UmeAiRT commited on
Commit
b8d2cf4
·
verified ·
1 Parent(s): bedbb5a

Sync upload config.json

Browse files
models/LLM/Qwen3-VL-8B-Instruct-FP8/config.json ADDED
@@ -0,0 +1,322 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3VLForConditionalGeneration"
4
+ ],
5
+ "image_token_id": 151655,
6
+ "model_type": "qwen3_vl",
7
+ "text_config": {
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "bos_token_id": 151643,
11
+ "dtype": "bfloat16",
12
+ "eos_token_id": 151645,
13
+ "head_dim": 128,
14
+ "hidden_act": "silu",
15
+ "hidden_size": 4096,
16
+ "initializer_range": 0.02,
17
+ "intermediate_size": 12288,
18
+ "max_position_embeddings": 262144,
19
+ "model_type": "qwen3_vl_text",
20
+ "num_attention_heads": 32,
21
+ "num_hidden_layers": 36,
22
+ "num_key_value_heads": 8,
23
+ "rms_norm_eps": 1e-06,
24
+ "rope_scaling": {
25
+ "mrope_interleaved": true,
26
+ "mrope_section": [
27
+ 24,
28
+ 20,
29
+ 20
30
+ ],
31
+ "rope_type": "default"
32
+ },
33
+ "rope_theta": 5000000,
34
+ "use_cache": true,
35
+ "vocab_size": 151936
36
+ },
37
+ "tie_word_embeddings": false,
38
+ "transformers_version": "4.57.0.dev0",
39
+ "video_token_id": 151656,
40
+ "vision_config": {
41
+ "deepstack_visual_indexes": [
42
+ 8,
43
+ 16,
44
+ 24
45
+ ],
46
+ "depth": 27,
47
+ "hidden_act": "gelu_pytorch_tanh",
48
+ "hidden_size": 1152,
49
+ "in_channels": 3,
50
+ "initializer_range": 0.02,
51
+ "intermediate_size": 4304,
52
+ "model_type": "qwen3_vl",
53
+ "num_heads": 16,
54
+ "num_position_embeddings": 2304,
55
+ "out_hidden_size": 4096,
56
+ "patch_size": 16,
57
+ "spatial_merge_size": 2,
58
+ "temporal_patch_size": 2
59
+ },
60
+ "vision_end_token_id": 151653,
61
+ "vision_start_token_id": 151652,
62
+ "quantization_config": {
63
+ "activation_scheme": "dynamic",
64
+ "fmt": "e4m3",
65
+ "quant_method": "fp8",
66
+ "ignored_layers": [
67
+ "lm_head",
68
+ "model.visual.merger.linear_fc1",
69
+ "model.visual.merger.linear_fc2",
70
+ "model.visual.merger.norm",
71
+ "model.visual.patch_embed.proj",
72
+ "model.visual.pos_embed",
73
+ "visual.merger.linear_fc1",
74
+ "visual.merger.linear_fc2",
75
+ "visual.merger.norm",
76
+ "visual.patch_embed.proj",
77
+ "visual.pos_embed",
78
+ "model.visual.blocks.0.attn.proj",
79
+ "model.visual.blocks.0.attn.qkv",
80
+ "model.visual.blocks.0.mlp.linear_fc1",
81
+ "model.visual.blocks.0.mlp.linear_fc2",
82
+ "visual.blocks.0.attn.proj",
83
+ "visual.blocks.0.attn.qkv_proj",
84
+ "visual.blocks.0.mlp.linear_fc1",
85
+ "visual.blocks.0.mlp.linear_fc2",
86
+ "model.visual.blocks.1.attn.proj",
87
+ "model.visual.blocks.1.attn.qkv",
88
+ "model.visual.blocks.1.mlp.linear_fc1",
89
+ "model.visual.blocks.1.mlp.linear_fc2",
90
+ "visual.blocks.1.attn.proj",
91
+ "visual.blocks.1.attn.qkv_proj",
92
+ "visual.blocks.1.mlp.linear_fc1",
93
+ "visual.blocks.1.mlp.linear_fc2",
94
+ "model.visual.blocks.2.attn.proj",
95
+ "model.visual.blocks.2.attn.qkv",
96
+ "model.visual.blocks.2.mlp.linear_fc1",
97
+ "model.visual.blocks.2.mlp.linear_fc2",
98
+ "visual.blocks.2.attn.proj",
99
+ "visual.blocks.2.attn.qkv_proj",
