festr2 commited on
Commit
977b21d
·
verified ·
1 Parent(s): 4c102b6

Upload config.json

Browse files
Files changed (1) hide show
  1. config.json +600 -0
config.json ADDED
@@ -0,0 +1,600 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GlmMoeDsaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "dtype": "bfloat16",
8
+ "eos_token_id": [
9
+ 154820,
10
+ 154827,
11
+ 154829
12
+ ],
13
+ "ep_size": 1,
14
+ "first_k_dense_replace": 3,
15
+ "head_dim": 64,
16
+ "hidden_act": "silu",
17
+ "hidden_size": 6144,
18
+ "index_head_dim": 128,
19
+ "index_n_heads": 32,
20
+ "index_topk": 2048,
21
+ "indexer_rope_interleave": true,
22
+ "initializer_range": 0.02,
23
+ "intermediate_size": 12288,
24
+ "kv_lora_rank": 512,
25
+ "max_position_embeddings": 202752,
26
+ "model_type": "glm_moe_dsa",
27
+ "moe_intermediate_size": 2048,
28
+ "moe_layer_freq": 1,
29
+ "n_group": 1,
30
+ "n_routed_experts": 256,
31
+ "n_shared_experts": 1,
32
+ "norm_topk_prob": true,
33
+ "num_attention_heads": 64,
34
+ "num_experts_per_tok": 8,
35
+ "num_hidden_layers": 78,
36
+ "num_key_value_heads": 64,
37
+ "num_nextn_predict_layers": 1,
38
+ "pad_token_id": 154820,
39
+ "pretraining_tp": 1,
40
+ "q_lora_rank": 2048,
41
+ "qk_head_dim": 256,
42
+ "qk_nope_head_dim": 192,
43
+ "qk_rope_head_dim": 64,
44
+ "quantization_config": {
45
+ "config_groups": {
46
+ "group_0": {
47
+ "input_activations": {
48
+ "dynamic": false,
49
+ "group_size": 16,
50
+ "num_bits": 4,
51
+ "type": "float"
52
+ },
53
+ "targets": [
54
+ "Linear"
55
+ ],
56
+ "weights": {
57
+ "dynamic": false,
58
+ "group_size": 16,
59
+ "num_bits": 4,
60
+ "type": "float"
61
+ }
62
+ },
63
+ "group_mxfp8_sparse_experts": {
64
+ "input_activations": {
65
+ "dynamic": true,
66
+ "group_size": 32,
67
+ "num_bits": 8,
68
+ "type": "float"
69
+ },
70
+ "targets": [
71
+ "model.layers.42.mlp.experts",
72
+ "model.layers.43.mlp.experts",
73
+ "model.layers.44.mlp.experts",
74
+ "model.layers.45.mlp.experts",
75
+ "model.layers.46.mlp.experts",
76
+ "model.layers.47.mlp.experts",
77
+ "model.layers.48.mlp.experts",
78
+ "model.layers.49.mlp.experts",
79
+ "model.layers.50.mlp.experts",
80
+ "model.layers.51.mlp.experts",
81
+ "model.layers.52.mlp.experts",
82
+ "model.layers.53.mlp.experts",
83
+ "model.layers.54.mlp.experts",
84
+ "model.layers.55.mlp.experts",
85
+ "model.layers.56.mlp.experts",
86
+ "model.layers.57.mlp.experts",
87
+ "model.layers.58.mlp.experts",
88
+ "model.layers.59.mlp.experts",
89
+ "model.layers.60.mlp.experts",
90
+ "model.layers.61.mlp.experts",
91
+ "model.layers.62.mlp.experts"
92
+ ],
93
+ "weights": {
94
+ "dynamic": false,
95
+ "group_size": 32,
96
+ "num_bits": 8,
97
+ "type": "float",
98
+ "weight_block_size": [
99
+ 1,
100
+ 32
101
+ ]
102
+ }
103
+ }
104
+ },
105
+ "ignore": [
106
+ "lm_head",
107
+ "model.layers.0.mlp*",
108
+ "model.layers.0.self_attn*",
109
+ "model.layers.0.self_attn.indexer*",
110
+ "model.layers.1.mlp*",
111
+ "model.layers.1.self_attn*",
