cpatonn commited on
Commit
eeec5f3
·
verified ·
1 Parent(s): 1c5b3f5

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +615 -436
config.json CHANGED
@@ -1,79 +1,11 @@
1
  {
2
  "architectures": [
3
- "Qwen3_5MoeForCausalLM"
4
  ],
5
- "attention_bias": false,
6
- "attention_dropout": 0.0,
7
- "attn_output_gate": true,
8
- "bos_token_id": 248044,
9
  "dtype": "float16",
10
- "eos_token_id": 248044,
11
- "full_attention_interval": 4,
12
- "head_dim": 256,
13
- "hidden_act": "silu",
14
  "hidden_size": 2048,
15
- "initializer_range": 0.02,
16
- "layer_types": [
17
- "linear_attention",
18
- "linear_attention",
19
- "linear_attention",
20
- "full_attention",
21
- "linear_attention",
22
- "linear_attention",
23
- "linear_attention",
24
- "full_attention",
25
- "linear_attention",
26
- "linear_attention",
27
- "linear_attention",
28
- "full_attention",
29
- "linear_attention",
30
- "linear_attention",
31
- "linear_attention",
32
- "full_attention",
33
- "linear_attention",
34
- "linear_attention",
35
- "linear_attention",
36
- "full_attention",
37
- "linear_attention",
38
- "linear_attention",
39
- "linear_attention",
40
- "full_attention",
41
- "linear_attention",
42
- "linear_attention",
43
- "linear_attention",
44
- "full_attention",
45
- "linear_attention",
46
- "linear_attention",
47
- "linear_attention",
48
- "full_attention",
49
- "linear_attention",
50
- "linear_attention",
51
- "linear_attention",
52
- "full_attention",
53
- "linear_attention",
54
- "linear_attention",
55
- "linear_attention",
56
- "full_attention"
57
- ],
58
- "linear_conv_kernel_dim": 4,
59
- "linear_key_head_dim": 128,
60
- "linear_num_key_heads": 16,
61
- "linear_num_value_heads": 32,
62
- "linear_value_head_dim": 128,
63
- "mamba_ssm_dtype": "float32",
64
- "max_position_embeddings": 262144,
65
- "model_type": "qwen3_5_moe_text",
66
- "moe_intermediate_size": 512,
67
- "mtp_num_hidden_layers": 1,
68
- "mtp_use_dedicated_embeddings": false,
69
- "num_attention_heads": 16,
70
- "num_experts": 256,
71
- "num_experts_per_tok": 8,
72
- "num_hidden_layers": 40,
73
- "num_key_value_heads": 2,
74
- "output_router_logits": false,
75
- "pad_token_id": 248044,
76
- "partial_rotary_factor": 0.25,
77
  "quantization_config": {
78
  "config_groups": {
79
  "group_0": {
@@ -102,356 +34,507 @@
102
  "format": "pack-quantized",
103
  "global_compression_ratio": null,
104
  "ignore": [
105
- "model.layers.0.linear_attn.out_proj",
106
- "model.layers.0.linear_attn.in_proj_qkv",
107
- "model.layers.0.linear_attn.in_proj_z",
108
- "model.layers.0.linear_attn.in_proj_b",
109
- "model.layers.0.linear_attn.in_proj_a",
110
- "model.layers.0.mlp.shared_expert.gate_proj",
111
- "model.layers.0.mlp.shared_expert.up_proj",
112
- "model.layers.0.mlp.shared_expert.down_proj",
113
- "model.layers.0.mlp.shared_expert_gate",
114
- "model.layers.1.linear_attn.out_proj",
115
- "model.layers.1.linear_attn.in_proj_qkv",
116
- "model.layers.1.linear_attn.in_proj_z",
117
- "model.layers.1.linear_attn.in_proj_b",
118
- "model.layers.1.linear_attn.in_proj_a",
119
- "model.layers.1.mlp.shared_expert.gate_proj",
120
- "model.layers.1.mlp.shared_expert.up_proj",
121
- "model.layers.1.mlp.shared_expert.down_proj",
122
- "model.layers.1.mlp.shared_expert_gate",
123
- "model.layers.2.linear_attn.out_proj",
124
- "model.layers.2.linear_attn.in_proj_qkv",
125
- "model.layers.2.linear_attn.in_proj_z",
126
- "model.layers.2.linear_attn.in_proj_b",
127
- "model.layers.2.linear_attn.in_proj_a",
128
- "model.layers.2.mlp.shared_expert.gate_proj",
129
- "model.layers.2.mlp.shared_expert.up_proj",
130
- "model.layers.2.mlp.shared_expert.down_proj",
131
- "model.layers.2.mlp.shared_expert_gate",
132
- "model.layers.3.self_attn.q_proj",
133
- "model.layers.3.self_attn.k_proj",
134
- "model.layers.3.self_attn.v_proj",
135
- "model.layers.3.self_attn.o_proj",
136
- "model.layers.3.mlp.shared_expert.gate_proj",
137
- "model.layers.3.mlp.shared_expert.up_proj",
138
- "model.layers.3.mlp.shared_expert.down_proj",
139
- "model.layers.3.mlp.shared_expert_gate",
140
- "model.layers.4.linear_attn.out_proj",
141
- "model.layers.4.linear_attn.in_proj_qkv",
142
- "model.layers.4.linear_attn.in_proj_z",
143
- "model.layers.4.linear_attn.in_proj_b",
144
- "model.layers.4.linear_attn.in_proj_a",
145
- "model.layers.4.mlp.shared_expert.gate_proj",
146
- "model.layers.4.mlp.shared_expert.up_proj",
147
- "model.layers.4.mlp.shared_expert.down_proj",
148
- "model.layers.4.mlp.shared_expert_gate",
149
- "model.layers.5.linear_attn.out_proj",
150
- "model.layers.5.linear_attn.in_proj_qkv",
151
- "model.layers.5.linear_attn.in_proj_z",
152
- "model.layers.5.linear_attn.in_proj_b",
153
- "model.layers.5.linear_attn.in_proj_a",
154
- "model.layers.5.mlp.shared_expert.gate_proj",
155
- "model.layers.5.mlp.shared_expert.up_proj",
156
- "model.layers.5.mlp.shared_expert.down_proj",
157
- "model.layers.5.mlp.shared_expert_gate",
158
- "model.layers.6.linear_attn.out_proj",
159
- "model.layers.6.linear_attn.in_proj_qkv",
160
- "model.layers.6.linear_attn.in_proj_z",
161
- "model.layers.6.linear_attn.in_proj_b",
162
- "model.layers.6.linear_attn.in_proj_a",
163
- "model.layers.6.mlp.shared_expert.gate_proj",
164
- "model.layers.6.mlp.shared_expert.up_proj",
165
- "model.layers.6.mlp.shared_expert.down_proj",
166
- "model.layers.6.mlp.shared_expert_gate",
167
- "model.layers.7.self_attn.q_proj",
168
- "model.layers.7.self_attn.k_proj",
169
- "model.layers.7.self_attn.v_proj",
170
- "model.layers.7.self_attn.o_proj",
171
- "model.layers.7.mlp.shared_expert.gate_proj",
172
- "model.layers.7.mlp.shared_expert.up_proj",
173
- "model.layers.7.mlp.shared_expert.down_proj",
174
- "model.layers.7.mlp.shared_expert_gate",
175
- "model.layers.8.linear_attn.out_proj",
176
- "model.layers.8.linear_attn.in_proj_qkv",
177
- "model.layers.8.linear_attn.in_proj_z",
178
- "model.layers.8.linear_attn.in_proj_b",
179
- "model.layers.8.linear_attn.in_proj_a",
180
- "model.layers.8.mlp.shared_expert.gate_proj",
181
- "model.layers.8.mlp.shared_expert.up_proj",
182
- "model.layers.8.mlp.shared_expert.down_proj",
183
- "model.layers.8.mlp.shared_expert_gate",
184
- "model.layers.9.linear_attn.out_proj",
185
- "model.layers.9.linear_attn.in_proj_qkv",
186
- "model.layers.9.linear_attn.in_proj_z",
187
- "model.layers.9.linear_attn.in_proj_b",
188
- "model.layers.9.linear_attn.in_proj_a",
189
- "model.layers.9.mlp.shared_expert.gate_proj",
190
- "model.layers.9.mlp.shared_expert.up_proj",
191
- "model.layers.9.mlp.shared_expert.down_proj",
192
- "model.layers.9.mlp.shared_expert_gate",
193
- "model.layers.10.linear_attn.out_proj",
194
- "model.layers.10.linear_attn.in_proj_qkv",