100
+ "visual.blocks.2.mlp.linear_fc1",
101
+ "visual.blocks.2.mlp.linear_fc2",
102
+ "model.visual.blocks.3.attn.proj",
103
+ "model.visual.blocks.3.attn.qkv",
104
+ "model.visual.blocks.3.mlp.linear_fc1",
105
+ "model.visual.blocks.3.mlp.linear_fc2",
106
+ "visual.blocks.3.attn.proj",
107
+ "visual.blocks.3.attn.qkv_proj",
108
+ "visual.blocks.3.mlp.linear_fc1",
109
+ "visual.blocks.3.mlp.linear_fc2",
110
+ "model.visual.blocks.4.attn.proj",
111
+ "model.visual.blocks.4.attn.qkv",
112
+ "model.visual.blocks.4.mlp.linear_fc1",
113
+ "model.visual.blocks.4.mlp.linear_fc2",
114
+ "visual.blocks.4.attn.proj",
115
+ "visual.blocks.4.attn.qkv_proj",
116
+ "visual.blocks.4.mlp.linear_fc1",
117
+ "visual.blocks.4.mlp.linear_fc2",
118
+ "model.visual.blocks.5.attn.proj",
119
+ "model.visual.blocks.5.attn.qkv",
120
+ "model.visual.blocks.5.mlp.linear_fc1",
121
+ "model.visual.blocks.5.mlp.linear_fc2",
122
+ "visual.blocks.5.attn.proj",
123
+ "visual.blocks.5.attn.qkv_proj",
124
+ "visual.blocks.5.mlp.linear_fc1",
125
+ "visual.blocks.5.mlp.linear_fc2",
126
+ "model.visual.blocks.6.attn.proj",
127
+ "model.visual.blocks.6.attn.qkv",
128
+ "model.visual.blocks.6.mlp.linear_fc1",
129
+ "model.visual.blocks.6.mlp.linear_fc2",
130
+ "visual.blocks.6.attn.proj",
131
+ "visual.blocks.6.attn.qkv_proj",
132
+ "visual.blocks.6.mlp.linear_fc1",
133
+ "visual.blocks.6.mlp.linear_fc2",
134
+ "model.visual.blocks.7.attn.proj",
135
+ "model.visual.blocks.7.attn.qkv",
136
+ "model.visual.blocks.7.mlp.linear_fc1",
137
+ "model.visual.blocks.7.mlp.linear_fc2",
138
+ "visual.blocks.7.attn.proj",
139
+ "visual.blocks.7.attn.qkv_proj",
140
+ "visual.blocks.7.mlp.linear_fc1",
141
+ "visual.blocks.7.mlp.linear_fc2",
142
+ "model.visual.blocks.8.attn.proj",
143
+ "model.visual.blocks.8.attn.qkv",
144
+ "model.visual.blocks.8.mlp.linear_fc1",
145
+ "model.visual.blocks.8.mlp.linear_fc2",
146
+ "visual.blocks.8.attn.proj",
147
+ "visual.blocks.8.attn.qkv_proj",
148
+ "visual.blocks.8.mlp.linear_fc1",
149
+ "visual.blocks.8.mlp.linear_fc2",
150
+ "model.visual.blocks.9.attn.proj",
151
+ "model.visual.blocks.9.attn.qkv",
152
+ "model.visual.blocks.9.mlp.linear_fc1",
153
+ "model.visual.blocks.9.mlp.linear_fc2",
154
+ "visual.blocks.9.attn.proj",
155
+ "visual.blocks.9.attn.qkv_proj",
156
+ "visual.blocks.9.mlp.linear_fc1",
157
+ "visual.blocks.9.mlp.linear_fc2",
158
+ "model.visual.blocks.10.attn.proj",
159
+ "model.visual.blocks.10.attn.qkv",
160
+ "model.visual.blocks.10.mlp.linear_fc1",
161
+ "model.visual.blocks.10.mlp.linear_fc2",
162
+ "visual.blocks.10.attn.proj",
163
+ "visual.blocks.10.attn.qkv_proj",
164
+ "visual.blocks.10.mlp.linear_fc1",
165
+ "visual.blocks.10.mlp.linear_fc2",
166
+ "model.visual.blocks.11.attn.proj",
167
+ "model.visual.blocks.11.attn.qkv",
168
+ "model.visual.blocks.11.mlp.linear_fc1",
169
+ "model.visual.blocks.11.mlp.linear_fc2",
170
+ "visual.blocks.11.attn.proj",
171
+ "visual.blocks.11.attn.qkv_proj",
172
+ "visual.blocks.11.mlp.linear_fc1",
173
+ "visual.blocks.11.mlp.linear_fc2",
174