112
+ "model.layers.1.self_attn.indexer*",
113
+ "*shared_experts*",
114
+ "model.layers.10.self_attn*",
115
+ "model.layers.10.self_attn.indexer*",
116
+ "model.layers.11.self_attn*",
117
+ "model.layers.11.self_attn.indexer*",
118
+ "model.layers.12.self_attn*",
119
+ "model.layers.12.self_attn.indexer*",
120
+ "model.layers.13.self_attn*",
121
+ "model.layers.13.self_attn.indexer*",
122
+ "model.layers.14.self_attn*",
123
+ "model.layers.14.self_attn.indexer*",
124
+ "model.layers.15.self_attn*",
125
+ "model.layers.15.self_attn.indexer*",
126
+ "model.layers.16.self_attn*",
127
+ "model.layers.16.self_attn.indexer*",
128
+ "model.layers.17.self_attn*",
129
+ "model.layers.17.self_attn.indexer*",
130
+ "model.layers.18.self_attn*",
131
+ "model.layers.18.self_attn.indexer*",
132
+ "model.layers.19.self_attn*",
133
+ "model.layers.19.self_attn.indexer*",
134
+ "model.layers.2.mlp*",
135
+ "model.layers.2.self_attn*",
136
+ "model.layers.2.self_attn.indexer*",
137
+ "model.layers.20.self_attn*",
138
+ "model.layers.20.self_attn.indexer*",
139
+ "model.layers.21.self_attn*",
140
+ "model.layers.21.self_attn.indexer*",
141
+ "model.layers.22.self_attn*",
142
+ "model.layers.22.self_attn.indexer*",
143
+ "model.layers.23.self_attn*",
144
+ "model.layers.23.self_attn.indexer*",
145
+ "model.layers.24.self_attn*",
146
+ "model.layers.24.self_attn.indexer*",
147
+ "model.layers.25.self_attn*",
148
+ "model.layers.25.self_attn.indexer*",
149
+ "model.layers.26.self_attn*",
150
+ "model.layers.26.self_attn.indexer*",
151
+ "model.layers.27.self_attn*",
152
+ "model.layers.27.self_attn.indexer*",
153
+ "model.layers.28.self_attn*",
154
+ "model.layers.28.self_attn.indexer*",
155
+ "model.layers.29.self_attn*",
156
+ "model.layers.29.self_attn.indexer*",
157
+ "model.layers.3.self_attn*",
158
+ "model.layers.3.self_attn.indexer*",
159
+ "model.layers.30.self_attn*",
160
+ "model.layers.30.self_attn.indexer*",
161
+ "model.layers.31.self_attn*",
162
+ "model.layers.31.self_attn.indexer*",
163
+ "model.layers.32.self_attn*",
164
+ "model.layers.32.self_attn.indexer*",
165
+ "model.layers.33.self_attn*",
166
+ "model.layers.33.self_attn.indexer*",
167
+ "model.layers.34.self_attn*",
168
+ "model.layers.34.self_attn.indexer*",
169
+ "model.layers.35.self_attn*",
170
+ "model.layers.35.self_attn.indexer*",
171
+ "model.layers.36.self_attn*",
172
+ "model.layers.36.self_attn.indexer*",
173
+ "model.layers.37.self_attn*",
174
+ "model.layers.37.self_attn.indexer*",
175
+ "model.layers.38.self_attn*",
176
+ "model.layers.38.self_attn.indexer*",
177
+ "model.layers.39.self_attn*",
178
+ "model.layers.39.self_attn.indexer*",
179
+ "model.layers.4.self_attn*",
180
+ "model.layers.4.self_attn.indexer*",
181
+ "model.layers.40.self_attn*",
182
+ "model.layers.40.self_attn.indexer*",
183
+ "model.layers.41.self_attn*",
184
+ "model.layers.41.self_attn.indexer*",
185
+ "model.layers.42.self_attn*",
186
+ "model.layers.42.self_attn.indexer*",