195
- "model.layers.10.linear_attn.in_proj_z",
196
- "model.layers.10.linear_attn.in_proj_b",
197
- "model.layers.10.linear_attn.in_proj_a",
198
- "model.layers.10.mlp.shared_expert.gate_proj",
199
- "model.layers.10.mlp.shared_expert.up_proj",
200
- "model.layers.10.mlp.shared_expert.down_proj",
201
- "model.layers.10.mlp.shared_expert_gate",
202
- "model.layers.11.self_attn.q_proj",
203
- "model.layers.11.self_attn.k_proj",
204
- "model.layers.11.self_attn.v_proj",
205
- "model.layers.11.self_attn.o_proj",
206
- "model.layers.11.mlp.shared_expert.gate_proj",
207
- "model.layers.11.mlp.shared_expert.up_proj",
208
- "model.layers.11.mlp.shared_expert.down_proj",
209
- "model.layers.11.mlp.shared_expert_gate",
210
- "model.layers.12.linear_attn.out_proj",
211
- "model.layers.12.linear_attn.in_proj_qkv",
212
- "model.layers.12.linear_attn.in_proj_z",
213
- "model.layers.12.linear_attn.in_proj_b",
214
- "model.layers.12.linear_attn.in_proj_a",
215
- "model.layers.12.mlp.shared_expert.gate_proj",
216
- "model.layers.12.mlp.shared_expert.up_proj",
217
- "model.layers.12.mlp.shared_expert.down_proj",
218
- "model.layers.12.mlp.shared_expert_gate",
219
- "model.layers.13.linear_attn.out_proj",
220
- "model.layers.13.linear_attn.in_proj_qkv",
221
- "model.layers.13.linear_attn.in_proj_z",
222
- "model.layers.13.linear_attn.in_proj_b",
223
- "model.layers.13.linear_attn.in_proj_a",
224
- "model.layers.13.mlp.shared_expert.gate_proj",
225
- "model.layers.13.mlp.shared_expert.up_proj",
226
- "model.layers.13.mlp.shared_expert.down_proj",
227
- "model.layers.13.mlp.shared_expert_gate",
228
- "model.layers.14.linear_attn.out_proj",
229
- "model.layers.14.linear_attn.in_proj_qkv",
230
- "model.layers.14.linear_attn.in_proj_z",
231
- "model.layers.14.linear_attn.in_proj_b",
232
- "model.layers.14.linear_attn.in_proj_a",
233
- "model.layers.14.mlp.shared_expert.gate_proj",
234
- "model.layers.14.mlp.shared_expert.up_proj",
235
- "model.layers.14.mlp.shared_expert.down_proj",
236
- "model.layers.14.mlp.shared_expert_gate",
237
- "model.layers.15.self_attn.q_proj",
238
- "model.layers.15.self_attn.k_proj",
239
- "model.layers.15.self_attn.v_proj",
240
- "model.layers.15.self_attn.o_proj",
241
- "model.layers.15.mlp.shared_expert.gate_proj",
242
- "model.layers.15.mlp.shared_expert.up_proj",
243
- "model.layers.15.mlp.shared_expert.down_proj",
244
- "model.layers.15.mlp.shared_expert_gate",
245
- "model.layers.16.linear_attn.out_proj",
246
- "model.layers.16.linear_attn.in_proj_qkv",
247
- "model.layers.16.linear_attn.in_proj_z",
248
- "model.layers.16.linear_attn.in_proj_b",
249
- "model.layers.16.linear_attn.in_proj_a",
250
- "model.layers.16.mlp.shared_expert.gate_proj",
251
- "model.layers.16.mlp.shared_expert.up_proj",
252
- "model.layers.16.mlp.shared_expert.down_proj",
253
- "model.layers.16.mlp.shared_expert_gate",
254
- "model.layers.17.linear_attn.out_proj",
255
- "model.layers.17.linear_attn.in_proj_qkv",
256
- "model.layers.17.linear_attn.in_proj_z",
257
- "model.layers.17.linear_attn.in_proj_b",
258
- "model.layers.17.linear_attn.in_proj_a",
259
- "model.layers.17.mlp.shared_expert.gate_proj",
260
- "model.layers.17.mlp.shared_expert.up_proj",
261
- "model.layers.17.mlp.shared_expert.down_proj",
262
- "model.layers.17.mlp.shared_expert_gate",
263
- "model.layers.18.linear_attn.out_proj",
264
- "model.layers.18.linear_attn.in_proj_qkv",
265
- "model.layers.18.linear_attn.in_proj_z",
266
- "model.layers.18.linear_attn.in_proj_b",
267
- "model.layers.18.linear_attn.in_proj_a",
268
- "model.layers.18.mlp.shared_expert.gate_proj",
269
- "model.layers.18.mlp.shared_expert.up_proj",
270
- "model.layers.18.mlp.shared_expert.down_proj",
271
- "model.layers.18.mlp.shared_expert_gate",
272
- "model.layers.19.self_attn.q_proj",
273
- "model.layers.19.self_attn.k_proj",
274
- "model.layers.19.self_attn.v_proj",
275
- "model.layers.19.self_attn.o_proj",
276
- "model.layers.19.mlp.shared_expert.gate_proj",
277
- "model.layers.19.mlp.shared_expert.up_proj",
278
- "model.layers.19.mlp.shared_expert.down_proj",
279
- "model.layers.19.mlp.shared_expert_gate",
280
- "model.layers.20.linear_attn.out_proj",
281
- "model.layers.20.linear_attn.in_proj_qkv",
282
- "model.layers.20.linear_attn.in_proj_z",
283
- "model.layers.20.linear_attn.in_proj_b",
284
- "model.layers.20.linear_attn.in_proj_a",
285
- "model.layers.20.mlp.shared_expert.gate_proj",
286
- "model.layers.20.mlp.shared_expert.up_proj",
287
- "model.layers.20.mlp.shared_expert.down_proj",
288
- "model.layers.20.mlp.shared_expert_gate",
289
- "model.layers.21.linear_attn.out_proj",
290
- "model.layers.21.linear_attn.in_proj_qkv",
291
- "model.layers.21.linear_attn.in_proj_z",
292
- "model.layers.21.linear_attn.in_proj_b",
293
- "model.layers.21.linear_attn.in_proj_a",
294
- "model.layers.21.mlp.shared_expert.gate_proj",
295
- "model.layers.21.mlp.shared_expert.up_proj",
296
- "model.layers.21.mlp.shared_expert.down_proj",
297
- "model.layers.21.mlp.shared_expert_gate",
298
- "model.layers.22.linear_attn.out_proj",
299
- "model.layers.22.linear_attn.in_proj_qkv",
300
- "model.layers.22.linear_attn.in_proj_z",
301
- "model.layers.22.linear_attn.in_proj_b",
302
- "model.layers.22.linear_attn.in_proj_a",
303
- "model.layers.22.mlp.shared_expert.gate_proj",
304
- "model.layers.22.mlp.shared_expert.up_proj",
305
- "model.layers.22.mlp.shared_expert.down_proj",
306
- "model.layers.22.mlp.shared_expert_gate",
307
- "model.layers.23.self_attn.q_proj",
308
- "model.layers.23.self_attn.k_proj",
309
- "model.layers.23.self_attn.v_proj",
310
- "model.layers.23.self_attn.o_proj",
311
- "model.layers.23.mlp.shared_expert.gate_proj",
312
- "model.layers.23.mlp.shared_expert.up_proj",
313
- "model.layers.23.mlp.shared_expert.down_proj",
314
- "model.layers.23.mlp.shared_expert_gate",
315
- "model.layers.24.linear_attn.out_proj",
316
- "model.layers.24.linear_attn.in_proj_qkv",
317
- "model.layers.24.linear_attn.in_proj_z",
318
- "model.layers.24.linear_attn.in_proj_b",
319
- "model.layers.24.linear_attn.in_proj_a",
320
- "model.layers.24.mlp.shared_expert.gate_proj",
321
- "model.layers.24.mlp.shared_expert.up_proj",
322
- "model.layers.24.mlp.shared_expert.down_proj",
323
- "model.layers.24.mlp.shared_expert_gate",
324
- "model.layers.25.linear_attn.out_proj",
325
- "model.layers.25.linear_attn.in_proj_qkv",
326
- "model.layers.25.linear_attn.in_proj_z",
327
- "model.layers.25.linear_attn.in_proj_b",
328
- "model.layers.25.linear_attn.in_proj_a",
329
- "model.layers.25.mlp.shared_expert.gate_proj",
330
- "model.layers.25.mlp.shared_expert.up_proj",
331
- "model.layers.25.mlp.shared_expert.down_proj",
332