+ "model.visual.blocks.12.attn.proj",
175
+ "model.visual.blocks.12.attn.qkv",
176
+ "model.visual.blocks.12.mlp.linear_fc1",
177
+ "model.visual.blocks.12.mlp.linear_fc2",
178
+ "visual.blocks.12.attn.proj",
179
+ "visual.blocks.12.attn.qkv_proj",
180
+ "visual.blocks.12.mlp.linear_fc1",
181
+ "visual.blocks.12.mlp.linear_fc2",
182
+ "model.visual.blocks.13.attn.proj",
183
+ "model.visual.blocks.13.attn.qkv",
184
+ "model.visual.blocks.13.mlp.linear_fc1",
185
+ "model.visual.blocks.13.mlp.linear_fc2",
186
+ "visual.blocks.13.attn.proj",
187
+ "visual.blocks.13.attn.qkv_proj",
188
+ "visual.blocks.13.mlp.linear_fc1",
189
+ "visual.blocks.13.mlp.linear_fc2",
190
+ "model.visual.blocks.14.attn.proj",
191
+ "model.visual.blocks.14.attn.qkv",
192
+ "model.visual.blocks.14.mlp.linear_fc1",
193
+ "model.visual.blocks.14.mlp.linear_fc2",
194
+ "visual.blocks.14.attn.proj",
195
+ "visual.blocks.14.attn.qkv_proj",
196
+ "visual.blocks.14.mlp.linear_fc1",
197
+ "visual.blocks.14.mlp.linear_fc2",
198
+ "model.visual.blocks.15.attn.proj",
199
+ "model.visual.blocks.15.attn.qkv",
200
+ "model.visual.blocks.15.mlp.linear_fc1",
201
+ "model.visual.blocks.15.mlp.linear_fc2",
202
+ "visual.blocks.15.attn.proj",
203
+ "visual.blocks.15.attn.qkv_proj",
204
+ "visual.blocks.15.mlp.linear_fc1",
205
+ "visual.blocks.15.mlp.linear_fc2",
206
+ "model.visual.blocks.16.attn.proj",
207
+ "model.visual.blocks.16.attn.qkv",
208
+ "model.visual.blocks.16.mlp.linear_fc1",
209
+ "model.visual.blocks.16.mlp.linear_fc2",
210
+ "visual.blocks.16.attn.proj",
211
+ "visual.blocks.16.attn.qkv_proj",
212
+ "visual.blocks.16.mlp.linear_fc1",
213
+ "visual.blocks.16.mlp.linear_fc2",
214
+ "model.visual.blocks.17.attn.proj",
215
+ "model.visual.blocks.17.attn.qkv",
216
+ "model.visual.blocks.17.mlp.linear_fc1",
217
+ "model.visual.blocks.17.mlp.linear_fc2",
218
+ "visual.blocks.17.attn.proj",
219
+ "visual.blocks.17.attn.qkv_proj",
220
+ "visual.blocks.17.mlp.linear_fc1",
221
+ "visual.blocks.17.mlp.linear_fc2",
222
+ "model.visual.blocks.18.attn.proj",
223
+ "model.visual.blocks.18.attn.qkv",
224
+ "model.visual.blocks.18.mlp.linear_fc1",
225
+ "model.visual.blocks.18.mlp.linear_fc2",
226
+ "visual.blocks.18.attn.proj",
227
+ "visual.blocks.18.attn.qkv_proj",
228
+ "visual.blocks.18.mlp.linear_fc1",
229
+ "visual.blocks.18.mlp.linear_fc2",
230
+ "model.visual.blocks.19.attn.proj",
231
+ "model.visual.blocks.19.attn.qkv",
232
+ "model.visual.blocks.19.mlp.linear_fc1",
233
+ "model.visual.blocks.19.mlp.linear_fc2",
234
+ "visual.blocks.19.attn.proj",
235
+ "visual.blocks.19.attn.qkv_proj",
236
+ "visual.blocks.19.mlp.linear_fc1",
237
+ "visual.blocks.19.mlp.linear_fc2",
238
+ "model.visual.blocks.20.attn.proj",
239
+ "model.visual.blocks.20.attn.qkv",
240
+ "model.visual.blocks.20.mlp.linear_fc1",
241
+ "model.visual.blocks.20.mlp.linear_fc2",
242
+ "visual.blocks.20.attn.proj",
243
+ "visual.blocks.20.attn.qkv_proj",
244
+ "visual.blocks.20.mlp.linear_fc1",
245
+ "visual.blocks.20.mlp.linear_fc2",
246
+ "model.visual.blocks.21.attn.proj",
247