187
+ "model.layers.43.self_attn*",
188
+ "model.layers.43.self_attn.indexer*",
189
+ "model.layers.44.self_attn*",
190
+ "model.layers.44.self_attn.indexer*",
191
+ "model.layers.45.self_attn*",
192
+ "model.layers.45.self_attn.indexer*",
193
+ "model.layers.46.self_attn*",
194
+ "model.layers.46.self_attn.indexer*",
195
+ "model.layers.47.self_attn*",
196
+ "model.layers.47.self_attn.indexer*",
197
+ "model.layers.48.self_attn*",
198
+ "model.layers.48.self_attn.indexer*",
199
+ "model.layers.49.self_attn*",
200
+ "model.layers.49.self_attn.indexer*",
201
+ "model.layers.5.self_attn*",
202
+ "model.layers.5.self_attn.indexer*",
203
+ "model.layers.50.self_attn*",
204
+ "model.layers.50.self_attn.indexer*",
205
+ "model.layers.51.self_attn*",
206
+ "model.layers.51.self_attn.indexer*",
207
+ "model.layers.52.self_attn*",
208
+ "model.layers.52.self_attn.indexer*",
209
+ "model.layers.53.self_attn*",
210
+ "model.layers.53.self_attn.indexer*",
211
+ "model.layers.54.self_attn*",
212
+ "model.layers.54.self_attn.indexer*",
213
+ "model.layers.55.self_attn*",
214
+ "model.layers.55.self_attn.indexer*",
215
+ "model.layers.56.self_attn*",
216
+ "model.layers.56.self_attn.indexer*",
217
+ "model.layers.57.self_attn*",
218
+ "model.layers.57.self_attn.indexer*",
219
+ "model.layers.58.self_attn*",
220
+ "model.layers.58.self_attn.indexer*",
221
+ "model.layers.59.self_attn*",
222
+ "model.layers.59.self_attn.indexer*",
223
+ "model.layers.6.self_attn*",
224
+ "model.layers.6.self_attn.indexer*",
225
+ "model.layers.60.self_attn*",
226
+ "model.layers.60.self_attn.indexer*",
227
+ "model.layers.61.self_attn*",
228
+ "model.layers.61.self_attn.indexer*",
229
+ "model.layers.62.self_attn*",
230
+ "model.layers.62.self_attn.indexer*",
231
+ "model.layers.63.self_attn*",
232
+ "model.layers.63.self_attn.indexer*",
233
+ "model.layers.64.self_attn*",
234
+ "model.layers.64.self_attn.indexer*",
235
+ "model.layers.65.self_attn*",
236
+ "model.layers.65.self_attn.indexer*",
237
+ "model.layers.66.self_attn*",
238
+ "model.layers.66.self_attn.indexer*",
239
+ "model.layers.67.self_attn*",
240
+ "model.layers.67.self_attn.indexer*",
241
+ "model.layers.68.self_attn*",
242
+ "model.layers.68.self_attn.indexer*",
243
+ "model.layers.69.self_attn*",
244
+ "model.layers.69.self_attn.indexer*",
245
+ "model.layers.7.self_attn*",
246
+ "model.layers.7.self_attn.indexer*",
247
+ "model.layers.70.self_attn*",
248
+ "model.layers.70.self_attn.indexer*",
249
+ "model.layers.71.self_attn*",
250
+ "model.layers.71.self_attn.indexer*",
251
+ "model.layers.72.self_attn*",
252
+ "model.layers.72.self_attn.indexer*",
253
+ "model.layers.73.self_attn*",
254
+ "model.layers.73.self_attn.indexer*",
255
+ "model.layers.74.self_attn*",
256
+ "model.layers.74.self_attn.indexer*",
257
+ "model.layers.75.self_attn*",
258
+ "model.layers.75.self_attn.indexer*",
259
+ "model.layers.76.self_attn*",
260
+ "model.layers.76.self_attn.indexer*",
261