- "model.layers.25.mlp.shared_expert_gate",
333
- "model.layers.26.linear_attn.out_proj",
334
- "model.layers.26.linear_attn.in_proj_qkv",
335
- "model.layers.26.linear_attn.in_proj_z",
336
- "model.layers.26.linear_attn.in_proj_b",
337
- "model.layers.26.linear_attn.in_proj_a",
338
- "model.layers.26.mlp.shared_expert.gate_proj",
339
- "model.layers.26.mlp.shared_expert.up_proj",
340
- "model.layers.26.mlp.shared_expert.down_proj",
341
- "model.layers.26.mlp.shared_expert_gate",
342
- "model.layers.27.self_attn.q_proj",
343
- "model.layers.27.self_attn.k_proj",
344
- "model.layers.27.self_attn.v_proj",
345
- "model.layers.27.self_attn.o_proj",
346
- "model.layers.27.mlp.shared_expert.gate_proj",
347
- "model.layers.27.mlp.shared_expert.up_proj",
348
- "model.layers.27.mlp.shared_expert.down_proj",
349
- "model.layers.27.mlp.shared_expert_gate",
350
- "model.layers.28.linear_attn.out_proj",
351
- "model.layers.28.linear_attn.in_proj_qkv",
352
- "model.layers.28.linear_attn.in_proj_z",
353
- "model.layers.28.linear_attn.in_proj_b",
354
- "model.layers.28.linear_attn.in_proj_a",
355
- "model.layers.28.mlp.shared_expert.gate_proj",
356
- "model.layers.28.mlp.shared_expert.up_proj",
357
- "model.layers.28.mlp.shared_expert.down_proj",
358
- "model.layers.28.mlp.shared_expert_gate",
359
- "model.layers.29.linear_attn.out_proj",
360
- "model.layers.29.linear_attn.in_proj_qkv",
361
- "model.layers.29.linear_attn.in_proj_z",
362
- "model.layers.29.linear_attn.in_proj_b",
363
- "model.layers.29.linear_attn.in_proj_a",
364
- "model.layers.29.mlp.shared_expert.gate_proj",
365
- "model.layers.29.mlp.shared_expert.up_proj",
366
- "model.layers.29.mlp.shared_expert.down_proj",
367
- "model.layers.29.mlp.shared_expert_gate",
368
- "model.layers.30.linear_attn.out_proj",
369
- "model.layers.30.linear_attn.in_proj_qkv",
370
- "model.layers.30.linear_attn.in_proj_z",
371
- "model.layers.30.linear_attn.in_proj_b",
372
- "model.layers.30.linear_attn.in_proj_a",
373
- "model.layers.30.mlp.shared_expert.gate_proj",
374
- "model.layers.30.mlp.shared_expert.up_proj",
375
- "model.layers.30.mlp.shared_expert.down_proj",
376
- "model.layers.30.mlp.shared_expert_gate",
377
- "model.layers.31.self_attn.q_proj",
378
- "model.layers.31.self_attn.k_proj",
379
- "model.layers.31.self_attn.v_proj",
380
- "model.layers.31.self_attn.o_proj",
381
- "model.layers.31.mlp.shared_expert.gate_proj",
382
- "model.layers.31.mlp.shared_expert.up_proj",
383
- "model.layers.31.mlp.shared_expert.down_proj",
384
- "model.layers.31.mlp.shared_expert_gate",
385
- "model.layers.32.linear_attn.out_proj",
386
- "model.layers.32.linear_attn.in_proj_qkv",
387
- "model.layers.32.linear_attn.in_proj_z",
388
- "model.layers.32.linear_attn.in_proj_b",
389
- "model.layers.32.linear_attn.in_proj_a",
390
- "model.layers.32.mlp.shared_expert.gate_proj",
391
- "model.layers.32.mlp.shared_expert.up_proj",
392
- "model.layers.32.mlp.shared_expert.down_proj",
393
- "model.layers.32.mlp.shared_expert_gate",
394
- "model.layers.33.linear_attn.out_proj",
395
- "model.layers.33.linear_attn.in_proj_qkv",
396
- "model.layers.33.linear_attn.in_proj_z",
397
- "model.layers.33.linear_attn.in_proj_b",
398
- "model.layers.33.linear_attn.in_proj_a",
399
- "model.layers.33.mlp.shared_expert.gate_proj",
400
- "model.layers.33.mlp.shared_expert.up_proj",
401
- "model.layers.33.mlp.shared_expert.down_proj",
402
- "model.layers.33.mlp.shared_expert_gate",
403
- "model.layers.34.linear_attn.out_proj",
404
- "model.layers.34.linear_attn.in_proj_qkv",
405
- "model.layers.34.linear_attn.in_proj_z",
406
- "model.layers.34.linear_attn.in_proj_b",
407
- "model.layers.34.linear_attn.in_proj_a",
408
- "model.layers.34.mlp.shared_expert.gate_proj",
409
- "model.layers.34.mlp.shared_expert.up_proj",
410
- "model.layers.34.mlp.shared_expert.down_proj",
411
- "model.layers.34.mlp.shared_expert_gate",
412
- "model.layers.35.self_attn.q_proj",
413
- "model.layers.35.self_attn.k_proj",
414
- "model.layers.35.self_attn.v_proj",
415
- "model.layers.35.self_attn.o_proj",
416
- "model.layers.35.mlp.shared_expert.gate_proj",
417
- "model.layers.35.mlp.shared_expert.up_proj",
418
- "model.layers.35.mlp.shared_expert.down_proj",
419
- "model.layers.35.mlp.shared_expert_gate",
420
- "model.layers.36.linear_attn.out_proj",
421
- "model.layers.36.linear_attn.in_proj_qkv",
422
- "model.layers.36.linear_attn.in_proj_z",
423
- "model.layers.36.linear_attn.in_proj_b",
424
- "model.layers.36.linear_attn.in_proj_a",
425
- "model.layers.36.mlp.shared_expert.gate_proj",
426
- "model.layers.36.mlp.shared_expert.up_proj",
427
- "model.layers.36.mlp.shared_expert.down_proj",
428
- "model.layers.36.mlp.shared_expert_gate",
429
- "model.layers.37.linear_attn.out_proj",
430
- "model.layers.37.linear_attn.in_proj_qkv",
431
- "model.layers.37.linear_attn.in_proj_z",
432
- "model.layers.37.linear_attn.in_proj_b",
433
- "model.layers.37.linear_attn.in_proj_a",
434
- "model.layers.37.mlp.shared_expert.gate_proj",
435
- "model.layers.37.mlp.shared_expert.up_proj",
436
- "model.layers.37.mlp.shared_expert.down_proj",
437
- "model.layers.37.mlp.shared_expert_gate",
438
- "model.layers.38.linear_attn.out_proj",
439
- "model.layers.38.linear_attn.in_proj_qkv",
440
- "model.layers.38.linear_attn.in_proj_z",
441
- "model.layers.38.linear_attn.in_proj_b",
442
- "model.layers.38.linear_attn.in_proj_a",
443
- "model.layers.38.mlp.shared_expert.gate_proj",
444
- "model.layers.38.mlp.shared_expert.up_proj",
445
- "model.layers.38.mlp.shared_expert.down_proj",
446
- "model.layers.38.mlp.shared_expert_gate",
447
- "model.layers.39.self_attn.q_proj",
448
- "model.layers.39.self_attn.k_proj",
449
- "model.layers.39.self_attn.v_proj",
450
- "model.layers.39.self_attn.o_proj",
451
- "model.layers.39.mlp.shared_expert.gate_proj",
452
- "model.layers.39.mlp.shared_expert.up_proj",
453
- "model.layers.39.mlp.shared_expert.down_proj",
454
- "model.layers.39.mlp.shared_expert_gate",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
455
  "lm_head"
456
  ],
457
  "kv_cache_scheme": null,
@@ -459,24 +542,120 @@
459
  "quantization_status": "compressed",
460
  "sparsity_config": {},
461
  "transform_config": {},
462
- "version": "0.1.dev490+gdde82c8"
463
  },
464
- "rms_norm_eps": 1e-06,
465
- "rope_parameters": {
466
- "mrope_interleaved": true,
467
- "mrope_section": [
468
- 11,
469
- 11,
470
- 10
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
471
  ],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
472
  "partial_rotary_factor": 0.25,
473
- "rope_theta": 10000000,
474
- "rope_type": "default"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
475
  },
476
- "router_aux_loss_coef": 0.0,