+ "model.visual.blocks.21.attn.qkv",
248
+ "model.visual.blocks.21.mlp.linear_fc1",
249
+ "model.visual.blocks.21.mlp.linear_fc2",
250
+ "visual.blocks.21.attn.proj",
251
+ "visual.blocks.21.attn.qkv_proj",
252
+ "visual.blocks.21.mlp.linear_fc1",
253
+ "visual.blocks.21.mlp.linear_fc2",
254
+ "model.visual.blocks.22.attn.proj",
255
+ "model.visual.blocks.22.attn.qkv",
256
+ "model.visual.blocks.22.mlp.linear_fc1",
257
+ "model.visual.blocks.22.mlp.linear_fc2",
258
+ "visual.blocks.22.attn.proj",
259
+ "visual.blocks.22.attn.qkv_proj",
260
+ "visual.blocks.22.mlp.linear_fc1",
261
+ "visual.blocks.22.mlp.linear_fc2",
262
+ "model.visual.blocks.23.attn.proj",
263
+ "model.visual.blocks.23.attn.qkv",
264
+ "model.visual.blocks.23.mlp.linear_fc1",
265
+ "model.visual.blocks.23.mlp.linear_fc2",
266
+ "visual.blocks.23.attn.proj",
267
+ "visual.blocks.23.attn.qkv_proj",
268
+ "visual.blocks.23.mlp.linear_fc1",
269
+ "visual.blocks.23.mlp.linear_fc2",
270
+ "model.visual.blocks.24.attn.proj",
271
+ "model.visual.blocks.24.attn.qkv",
272
+ "model.visual.blocks.24.mlp.linear_fc1",
273
+ "model.visual.blocks.24.mlp.linear_fc2",
274
+ "visual.blocks.24.attn.proj",
275
+ "visual.blocks.24.attn.qkv_proj",
276
+ "visual.blocks.24.mlp.linear_fc1",
277
+ "visual.blocks.24.mlp.linear_fc2",
278
+ "model.visual.blocks.25.attn.proj",
279
+ "model.visual.blocks.25.attn.qkv",
280
+ "model.visual.blocks.25.mlp.linear_fc1",
281
+ "model.visual.blocks.25.mlp.linear_fc2",
282
+ "visual.blocks.25.attn.proj",
283
+ "visual.blocks.25.attn.qkv_proj",
284
+ "visual.blocks.25.mlp.linear_fc1",
285
+ "visual.blocks.25.mlp.linear_fc2",
286
+ "model.visual.blocks.26.attn.proj",
287
+ "model.visual.blocks.26.attn.qkv",
288
+ "model.visual.blocks.26.mlp.linear_fc1",
289
+ "model.visual.blocks.26.mlp.linear_fc2",
290
+ "visual.blocks.26.attn.proj",
291
+ "visual.blocks.26.attn.qkv_proj",
292
+ "visual.blocks.26.mlp.linear_fc1",
293
+ "visual.blocks.26.mlp.linear_fc2",
294
+ "model.visual.deepstack_merger_list.0.linear_fc1",
295
+ "model.visual.deepstack_merger_list.0.linear_fc2",
296
+ "model.visual.deepstack_merger_list.0.norm",
297
+ "visual.deepstack_merger_list.0.linear_fc1",
298
+ "visual.deepstack_merger_list.0.linear_fc2",
299
+ "visual.deepstack_merger_list.0.norm",
300
+ "model.visual.deepstack_merger_list.1.linear_fc1",
301
+ "model.visual.deepstack_merger_list.1.linear_fc2",
302
+ "model.visual.deepstack_merger_list.1.norm",
303
+ "visual.deepstack_merger_list.1.linear_fc1",
304
+ "visual.deepstack_merger_list.1.linear_fc2",
305
+ "visual.deepstack_merger_list.1.norm",
306
+ "model.visual.deepstack_merger_list.2.linear_fc1",
307
+ "model.visual.deepstack_merger_list.2.linear_fc2",
308
+ "model.visual.deepstack_merger_list.2.norm",
309
+ "visual.deepstack_merger_list.2.linear_fc1",
310
+ "visual.deepstack_merger_list.2.linear_fc2",
311
+ "visual.deepstack_merger_list.2.norm"
312
+ ],
313
+ "modules_to_not_convert":[
314
+ "lm_head",
315
+ "model.visual"
316
+ ],
317
+ "weight_block_size": [
318
+ 128,
319
+ 128
320
+ ]
321
+ }
322
+ }