+ "model.layers.77.self_attn*",
262
+ "model.layers.77.self_attn.indexer*",
263
+ "model.layers.8.self_attn*",
264
+ "model.layers.8.self_attn.indexer*",
265
+ "model.layers.9.self_attn*",
266
+ "model.layers.9.self_attn.indexer*"
267
+ ],
268
+ "producer": {
269
+ "glm51_mtp_layer78_patch": "NVFP4 MTP layer transplanted in-place on 2026-05-26",
270
+ "glm51_mtp_layer78_source": "lukealonso/GLM-5.1-NVFP4-MTP@78b7fe365f3905b4e0261a85182fefdbd5137989",
271
+ "glm51_mxfp8_bf16_source": "/root/.cache/huggingface/hub/models--zai-org--GLM-5.1/snapshots/26e1bd6e011feb778d25ae34b09b07074139d92d",
272
+ "glm51_mxfp8_layers": "42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62",
273
+ "name": "modelopt",
274
+ "version": "0.39.0.dev290+gf9d9a71de.d20260214"
275
+ },
276
+ "quant_algo": "MIXED_PRECISION",
277
+ "quant_method": "modelopt",
278
+ "quantized_layers": {
279
+ "model.layers.10.mlp.experts": {
280
+ "group_size": 16,
281
+ "quant_algo": "NVFP4"
282
+ },
283
+ "model.layers.11.mlp.experts": {
284
+ "group_size": 16,
285
+ "quant_algo": "NVFP4"
286
+ },
287
+ "model.layers.12.mlp.experts": {
288
+ "group_size": 16,
289
+ "quant_algo": "NVFP4"
290
+ },
291
+ "model.layers.13.mlp.experts": {
292
+ "group_size": 16,
293
+ "quant_algo": "NVFP4"
294
+ },
295
+ "model.layers.14.mlp.experts": {
296
+ "group_size": 16,
297
+ "quant_algo": "NVFP4"
298
+ },
299
+ "model.layers.15.mlp.experts": {
300
+ "group_size": 16,
301
+ "quant_algo": "NVFP4"
302
+ },
303
+ "model.layers.16.mlp.experts": {
304
+ "group_size": 16,
305
+ "quant_algo": "NVFP4"
306
+ },
307
+ "model.layers.17.mlp.experts": {
308
+ "group_size": 16,
309
+ "quant_algo": "NVFP4"
310
+ },
311
+ "model.layers.18.mlp.experts": {
312
+ "group_size": 16,
313
+ "quant_algo": "NVFP4"
314
+ },
315
+ "model.layers.19.mlp.experts": {
316
+ "group_size": 16,
317
+ "quant_algo": "NVFP4"
318
+ },
319
+ "model.layers.20.mlp.experts": {
320
+ "group_size": 16,
321
+ "quant_algo": "NVFP4"
322
+ },
323
+ "model.layers.21.mlp.experts": {
324
+ "group_size": 16,
325
+ "quant_algo": "NVFP4"
326
+ },
327
+ "model.layers.22.mlp.experts": {
328
+ "group_size": 16,
329
+ "quant_algo": "NVFP4"
330
+ },
331
+ "model.layers.23.mlp.experts": {
332
+ "group_size": 16,
333
+ "quant_algo": "NVFP4"
334
+ },
335
+ "model.layers.24.mlp.experts": {
336
+ "group_size": 16,
337
+ "quant_algo": "NVFP4"
338
+ },
339
+ "model.layers.25.mlp.experts": {
340
+ "group_size": 16,
341
+ "quant_algo": "NVFP4"
342
+ },
343
+ "model.layers.26.mlp.experts": {
344
+ "group_size": 16,
345
+ "quant_algo": "NVFP4"
346
+ },
347
+ "model.layers.27.mlp.experts": {
348
+ "group_size": 16,
349
+ "quant_algo": "NVFP4"
350
+ },
351
+ "model.layers.28.mlp.experts": {
352
+ "group_size": 16,
353
+ "quant_algo": "NVFP4"
354
+ },
355
+ "model.layers.29.mlp.experts": {
356
+ "group_size": 16,
357
+ "quant_algo": "NVFP4"
358
+ },
359
+ "model.layers.3.mlp.experts": {
360
+ "group_size": 16,
361
+ "quant_algo": "NVFP4"
362
+ },
363