477
- "shared_expert_intermediate_size": 512,
478
  "tie_word_embeddings": false,
479
- "transformers_version": "5.12.1",
480
- "use_cache": false,
481
- "vocab_size": 248320
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
482
  }
 
1
  {
2
  "architectures": [
3
+ "Qwen3_5MoeForConditionalGeneration"
4
  ],
 
 
 
 
5
  "dtype": "float16",
 
 
 
 
6
  "hidden_size": 2048,
7
+ "image_token_id": 248056,
8
+ "model_type": "qwen3_5_moe",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  "quantization_config": {
10
  "config_groups": {
11
  "group_0": {
 
34
  "format": "pack-quantized",
35
  "global_compression_ratio": null,
36
  "ignore": [
37
+ "model.visual.blocks.0.attn.qkv",
38
+ "model.visual.blocks.0.attn.proj",
39
+ "model.visual.blocks.0.mlp.linear_fc1",
40
+ "model.visual.blocks.0.mlp.linear_fc2",
41
+ "model.visual.blocks.1.attn.qkv",
42
+ "model.visual.blocks.1.attn.proj",
43
+ "model.visual.blocks.1.mlp.linear_fc1",
44
+ "model.visual.blocks.1.mlp.linear_fc2",
45
+ "model.visual.blocks.2.attn.qkv",
46
+ "model.visual.blocks.2.attn.proj",
47
+ "model.visual.blocks.2.mlp.linear_fc1",
48
+ "model.visual.blocks.2.mlp.linear_fc2",
49
+ "model.visual.blocks.3.attn.qkv",
50
+ "model.visual.blocks.3.attn.proj",
51
+ "model.visual.blocks.3.mlp.linear_fc1",
52
+ "model.visual.blocks.3.mlp.linear_fc2",
53
+ "model.visual.blocks.4.attn.qkv",
54
+ "model.visual.blocks.4.attn.proj",
55
+ "model.visual.blocks.4.mlp.linear_fc1",
56
+ "model.visual.blocks.4.mlp.linear_fc2",
57
+ "model.visual.blocks.5.attn.qkv",
58
+ "model.visual.blocks.5.attn.proj",
59
+ "model.visual.blocks.5.mlp.linear_fc1",
60
+ "model.visual.blocks.5.mlp.linear_fc2",
61
+ "model.visual.blocks.6.attn.qkv",
62
+ "model.visual.blocks.6.attn.proj",
63
+ "model.visual.blocks.6.mlp.linear_fc1",
64
+ "model.visual.blocks.6.mlp.linear_fc2",
65
+ "model.visual.blocks.7.attn.qkv",
66
+ "model.visual.blocks.7.attn.proj",
67
+ "model.visual.blocks.7.mlp.linear_fc1",
68
+ "model.visual.blocks.7.mlp.linear_fc2",
69
+ "model.visual.blocks.8.attn.qkv",
70
+ "model.visual.blocks.8.attn.proj",
71
+ "model.visual.blocks.8.mlp.linear_fc1",
72
+ "model.visual.blocks.8.mlp.linear_fc2",
73
+ "model.visual.blocks.9.attn.qkv",
74
+ "model.visual.blocks.9.attn.proj",
75
+ "model.visual.blocks.9.mlp.linear_fc1",
76
+ "model.visual.blocks.9.mlp.linear_fc2",
77
+ "model.visual.blocks.10.attn.qkv",
78
+ "model.visual.blocks.10.attn.proj",
79
+ "model.visual.blocks.10.mlp.linear_fc1",
80
+ "model.visual.blocks.10.mlp.linear_fc2",
81
+ "model.visual.blocks.11.attn.qkv",
82
+ "model.visual.blocks.11.attn.proj",
83
+ "model.visual.blocks.11.mlp.linear_fc1",
84
+ "model.visual.blocks.11.mlp.linear_fc2",
85
+ "model.visual.blocks.12.attn.qkv",
86
+ "model.visual.blocks.12.attn.proj",
87
+ "model.visual.blocks.12.mlp.linear_fc1",
88
+ "model.visual.blocks.12.mlp.linear_fc2",
89
+ "model.visual.blocks.13.attn.qkv",
90
+ "model.visual.blocks.13.attn.proj",
91
+ "model.visual.blocks.13.mlp.linear_fc1",
92
+ "model.visual.blocks.13.mlp.linear_fc2",
93
+ "model.visual.blocks.14.attn.qkv",
94
+ "model.visual.blocks.14.attn.proj",
95
+ "model.visual.blocks.14.mlp.linear_fc1",
96
+ "model.visual.blocks.14.mlp.linear_fc2",
97
+ "model.visual.blocks.15.attn.qkv",
98
+ "model.visual.blocks.15.attn.proj",
99
+ "model.visual.blocks.15.mlp.linear_fc1",
100
+ "model.visual.blocks.15.mlp.linear_fc2",
101
+ "model.visual.blocks.16.attn.qkv",
102
+ "model.visual.blocks.16.attn.proj",
103
+ "model.visual.blocks.16.mlp.linear_fc1",
104
+ "model.visual.blocks.16.mlp.linear_fc2",
105
+ "model.visual.blocks.17.attn.qkv",
106
+ "model.visual.blocks.17.attn.proj",
107
+ "model.visual.blocks.17.mlp.linear_fc1",
108
+ "model.visual.blocks.17.mlp.linear_fc2",
109
+ "model.visual.blocks.18.attn.qkv",
110
+ "model.visual.blocks.18.attn.proj",
111
+ "model.visual.blocks.18.mlp.linear_fc1",
112
+ "model.visual.blocks.18.mlp.linear_fc2",
113
+ "model.visual.blocks.19.attn.qkv",
114
+ "model.visual.blocks.19.attn.proj",
115
+ "model.visual.blocks.19.mlp.linear_fc1",
116
+ "model.visual.blocks.19.mlp.linear_fc2",
117
+ "model.visual.blocks.20.attn.qkv",
118
+ "model.visual.blocks.20.attn.proj",
119
+ "model.visual.blocks.20.mlp.linear_fc1",
120
+ "model.visual.blocks.20.mlp.linear_fc2",
121
+ "model.visual.blocks.21.attn.qkv",
122
+ "model.visual.blocks.21.attn.proj",
123
+ "model.visual.blocks.21.mlp.linear_fc1",
124
+ "model.visual.blocks.21.mlp.linear_fc2",
125
+ "model.visual.blocks.22.attn.qkv",
126
+ "model.visual.blocks.22.attn.proj",
127
+ "model.visual.blocks.22.mlp.linear_fc1",
128
+ "model.visual.blocks.22.mlp.linear_fc2",
129
+ "model.visual.blocks.23.attn.qkv",
130
+ "model.visual.blocks.23.attn.proj",
131
+ "model.visual.blocks.23.mlp.linear_fc1",
132
+ "model.visual.blocks.23.mlp.linear_fc2",
133
+ "model.visual.blocks.24.attn.qkv",
134
+ "model.visual.blocks.24.attn.proj",
135
+ "model.visual.blocks.24.mlp.linear_fc1",
136
+ "model.visual.blocks.24.mlp.linear_fc2",
137
+ "model.visual.blocks.25.attn.qkv",
138
+ "model.visual.blocks.25.attn.proj",
139
+ "model.visual.blocks.25.mlp.linear_fc1",
140
+ "model.visual.blocks.25.mlp.linear_fc2",
141
+ "model.visual.blocks.26.attn.qkv",
142
+ "model.visual.blocks.26.attn.proj",
143
+ "model.visual.blocks.26.mlp.linear_fc1",
144
+ "model.visual.blocks.26.mlp.linear_fc2",
145
+ "model.visual.merger.linear_fc1",
146
+ "model.visual.merger.linear_fc2",
147
+ "model.language_model.layers.0.linear_attn.out_proj",
148
+ "model.language_model.layers.0.linear_attn.in_proj_qkv",
149
+ "model.language_model.layers.0.linear_attn.in_proj_z",
150
+ "model.language_model.layers.0.linear_attn.in_proj_b",
151
+ "model.language_model.layers.0.linear_attn.in_proj_a",
152
+ "model.language_model.layers.0.linear_attn.in_proj_ba",
153
+ "model.language_model.layers.0.mlp.shared_expert.gate_proj",
154
+ "model.language_model.layers.0.mlp.shared_expert.up_proj",
155
+ "model.language_model.layers.0.mlp.shared_expert.down_proj",
156
+ "model.language_model.layers.0.mlp.shared_expert_gate",
157
+ "model.language_model.layers.1.linear_attn.out_proj",
158
+ "model.language_model.layers.1.linear_attn.in_proj_qkv",
159
+ "model.language_model.layers.1.linear_attn.in_proj_z",
160
+ "model.language_model.layers.1.linear_attn.in_proj_b",
161
+ "model.language_model.layers.1.linear_attn.in_proj_a",
162
+ "model.language_model.layers.1.linear_attn.in_proj_ba",
163
+ "model.language_model.layers.1.mlp.shared_expert.gate_proj",
164
+ "model.language_model.layers.1.mlp.shared_expert.up_proj",
165
+ "model.language_model.layers.1.mlp.shared_expert.down_proj",
166
+ "model.language_model.layers.1.mlp.shared_expert_gate",
