34.post_attention_layernorm": false,
1380
+ "language_model.layers.35.self_attn.q_proj": true,
1381
+ "language_model.layers.35.self_attn.k_proj": true,
1382
+ "language_model.layers.35.self_attn.v_proj": true,
1383
+ "language_model.layers.35.self_attn.o_proj": true,
1384
+ "language_model.layers.35.self_attn.rope": false,
1385
+ "language_model.layers.35.mlp.gate_proj": true,
1386
+ "language_model.layers.35.mlp.up_proj": true,
1387
+ "language_model.layers.35.mlp.down_proj": true,
1388
+ "language_model.layers.35.input_layernorm": false,
1389
+ "language_model.layers.35.post_attention_layernorm": false,
1390
+ "language_model.layers.36.self_attn.q_proj": true,
1391
+ "language_model.layers.36.self_attn.k_proj": true,
1392
+ "language_model.layers.36.self_attn.v_proj": true,
1393
+ "language_model.layers.36.self_attn.o_proj": true,
1394
+ "language_model.layers.36.self_attn.rope": false,
1395
+ "language_model.layers.36.mlp.gate_proj": true,
1396
+ "language_model.layers.36.mlp.up_proj": true,
1397
+ "language_model.layers.36.mlp.down_proj": true,
1398
+ "language_model.layers.36.input_layernorm": false,
1399
+ "language_model.layers.36.post_attention_layernorm": false,
1400
+ "language_model.layers.37.self_attn.q_proj": true,
1401
+ "language_model.layers.37.self_attn.k_proj": true,
1402
+ "language_model.layers.37.self_attn.v_proj": true,
1403
+ "language_model.layers.37.self_attn.o_proj": true,
1404
+ "language_model.layers.37.self_attn.rope": false,
1405
+ "language_model.layers.37.mlp.gate_proj": true,
1406
+ "language_model.layers.37.mlp.up_proj": true,
1407
+ "language_model.layers.37.mlp.down_proj": true,
1408
+ "language_model.layers.37.input_layernorm": false,
1409
+ "language_model.layers.37.post_attention_layernorm": false,
1410
+ "language_model.layers.38.self_attn.q_proj": true,
1411
+ "language_model.layers.38.self_attn.k_proj": true,
1412
+ "language_model.layers.38.self_attn.v_proj": true,
1413
+ "language_model.layers.38.self_attn.o_proj": true,
1414
+ "language_model.layers.38.self_attn.rope": false,
1415
+ "language_model.layers.38.mlp.gate_proj": true,
1416
+ "language_model.layers.38.mlp.up_proj": true,
1417
+ "language_model.layers.38.mlp.down_proj": true,
1418
+ "language_model.layers.38.input_layernorm": false,
1419
+ "language_model.layers.38.post_attention_layernorm": false,
1420
+ "language_model.layers.39.self_attn.q_proj": true,
1421
+ "language_model.layers.39.self_attn.k_proj": true,
1422
+ "language_model.layers.39.self_attn.v_proj": true,
1423
+ "language_model.layers.39.self_attn.o_proj": true,
1424
+ "language_model.layers.39.self_attn.rope": false,
1425
+ "language_model.layers.39.mlp.gate_proj": true,
1426
+ "language_model.layers.39.mlp.up_proj": true,
1427
+ "language_model.layers.39.mlp.down_proj": true,
1428
+ "language_model.layers.39.input_layernorm": false,
1429
+ "language_model.layers.39.post_attention_layernorm": false,
1430
+ "language_model.norm": false,
1431
+ "embed_tokens": true,
1432
+ "lm_head": true
1433
+ },
1434
+ "text_config": {
1435
+ "attention_bias": false,
1436
+ "attention_dropout": 0.0,
1437
+ "head_dim": 128,
1438
+ "hidden_act": "silu",
1439
+ "hidden_size": 5120,
1440
+ "initializer_range": 0.02,
1441
+ "intermediate_size": 32768,
1442
+ "max_position_embeddings": 131072,
1443
+ "mlp_bias": false,
1444
+ "model_type": "llama",
1445
+ "num_attention_heads": 32,
1446
+ "num_hidden_layers": 40,
1447
+ "num_key_value_heads": 8,
1448
+ "pretraining_tp": 1,
1449
+ "rms_norm_eps": 1e-05,
1450
+ "rope_scaling": null,
1451
+ "rope_theta": 100000000.0,
1452
+ "sliding_window": null,
1453
+ "use_cache": true,
1454
+ "vocab_size": 131072
1455
+ },
1456
+ "torch_dtype": "bfloat16",
1457
+ "transformers_version": "4.54.0.dev0",
1458
+ "vocab_size": 131072
1459
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 1,
3
+ "eos_token_id": 2,
4
+ "pad_token_id": 11,
5
+ "transformers_version": "4.54.0.dev0"
6
+ }
model-00001-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a65cbe382fed616424f9693db1e5673d01d344f9db71b921e2294a0124a68c04
3
+ size 5229934542
model-00002-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f21a585d2145dfd58ec0c53c840bdbc33d832f9164184820aa397b11a23d511
3
+ size 5314524015
model-00003-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d977dcebf6cdd09bf2aa711a9c23bf2f0f4d85f94198742b7b2fde1220d778fa
3
+ size 5314524070
model-00004-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33c5a38e302ce449fc72eb6d1214b5c15bd21d08a197dc49748e46f8c32a9c95
3
+ size 5314524062
model-00005-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:019ab4d90201ee57223fe6f29f5e57568ca11a5f9c9bf6241ded6518279a8a67
3
+ size 5325627680
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
params.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dim": 5120,
3
+ "n_layers": 40,
4
+ "head_dim": 128,
5
+ "hidden_dim": 32768,
6
+ "n_heads": 32,
7
+ "n_kv_heads": 8,
8
+ "rope_theta": 100000000.0,
9
+ "norm_eps": 1e-05,
10
+ "vocab_size": 131072,
11
+ "max_position_embeddings": 32768,
12
+ "multimodal": {
13
+ "whisper_model_args": {
14
+ "encoder_args": {
15
+ "dim": 1280,
16
+ "n_layers": 32,
17
+ "head_dim": 64,
18
+ "hidden_dim": 5120,
19
+ "n_heads": 20,
20
+ "vocab_size": 51866,
21
+ "max_source_positions": 1500,
22
+ "audio_encoding_args": {
23
+ "sampling_rate": 16000,
24
+ "num_mel_bins": 128,
25
+ "hop_length": 160,
26
+ "window_size": 400
27
+ }
28
+ },
29
+ "downsample_args": {
30
+ "downsample_factor": 4
31
+ }
32
+ }
33
+ }
34
+ }
preprocessor_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_length": 30,
3
+ "dither": 0.0,
4
+ "feature_extractor_type": "WhisperFeatureExtractor",
5
+ "feature_size": 128,
6
+ "hop_length": 160,
7
+ "n_fft": 400,
8
+ "n_samples": 480000,
9
+ "nb_max_frames": 3000,
10
+ "padding_side": "right",
11
+ "padding_value": 0.0,
12
+ "processor_class": "VoxtralProcessor",
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ }
tekken.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4aaf3836c2a5332f029ce85a7a62255c966f47b6797ef81dedd0ade9c862e4a8
3
+ size 14894206