+ "model.layers.30.mlp.experts": {
364
+ "group_size": 16,
365
+ "quant_algo": "NVFP4"
366
+ },
367
+ "model.layers.31.mlp.experts": {
368
+ "group_size": 16,
369
+ "quant_algo": "NVFP4"
370
+ },
371
+ "model.layers.32.mlp.experts": {
372
+ "group_size": 16,
373
+ "quant_algo": "NVFP4"
374
+ },
375
+ "model.layers.33.mlp.experts": {
376
+ "group_size": 16,
377
+ "quant_algo": "NVFP4"
378
+ },
379
+ "model.layers.34.mlp.experts": {
380
+ "group_size": 16,
381
+ "quant_algo": "NVFP4"
382
+ },
383
+ "model.layers.35.mlp.experts": {
384
+ "group_size": 16,
385
+ "quant_algo": "NVFP4"
386
+ },
387
+ "model.layers.36.mlp.experts": {
388
+ "group_size": 16,
389
+ "quant_algo": "NVFP4"
390
+ },
391
+ "model.layers.37.mlp.experts": {
392
+ "group_size": 16,
393
+ "quant_algo": "NVFP4"
394
+ },
395
+ "model.layers.38.mlp.experts": {
396
+ "group_size": 16,
397
+ "quant_algo": "NVFP4"
398
+ },
399
+ "model.layers.39.mlp.experts": {
400
+ "group_size": 16,
401
+ "quant_algo": "NVFP4"
402
+ },
403
+ "model.layers.4.mlp.experts": {
404
+ "group_size": 16,
405
+ "quant_algo": "NVFP4"
406
+ },
407
+ "model.layers.40.mlp.experts": {
408
+ "group_size": 16,
409
+ "quant_algo": "NVFP4"
410
+ },
411
+ "model.layers.41.mlp.experts": {
412
+ "group_size": 16,
413
+ "quant_algo": "NVFP4"
414
+ },
415
+ "model.layers.42.mlp.experts": {
416
+ "group_size": 32,
417
+ "quant_algo": "MXFP8"
418
+ },
419
+ "model.layers.43.mlp.experts": {
420
+ "group_size": 32,
421
+ "quant_algo": "MXFP8"
422
+ },
423
+ "model.layers.44.mlp.experts": {
424
+ "group_size": 32,
425
+ "quant_algo": "MXFP8"
426
+ },
427
+ "model.layers.45.mlp.experts": {
428
+ "group_size": 32,
429
+ "quant_algo": "MXFP8"
430
+ },
431
+ "model.layers.46.mlp.experts": {
432
+ "group_size": 32,
433
+ "quant_algo": "MXFP8"
434
+ },
435
+ "model.layers.47.mlp.experts": {
436
+ "group_size": 32,
437
+ "quant_algo": "MXFP8"
438
+ },
439
+ "model.layers.48.mlp.experts": {
440
+ "group_size": 32,
441
+ "quant_algo": "MXFP8"
442
+ },
443
+ "model.layers.49.mlp.experts": {
444
+ "group_size": 32,
445
+ "quant_algo": "MXFP8"
446
+ },
447
+ "model.layers.5.mlp.experts": {
448
+ "group_size": 16,
449
+ "quant_algo": "NVFP4"
450
+ },
451
+ "model.layers.50.mlp.experts": {
452
+ "group_size": 32,
453
+ "quant_algo": "MXFP8"
454
+ },
455
+ "model.layers.51.mlp.experts": {
456
+ "group_size": 32,
457
+ "quant_algo": "MXFP8"
458
+ },
459
+ "model.layers.52.mlp.experts": {
460
+ "group_size": 32,
461
+ "quant_algo": "MXFP8"
462
+ },
463
+ "model.layers.53.mlp.experts": {
464
+ "group_size": 32,
465
+ "quant_algo": "MXFP8"
466
+ },
467
+ "model.layers.54.mlp.experts": {
468
+ "group_size": 32,
469
+ "quant_algo": "MXFP8"
470
+ },
471
+ "model.layers.55.mlp.experts": {
472
+ "group_size": 32,
473
+ "quant_algo": "MXFP8"
474
+ },
475
+ "model.layers.56.mlp.experts": {
476
+ "group_size": 32,
477
+ "quant_algo": "MXFP8"
478
+ },
479
+ "model.layers.57.mlp.experts": {
480
+ "group_size": 32,
481