167
+ "model.language_model.layers.2.linear_attn.out_proj",
168
+ "model.language_model.layers.2.linear_attn.in_proj_qkv",
169
+ "model.language_model.layers.2.linear_attn.in_proj_z",
170
+ "model.language_model.layers.2.linear_attn.in_proj_b",
171
+ "model.language_model.layers.2.linear_attn.in_proj_a",
172
+ "model.language_model.layers.2.linear_attn.in_proj_ba",
173
+ "model.language_model.layers.2.mlp.shared_expert.gate_proj",
174
+ "model.language_model.layers.2.mlp.shared_expert.up_proj",
175
+ "model.language_model.layers.2.mlp.shared_expert.down_proj",
176
+ "model.language_model.layers.2.mlp.shared_expert_gate",
177
+ "model.language_model.layers.3.self_attn.q_proj",
178
+ "model.language_model.layers.3.self_attn.k_proj",
179
+ "model.language_model.layers.3.self_attn.v_proj",
180
+ "model.language_model.layers.3.self_attn.o_proj",
181
+ "model.language_model.layers.3.mlp.shared_expert.gate_proj",
182
+ "model.language_model.layers.3.mlp.shared_expert.up_proj",
183
+ "model.language_model.layers.3.mlp.shared_expert.down_proj",
184
+ "model.language_model.layers.3.mlp.shared_expert_gate",
185
+ "model.language_model.layers.4.linear_attn.out_proj",
186
+ "model.language_model.layers.4.linear_attn.in_proj_qkv",
187
+ "model.language_model.layers.4.linear_attn.in_proj_z",
188
+ "model.language_model.layers.4.linear_attn.in_proj_b",
189
+ "model.language_model.layers.4.linear_attn.in_proj_a",
190
+ "model.language_model.layers.4.linear_attn.in_proj_ba",
191
+ "model.language_model.layers.4.mlp.shared_expert.gate_proj",
192
+ "model.language_model.layers.4.mlp.shared_expert.up_proj",
193
+ "model.language_model.layers.4.mlp.shared_expert.down_proj",
194
+ "model.language_model.layers.4.mlp.shared_expert_gate",
195
+ "model.language_model.layers.5.linear_attn.out_proj",
196
+ "model.language_model.layers.5.linear_attn.in_proj_qkv",
197
+ "model.language_model.layers.5.linear_attn.in_proj_z",
198
+ "model.language_model.layers.5.linear_attn.in_proj_b",
199
+ "model.language_model.layers.5.linear_attn.in_proj_a",
200
+ "model.language_model.layers.5.linear_attn.in_proj_ba",
201
+ "model.language_model.layers.5.mlp.shared_expert.gate_proj",
202
+ "model.language_model.layers.5.mlp.shared_expert.up_proj",
203
+ "model.language_model.layers.5.mlp.shared_expert.down_proj",
204
+ "model.language_model.layers.5.mlp.shared_expert_gate",
205
+ "model.language_model.layers.6.linear_attn.out_proj",
206
+ "model.language_model.layers.6.linear_attn.in_proj_qkv",
207
+ "model.language_model.layers.6.linear_attn.in_proj_z",
208
+ "model.language_model.layers.6.linear_attn.in_proj_b",
209
+ "model.language_model.layers.6.linear_attn.in_proj_a",
210
+ "model.language_model.layers.6.linear_attn.in_proj_ba",
211
+ "model.language_model.layers.6.mlp.shared_expert.gate_proj",
212
+ "model.language_model.layers.6.mlp.shared_expert.up_proj",
213
+ "model.language_model.layers.6.mlp.shared_expert.down_proj",
214
+ "model.language_model.layers.6.mlp.shared_expert_gate",
215
+ "model.language_model.layers.7.self_attn.q_proj",
216
+ "model.language_model.layers.7.self_attn.k_proj",
217
+ "model.language_model.layers.7.self_attn.v_proj",
218
+ "model.language_model.layers.7.self_attn.o_proj",
219
+ "model.language_model.layers.7.mlp.shared_expert.gate_proj",
220
+ "model.language_model.layers.7.mlp.shared_expert.up_proj",
221
+ "model.language_model.layers.7.mlp.shared_expert.down_proj",
222
+ "model.language_model.layers.7.mlp.shared_expert_gate",
223
+ "model.language_model.layers.8.linear_attn.out_proj",
224
+ "model.language_model.layers.8.linear_attn.in_proj_qkv",
225
+ "model.language_model.layers.8.linear_attn.in_proj_z",
226
+ "model.language_model.layers.8.linear_attn.in_proj_b",
227
+ "model.language_model.layers.8.linear_attn.in_proj_a",
228
+ "model.language_model.layers.8.linear_attn.in_proj_ba",
229
+ "model.language_model.layers.8.mlp.shared_expert.gate_proj",
230
+ "model.language_model.layers.8.mlp.shared_expert.up_proj",
231
+ "model.language_model.layers.8.mlp.shared_expert.down_proj",
232
+ "model.language_model.layers.8.mlp.shared_expert_gate",
233
+ "model.language_model.layers.9.linear_attn.out_proj",
234
+ "model.language_model.layers.9.linear_attn.in_proj_qkv",
235
+ "model.language_model.layers.9.linear_attn.in_proj_z",
236
+ "model.language_model.layers.9.linear_attn.in_proj_b",
237
+ "model.language_model.layers.9.linear_attn.in_proj_a",
238
+ "model.language_model.layers.9.linear_attn.in_proj_ba",
239
+ "model.language_model.layers.9.mlp.shared_expert.gate_proj",
240
+ "model.language_model.layers.9.mlp.shared_expert.up_proj",
241
+ "model.language_model.layers.9.mlp.shared_expert.down_proj",
242
+ "model.language_model.layers.9.mlp.shared_expert_gate",
243
+ "model.language_model.layers.10.linear_attn.out_proj",
244
+ "model.language_model.layers.10.linear_attn.in_proj_qkv",
245
+ "model.language_model.layers.10.linear_attn.in_proj_z",
246
+ "model.language_model.layers.10.linear_attn.in_proj_b",
247
+ "model.language_model.layers.10.linear_attn.in_proj_a",
248
+ "model.language_model.layers.10.linear_attn.in_proj_ba",
249
+ "model.language_model.layers.10.mlp.shared_expert.gate_proj",
250
+ "model.language_model.layers.10.mlp.shared_expert.up_proj",
251
+ "model.language_model.layers.10.mlp.shared_expert.down_proj",
252
+ "model.language_model.layers.10.mlp.shared_expert_gate",
253
+ "model.language_model.layers.11.self_attn.q_proj",
254
+ "model.language_model.layers.11.self_attn.k_proj",
255
+ "model.language_model.layers.11.self_attn.v_proj",
256
+ "model.language_model.layers.11.self_attn.o_proj",
257
+ "model.language_model.layers.11.mlp.shared_expert.gate_proj",
258
+ "model.language_model.layers.11.mlp.shared_expert.up_proj",
259
+ "model.language_model.layers.11.mlp.shared_expert.down_proj",
260
+ "model.language_model.layers.11.mlp.shared_expert_gate",
261
+ "model.language_model.layers.12.linear_attn.out_proj",
262
+ "model.language_model.layers.12.linear_attn.in_proj_qkv",
263
+ "model.language_model.layers.12.linear_attn.in_proj_z",
264
+ "model.language_model.layers.12.linear_attn.in_proj_b",
265
+ "model.language_model.layers.12.linear_attn.in_proj_a",
266
+ "model.language_model.layers.12.linear_attn.in_proj_ba",
267
+ "model.language_model.layers.12.mlp.shared_expert.gate_proj",
268
+ "model.language_model.layers.12.mlp.shared_expert.up_proj",
269
+ "model.language_model.layers.12.mlp.shared_expert.down_proj",
270
+ "model.language_model.layers.12.mlp.shared_expert_gate",
271
+ "model.language_model.layers.13.linear_attn.out_proj",
272
+ "model.language_model.layers.13.linear_attn.in_proj_qkv",
273
+ "model.language_model.layers.13.linear_attn.in_proj_z",
274
+ "model.language_model.layers.13.linear_attn.in_proj_b",
275
+ "model.language_model.layers.13.linear_attn.in_proj_a",
276
+ "model.language_model.layers.13.linear_attn.in_proj_ba",
277
+ "model.language_model.layers.13.mlp.shared_expert.gate_proj",
278
+ "model.language_model.layers.13.mlp.shared_expert.up_proj",
279
+ "model.language_model.layers.13.mlp.shared_expert.down_proj",
280
+ "model.language_model.layers.13.mlp.shared_expert_gate",
281
+ "model.language_model.layers.14.linear_attn.out_proj",
282
+ "model.language_model.layers.14.linear_attn.in_proj_qkv",
283
+ "model.language_model.layers.14.linear_attn.in_proj_z",
284
+ "model.language_model.layers.14.linear_attn.in_proj_b",
285
+ "model.language_model.layers.14.linear_attn.in_proj_a",
286
+ "model.language_model.layers.14.linear_attn.in_proj_ba",
287
+ "model.language_model.layers.14.mlp.shared_expert.gate_proj",
288
+ "model.language_model.layers.14.mlp.shared_expert.up_proj",
289
+ "model.language_model.layers.14.mlp.shared_expert.down_proj",
290
+ "model.language_model.layers.14.mlp.shared_expert_gate",
291
+ "model.language_model.layers.15.self_attn.q_proj",
292
+ "model.language_model.layers.15.self_attn.k_proj",
293
+ "model.language_model.layers.15.self_attn.v_proj",
294
+ "model.language_model.layers.15.self_attn.o_proj",
295
+ "model.language_model.layers.15.mlp.shared_expert.gate_proj",
296
+ "model.language_model.layers.15.mlp.shared_expert.up_proj",
297
+ "model.language_model.layers.15.mlp.shared_expert.down_proj",
298
+ "model.language_model.layers.15.mlp.shared_expert_gate",
299
+ "model.language_model.layers.16.linear_attn.out_proj",
300
+ "model.language_model.layers.16.linear_attn.in_proj_qkv",
301
+ "model.language_model.layers.16.linear_attn.in_proj_z",
302
+ "model.language_model.layers.16.linear_attn.in_proj_b",
303
+ "model.language_model.layers.16.linear_attn.in_proj_a",
304
+ "model.language_model.layers.16.linear_attn.in_proj_ba",
305
+ "model.language_model.layers.16.mlp.shared_expert.gate_proj",
306
+ "model.language_model.layers.16.mlp.shared_expert.up_proj",
307
+ "model.language_model.layers.16.mlp.shared_expert.down_proj",
308
+ "model.language_model.layers.16.mlp.shared_expert_gate",
309
+ "model.language_model.layers.17.linear_attn.out_proj",
310
+ "model.language_model.layers.17.linear_attn.in_proj_qkv",
311
+ "model.language_model.layers.17.linear_attn.in_proj_z",
312
+ "model.language_model.layers.17.linear_attn.in_proj_b",
313
+ "model.language_model.layers.17.linear_attn.in_proj_a",
314
+ "model.language_model.layers.17.linear_attn.in_proj_ba",
315
+ "model.language_model.layers.17.mlp.shared_expert.gate_proj",
316
+ "model.language_model.layers.17.mlp.shared_expert.up_proj",
317
+ "model.language_model.layers.17.mlp.shared_expert.down_proj",
318
+ "model.language_model.layers.17.mlp.shared_expert_gate",
319
+ "model.language_model.layers.18.linear_attn.out_proj",
320
+ "model.language_model.layers.18.linear_attn.in_proj_qkv",
321
+ "model.language_model.layers.18.linear_attn.in_proj_z",
322
+ "model.language_model.layers.18.linear_attn.in_proj_b",
323
+ "model.language_model.layers.18.linear_attn.in_proj_a",
324
+ "model.language_model.layers.18.linear_attn.in_proj_ba",
325
+ "model.language_model.layers.18.mlp.shared_expert.gate_proj",
326
+ "model.language_model.layers.18.mlp.shared_expert.up_proj",
327
+ "model.language_model.layers.18.mlp.shared_expert.down_proj",
328
+ "model.language_model.layers.18.mlp.shared_expert_gate",
329
+ "model.language_model.layers.19.self_attn.q_proj",
330
+ "model.language_model.layers.19.self_attn.k_proj",
331
+ "model.language_model.layers.19.self_attn.v_proj",
332
+ "model.language_model.layers.19.self_attn.o_proj",
333
+ "model.language_model.layers.19.mlp.shared_expert.gate_proj",
334
+ "model.language_model.layers.19.mlp.shared_expert.up_proj",
335
+ "model.language_model.layers.19.mlp.shared_expert.down_proj",
336
+ "model.language_model.layers.19.mlp.shared_expert_gate",
337
+ "model.language_model.layers.20.linear_attn.out_proj",
338
+ "model.language_model.layers.20.linear_attn.in_proj_qkv",
339
+ "model.language_model.layers.20.linear_attn.in_proj_z",
340
+ "model.language_model.layers.20.linear_attn.in_proj_b",
341
+ "model.language_model.layers.20.linear_attn.in_proj_a",
342
+ "model.language_model.layers.20.linear_attn.in_proj_ba",
343
+ "model.language_model.layers.20.mlp.shared_expert.gate_proj",
344
+ "model.language_model.layers.20.mlp.shared_expert.up_proj",
345
+ "model.language_model.layers.20.mlp.shared_expert.down_proj",
346
+ "model.language_model.layers.20.mlp.shared_expert_gate",
347
+ "model.language_model.layers.21.linear_attn.out_proj",
348
+ "model.language_model.layers.21.linear_attn.in_proj_qkv",
349
+ "model.language_model.layers.21.linear_attn.in_proj_z",
350
+ "model.language_model.layers.21.linear_attn.in_proj_b",
351
+ "model.language_model.layers.21.linear_attn.in_proj_a",
352
+ "model.language_model.layers.21.linear_attn.in_proj_ba",
353
+ "model.language_model.layers.21.mlp.shared_expert.gate_proj",
354
+ "model.language_model.layers.21.mlp.shared_expert.up_proj",
355
+ "model.language_model.layers.21.mlp.shared_expert.down_proj",
356
+ "model.language_model.layers.21.mlp.shared_expert_gate",
357
+ "model.language_model.layers.22.linear_attn.out_proj",
358
+ "model.language_model.layers.22.linear_attn.in_proj_qkv",
359
+ "model.language_model.layers.22.linear_attn.in_proj_z",
360
+ "model.language_model.layers.22.linear_attn.in_proj_b",
361
+ "model.language_model.layers.22.linear_attn.in_proj_a",
362
+ "model.language_model.layers.22.linear_attn.in_proj_ba",
363
+ "model.language_model.layers.22.mlp.shared_expert.gate_proj",
364
+ "model.language_model.layers.22.mlp.shared_expert.up_proj",
365
+ "model.language_model.layers.22.mlp.shared_expert.down_proj",
366
+ "model.language_model.layers.22.mlp.shared_expert_gate",
367
+ "model.language_model.layers.23.self_attn.q_proj",
368
+ "model.language_model.layers.23.self_attn.k_proj",
369
+ "model.language_model.layers.23.self_attn.v_proj",
370
+ "model.language_model.layers.23.self_attn.o_proj",
371
+ "model.language_model.layers.23.mlp.shared_expert.gate_proj",
372
+ "model.language_model.layers.23.mlp.shared_expert.up_proj",
373
+ "model.language_model.layers.23.mlp.shared_expert.down_proj",
374
+ "model.language_model.layers.23.mlp.shared_expert_gate",
375
+ "model.language_model.layers.24.linear_attn.out_proj",
376
+ "model.language_model.layers.24.linear_attn.in_proj_qkv",
377
+ "model.language_model.layers.24.linear_attn.in_proj_z",
378