+ "quant_algo": "MXFP8"
482
+ },
483
+ "model.layers.58.mlp.experts": {
484
+ "group_size": 32,
485
+ "quant_algo": "MXFP8"
486
+ },
487
+ "model.layers.59.mlp.experts": {
488
+ "group_size": 32,
489
+ "quant_algo": "MXFP8"
490
+ },
491
+ "model.layers.6.mlp.experts": {
492
+ "group_size": 16,
493
+ "quant_algo": "NVFP4"
494
+ },
495
+ "model.layers.60.mlp.experts": {
496
+ "group_size": 32,
497
+ "quant_algo": "MXFP8"
498
+ },
499
+ "model.layers.61.mlp.experts": {
500
+ "group_size": 32,
501
+ "quant_algo": "MXFP8"
502
+ },
503
+ "model.layers.62.mlp.experts": {
504
+ "group_size": 32,
505
+ "quant_algo": "MXFP8"
506
+ },
507
+ "model.layers.63.mlp.experts": {
508
+ "group_size": 16,
509
+ "quant_algo": "NVFP4"
510
+ },
511
+ "model.layers.64.mlp.experts": {
512
+ "group_size": 16,
513
+ "quant_algo": "NVFP4"
514
+ },
515
+ "model.layers.65.mlp.experts": {
516
+ "group_size": 16,
517
+ "quant_algo": "NVFP4"
518
+ },
519
+ "model.layers.66.mlp.experts": {
520
+ "group_size": 16,
521
+ "quant_algo": "NVFP4"
522
+ },
523
+ "model.layers.67.mlp.experts": {
524
+ "group_size": 16,
525
+ "quant_algo": "NVFP4"
526
+ },
527
+ "model.layers.68.mlp.experts": {
528
+ "group_size": 16,
529
+ "quant_algo": "NVFP4"
530
+ },
531
+ "model.layers.69.mlp.experts": {
532
+ "group_size": 16,
533
+ "quant_algo": "NVFP4"
534
+ },
535
+ "model.layers.7.mlp.experts": {
536
+ "group_size": 16,
537
+ "quant_algo": "NVFP4"
538
+ },
539
+ "model.layers.70.mlp.experts": {
540
+ "group_size": 16,
541
+ "quant_algo": "NVFP4"
542
+ },
543
+ "model.layers.71.mlp.experts": {
544
+ "group_size": 16,
545
+ "quant_algo": "NVFP4"
546
+ },
547
+ "model.layers.72.mlp.experts": {
548
+ "group_size": 16,
549
+ "quant_algo": "NVFP4"
550
+ },
551
+ "model.layers.73.mlp.experts": {
552
+ "group_size": 16,
553
+ "quant_algo": "NVFP4"
554
+ },
555
+ "model.layers.74.mlp.experts": {
556
+ "group_size": 16,
557
+ "quant_algo": "NVFP4"
558
+ },
559
+ "model.layers.75.mlp.experts": {
560
+ "group_size": 16,
561
+ "quant_algo": "NVFP4"
562
+ },
563
+ "model.layers.76.mlp.experts": {
564
+ "group_size": 16,
565
+ "quant_algo": "NVFP4"
566
+ },
567
+ "model.layers.77.mlp.experts": {
568
+ "group_size": 16,
569
+ "quant_algo": "NVFP4"
570
+ },
571
+ "model.layers.78.mlp.experts": {
572
+ "group_size": 16,
573
+ "quant_algo": "NVFP4"
574
+ },
575
+ "model.layers.8.mlp.experts": {
576
+ "group_size": 16,
577
+ "quant_algo": "NVFP4"
578
+ },
579
+ "model.layers.9.mlp.experts": {
580
+ "group_size": 16,
581
+ "quant_algo": "NVFP4"
582
+ }
583
+ }
584
+ },
585
+ "rms_norm_eps": 1e-05,
586
+ "rope_interleave": true,
587
+ "rope_parameters": {
588
+ "rope_theta": 1000000,
589
+ "rope_type": "default"
590
+ },
591
+ "routed_scaling_factor": 2.5,
592
+ "scoring_func": "sigmoid",
593
+ "tie_word_embeddings": false,
594
+ "topk_group": 1,
595
+ "topk_method": "noaux_tc",
596
+ "transformers_version": "5.4.0",
597
+ "use_cache": true,
598
+ "v_head_dim": 256,
599
+ "vocab_size": 154880
600
+ }