+ "model.language_model.layers.24.linear_attn.in_proj_b",
379
+ "model.language_model.layers.24.linear_attn.in_proj_a",
380
+ "model.language_model.layers.24.linear_attn.in_proj_ba",
381
+ "model.language_model.layers.24.mlp.shared_expert.gate_proj",
382
+ "model.language_model.layers.24.mlp.shared_expert.up_proj",
383
+ "model.language_model.layers.24.mlp.shared_expert.down_proj",
384
+ "model.language_model.layers.24.mlp.shared_expert_gate",
385
+ "model.language_model.layers.25.linear_attn.out_proj",
386
+ "model.language_model.layers.25.linear_attn.in_proj_qkv",
387
+ "model.language_model.layers.25.linear_attn.in_proj_z",
388
+ "model.language_model.layers.25.linear_attn.in_proj_b",
389
+ "model.language_model.layers.25.linear_attn.in_proj_a",
390
+ "model.language_model.layers.25.linear_attn.in_proj_ba",
391
+ "model.language_model.layers.25.mlp.shared_expert.gate_proj",
392
+ "model.language_model.layers.25.mlp.shared_expert.up_proj",
393
+ "model.language_model.layers.25.mlp.shared_expert.down_proj",
394
+ "model.language_model.layers.25.mlp.shared_expert_gate",
395
+ "model.language_model.layers.26.linear_attn.out_proj",
396
+ "model.language_model.layers.26.linear_attn.in_proj_qkv",
397
+ "model.language_model.layers.26.linear_attn.in_proj_z",
398
+ "model.language_model.layers.26.linear_attn.in_proj_b",
399
+ "model.language_model.layers.26.linear_attn.in_proj_a",
400
+ "model.language_model.layers.26.linear_attn.in_proj_ba",
401
+ "model.language_model.layers.26.mlp.shared_expert.gate_proj",
402
+ "model.language_model.layers.26.mlp.shared_expert.up_proj",
403
+ "model.language_model.layers.26.mlp.shared_expert.down_proj",
404
+ "model.language_model.layers.26.mlp.shared_expert_gate",
405
+ "model.language_model.layers.27.self_attn.q_proj",
406
+ "model.language_model.layers.27.self_attn.k_proj",
407
+ "model.language_model.layers.27.self_attn.v_proj",
408
+ "model.language_model.layers.27.self_attn.o_proj",
409
+ "model.language_model.layers.27.mlp.shared_expert.gate_proj",
410
+ "model.language_model.layers.27.mlp.shared_expert.up_proj",
411
+ "model.language_model.layers.27.mlp.shared_expert.down_proj",
412
+ "model.language_model.layers.27.mlp.shared_expert_gate",
413
+ "model.language_model.layers.28.linear_attn.out_proj",
414
+ "model.language_model.layers.28.linear_attn.in_proj_qkv",
415
+ "model.language_model.layers.28.linear_attn.in_proj_z",
416
+ "model.language_model.layers.28.linear_attn.in_proj_b",
417
+ "model.language_model.layers.28.linear_attn.in_proj_a",
418
+ "model.language_model.layers.28.linear_attn.in_proj_ba",
419
+ "model.language_model.layers.28.mlp.shared_expert.gate_proj",
420
+ "model.language_model.layers.28.mlp.shared_expert.up_proj",
421
+ "model.language_model.layers.28.mlp.shared_expert.down_proj",
422
+ "model.language_model.layers.28.mlp.shared_expert_gate",
423
+ "model.language_model.layers.29.linear_attn.out_proj",
424
+ "model.language_model.layers.29.linear_attn.in_proj_qkv",
425
+ "model.language_model.layers.29.linear_attn.in_proj_z",
426
+ "model.language_model.layers.29.linear_attn.in_proj_b",
427
+ "model.language_model.layers.29.linear_attn.in_proj_a",
428
+ "model.language_model.layers.29.linear_attn.in_proj_ba",
429
+ "model.language_model.layers.29.mlp.shared_expert.gate_proj",
430
+ "model.language_model.layers.29.mlp.shared_expert.up_proj",
431
+ "model.language_model.layers.29.mlp.shared_expert.down_proj",
432
+ "model.language_model.layers.29.mlp.shared_expert_gate",
433
+ "model.language_model.layers.30.linear_attn.out_proj",
434
+ "model.language_model.layers.30.linear_attn.in_proj_qkv",
435
+ "model.language_model.layers.30.linear_attn.in_proj_z",
436
+ "model.language_model.layers.30.linear_attn.in_proj_b",
437
+ "model.language_model.layers.30.linear_attn.in_proj_a",
438
+ "model.language_model.layers.30.linear_attn.in_proj_ba",
439
+ "model.language_model.layers.30.mlp.shared_expert.gate_proj",
440
+ "model.language_model.layers.30.mlp.shared_expert.up_proj",
441
+ "model.language_model.layers.30.mlp.shared_expert.down_proj",
442
+ "model.language_model.layers.30.mlp.shared_expert_gate",
443
+ "model.language_model.layers.31.self_attn.q_proj",
444
+ "model.language_model.layers.31.self_attn.k_proj",
445
+ "model.language_model.layers.31.self_attn.v_proj",
446
+ "model.language_model.layers.31.self_attn.o_proj",
447
+ "model.language_model.layers.31.mlp.shared_expert.gate_proj",
448
+ "model.language_model.layers.31.mlp.shared_expert.up_proj",
449
+ "model.language_model.layers.31.mlp.shared_expert.down_proj",
450
+ "model.language_model.layers.31.mlp.shared_expert_gate",
451
+ "model.language_model.layers.32.linear_attn.out_proj",
452
+ "model.language_model.layers.32.linear_attn.in_proj_qkv",
453
+ "model.language_model.layers.32.linear_attn.in_proj_z",
454
+ "model.language_model.layers.32.linear_attn.in_proj_b",
455
+ "model.language_model.layers.32.linear_attn.in_proj_a",
456
+ "model.language_model.layers.32.linear_attn.in_proj_ba",
457
+ "model.language_model.layers.32.mlp.shared_expert.gate_proj",
458
+ "model.language_model.layers.32.mlp.shared_expert.up_proj",
459
+ "model.language_model.layers.32.mlp.shared_expert.down_proj",
460
+ "model.language_model.layers.32.mlp.shared_expert_gate",
461
+ "model.language_model.layers.33.linear_attn.out_proj",
462
+ "model.language_model.layers.33.linear_attn.in_proj_qkv",
463
+ "model.language_model.layers.33.linear_attn.in_proj_z",
464
+ "model.language_model.layers.33.linear_attn.in_proj_b",
465
+ "model.language_model.layers.33.linear_attn.in_proj_a",
466
+ "model.language_model.layers.33.linear_attn.in_proj_ba",
467
+ "model.language_model.layers.33.mlp.shared_expert.gate_proj",
468
+ "model.language_model.layers.33.mlp.shared_expert.up_proj",
469
+ "model.language_model.layers.33.mlp.shared_expert.down_proj",
470
+ "model.language_model.layers.33.mlp.shared_expert_gate",
471
+ "model.language_model.layers.34.linear_attn.out_proj",
472
+ "model.language_model.layers.34.linear_attn.in_proj_qkv",
473
+ "model.language_model.layers.34.linear_attn.in_proj_z",
474
+ "model.language_model.layers.34.linear_attn.in_proj_b",
475
+ "model.language_model.layers.34.linear_attn.in_proj_a",
476
+ "model.language_model.layers.34.linear_attn.in_proj_ba",
477
+ "model.language_model.layers.34.mlp.shared_expert.gate_proj",
478
+ "model.language_model.layers.34.mlp.shared_expert.up_proj",
479
+ "model.language_model.layers.34.mlp.shared_expert.down_proj",
480
+ "model.language_model.layers.34.mlp.shared_expert_gate",
481
+ "model.language_model.layers.35.self_attn.q_proj",
482
+ "model.language_model.layers.35.self_attn.k_proj",
483
+ "model.language_model.layers.35.self_attn.v_proj",
484
+ "model.language_model.layers.35.self_attn.o_proj",
485
+ "model.language_model.layers.35.mlp.shared_expert.gate_proj",
486
+ "model.language_model.layers.35.mlp.shared_expert.up_proj",
487
+ "model.language_model.layers.35.mlp.shared_expert.down_proj",
488
+ "model.language_model.layers.35.mlp.shared_expert_gate",
489
+ "model.language_model.layers.36.linear_attn.out_proj",
490
+ "model.language_model.layers.36.linear_attn.in_proj_qkv",
491
+ "model.language_model.layers.36.linear_attn.in_proj_z",
492
+ "model.language_model.layers.36.linear_attn.in_proj_b",
493
+ "model.language_model.layers.36.linear_attn.in_proj_a",
494
+ "model.language_model.layers.36.linear_attn.in_proj_ba",
495
+ "model.language_model.layers.36.mlp.shared_expert.gate_proj",
496
+ "model.language_model.layers.36.mlp.shared_expert.up_proj",
497
+ "model.language_model.layers.36.mlp.shared_expert.down_proj",
498
+ "model.language_model.layers.36.mlp.shared_expert_gate",
499
+ "model.language_model.layers.37.linear_attn.out_proj",
500
+ "model.language_model.layers.37.linear_attn.in_proj_qkv",
501
+ "model.language_model.layers.37.linear_attn.in_proj_z",
502
+ "model.language_model.layers.37.linear_attn.in_proj_b",
503
+ "model.language_model.layers.37.linear_attn.in_proj_a",
504
+ "model.language_model.layers.37.linear_attn.in_proj_ba",
505
+ "model.language_model.layers.37.mlp.shared_expert.gate_proj",
506
+ "model.language_model.layers.37.mlp.shared_expert.up_proj",
507
+ "model.language_model.layers.37.mlp.shared_expert.down_proj",
508
+ "model.language_model.layers.37.mlp.shared_expert_gate",
509
+ "model.language_model.layers.38.linear_attn.out_proj",
510
+ "model.language_model.layers.38.linear_attn.in_proj_qkv",
511
+ "model.language_model.layers.38.linear_attn.in_proj_z",
512
+ "model.language_model.layers.38.linear_attn.in_proj_b",
513
+ "model.language_model.layers.38.linear_attn.in_proj_a",
514
+ "model.language_model.layers.38.linear_attn.in_proj_ba",
515
+ "model.language_model.layers.38.mlp.shared_expert.gate_proj",
516
+ "model.language_model.layers.38.mlp.shared_expert.up_proj",
517
+ "model.language_model.layers.38.mlp.shared_expert.down_proj",
518
+ "model.language_model.layers.38.mlp.shared_expert_gate",
519
+ "model.language_model.layers.39.self_attn.q_proj",
520
+ "model.language_model.layers.39.self_attn.k_proj",
521
+ "model.language_model.layers.39.self_attn.v_proj",
522
+ "model.language_model.layers.39.self_attn.o_proj",
523
+ "model.language_model.layers.39.mlp.shared_expert.gate_proj",
524
+ "model.language_model.layers.39.mlp.shared_expert.up_proj",
525
+ "model.language_model.layers.39.mlp.shared_expert.down_proj",
526
+ "model.language_model.layers.39.mlp.shared_expert_gate",
527
+ "mtp.fc",
528
+ "mtp.layers.0.mlp.shared_expert.gate_up_proj",
529
+ "mtp.layers.0.mlp.shared_expert.gate_proj",
530
+ "mtp.layers.0.mlp.shared_expert.up_proj",
531
+ "mtp.layers.0.mlp.shared_expert.down_proj",
532
+ "mtp.layers.0.mlp.shared_expert_gate",
533
+ "mtp.layers.0.mlp.gate",
534
+ "mtp.layers.0.self_attn.k_proj",
535
+ "mtp.layers.0.self_attn.o_proj",
536
+ "mtp.layers.0.self_attn.q_proj",
537
+ "mtp.layers.0.self_attn.v_proj",
538
  "lm_head"
539
  ],
540
  "kv_cache_scheme": null,
 
542
  "quantization_status": "compressed",
543
  "sparsity_config": {},
544
  "transform_config": {},
545
+ "version": "0.1.dev489+g6d69620"
546
  },
547
+ "text_config": {
548
+ "attention_bias": false,
549
+ "attention_dropout": 0.0,
550
+ "attn_output_gate": true,
551
+ "bos_token_id": null,
552
+ "dtype": "bfloat16",
553
+ "eos_token_id": 248044,
554
+ "full_attention_interval": 4,
555
+ "head_dim": 256,
556
+ "hidden_act": "silu",
557
+ "hidden_size": 2048,
558
+ "initializer_range": 0.02,
559
+ "layer_types": [
560
+ "linear_attention",
561
+ "linear_attention",
562
+ "linear_attention",
563
+ "full_attention",
564
+ "linear_attention",
565
+ "linear_attention",
566
+ "linear_attention",
567
+ "full_attention",
568
+ "linear_attention",
569
+ "linear_attention",
570
+ "linear_attention",
571
+ "full_attention",
572
+ "linear_attention",
573
+ "linear_attention",
574
+ "linear_attention",
575
+ "full_attention",
576
+ "linear_attention",
577
+ "linear_attention",
578
+ "linear_attention",
579
+ "full_attention",
580
+ "linear_attention",
581
+ "linear_attention",
582
+ "linear_attention",
583
+ "full_attention",
584
+ "linear_attention",
585
+ "linear_attention",
586
+ "linear_attention",
587
+ "full_attention",
588
+ "linear_attention",
589
+ "linear_attention",
590
+ "linear_attention",
591
+ "full_attention",
592
+ "linear_attention",
593
+ "linear_attention",
594
+ "linear_attention",
595
+ "full_attention",
596
+ "linear_attention",
597
+ "linear_attention",
598
+ "linear_attention",
599
+ "full_attention"
600
  ],
601
+ "linear_conv_kernel_dim": 4,
602
+ "linear_key_head_dim": 128,
603
+ "linear_num_key_heads": 16,
604
+ "linear_num_value_heads": 32,
605
+ "linear_value_head_dim": 128,
606
+ "mamba_ssm_dtype": "float32",
607
+ "max_position_embeddings": 262144,
608
+ "mlp_only_layers": [],
609
+ "model_type": "qwen3_5_moe_text",
610
+ "moe_intermediate_size": 512,
611
+ "mtp_num_hidden_layers": 1,
612
+ "mtp_use_dedicated_embeddings": false,
613
+ "num_attention_heads": 16,
614
+ "num_experts": 256,
615
+ "num_experts_per_tok": 8,
616
+ "num_hidden_layers": 40,
617
+ "num_key_value_heads": 2,
618
+ "output_router_logits": false,
619
+ "pad_token_id": null,
620
  "partial_rotary_factor": 0.25,
621
+ "rms_norm_eps": 1e-06,
622
+ "rope_parameters": {
623
+ "mrope_interleaved": true,
624
+ "mrope_section": [
625
+ 11,
626
+ 11,
627
+ 10
628
+ ],
629
+ "partial_rotary_factor": 0.25,
630
+ "rope_theta": 10000000,
631
+ "rope_type": "default"
632
+ },
633
+ "router_aux_loss_coef": 0.001,
634
+ "shared_expert_intermediate_size": 512,
635
+ "tie_word_embeddings": false,
636
+ "use_cache": true,
637
+ "vocab_size": 248320
638
  },
 
 
639
  "tie_word_embeddings": false,
640
+ "transformers_version": "5.2.0",
641
+ "video_token_id": 248057,
642
+ "vision_config": {
643
+ "deepstack_visual_indexes": [],
644
+ "depth": 27,
645
+ "dtype": "bfloat16",
646
+ "hidden_act": "gelu_pytorch_tanh",
647
+ "hidden_size": 1152,
648
+ "in_channels": 3,
649
+ "initializer_range": 0.02,
650
+ "intermediate_size": 4304,
651
+ "model_type": "qwen3_5_moe",
652
+ "num_heads": 16,
653
+ "num_position_embeddings": 2304,
654
+ "out_hidden_size": 2048,
655
+ "patch_size": 16,
656
+ "spatial_merge_size": 2,
657
+ "temporal_patch_size": 2
658
+ },
659
+ "vision_end_token_id": 248054,
660
+ "vision_start_token_id": 248053
661
  }