Hugging Face's logo

inferencerlabs
/
DeepSeek-V4-Flash-MTP-DSpark-MLX

Text Generation
MLX
Safetensors
English
deepseek_v4_dspark
8-bit precision
Model card Files
xet
 Community
inferencerlabs commited on
Commit
b7aba47
·
0 Parent(s):

Squash history to free storage

Browse files
Files changed (8) hide show
  1. .gitattributes +35 -0
  2. LICENSE +21 -0
  3. README.md +42 -0
  4. config.json +460 -0
  5. generation_config.json +9 -0
  6. model.safetensors +3 -0
  7. tokenizer.json +0 -0
  8. tokenizer_config.json +34 -0
.gitattributes ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz 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
LICENSE ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ MIT License
2
+
3
+ Copyright (c) 2023 DeepSeek
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
README.md ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ tags:
4
+ - mlx
5
+ base_model:
6
+ - deepseek-ai/DeepSeek-V4-Flash-0731
7
+ library_name: mlx
8
+ pipeline_tag: text-generation
9
+ ---
10
+ # DeepSeek-V4-Flash-0731 DSpark
11
+ See DeepSeek-V4-Flash with MTP in action: [demonstration videos](https://youtube.com/xcreate)
12
+
13
+ This draft model contains the extracted **DSpark** layers to be used alongside the [DeepSeek-V4-Flash-0731-MLX](https://huggingface.co/models?search=inferencerlabs/deepseek-v4-flash-0731-mlx) model as a speculative decoder for improved performance.
14
+
15
+ #### Tested on an M3 Ultra 512 GiB RAM using [Inferencer app v2.2.3](https://inferencer.com)
16
+ ### Tetris HTML
17
+ <table style="border-collapse: collapse; border: none; text-align:left; margin-top:10px; margin-bottom:0px;">
18
+ <tr><td><strong>Text inference</strong></td><td>~31 tokens/s @ 500 tokens ~148.5 GiB</td></tr>
19
+ <tr><td><strong>Text inference (with MTP)</strong></td><td>~35.8 tokens/s @ 500 tokens ~152.9 GiB</td></tr>
20
+ <tr><td><strong>Text inference (with DSpark)</strong></td><td>~42.7 tokens/s @ 500 tokens ~165.6 GiB</td></tr>
21
+ </table>
22
+
23
+ ### Tell me about Paris
24
+ <table style="border-collapse: collapse; border: none; text-align:left; margin-top:10px; margin-bottom:0px;">
25
+ <tr><td><strong>Text inference</strong></td><td>~30.5 tokens/s @ 500 tokens ~148.5 GiB</td></tr>
26
+ <tr><td><strong>Text inference (with MTP)</strong></td><td>~37.3 tokens/s @ 500 tokens ~152.9 GiB</td></tr>
27
+ <tr><td><strong>Text inference (with DSpark)</strong></td><td>~28.8 tokens/s @ 500 tokens ~165.6 GiB</td></tr>
28
+ </table>
29
+
30
+ <p style="margin-bottom:0;padding-bottom:0;">Enable speculative decoding in Inference Contols:</p>
31
+ <div style="margin-left:10px;">
32
+ <img
33
+ src="https://cdn-uploads.huggingface.co/production/uploads/688479d616f1ec82fa645019/weJgx7wZI7VgOdQe827yS.jpeg"
34
+ style="display:block;margin:0;">
35
+ </div>
36
+
37
+ ##### Extracted with a modified version of [MLX-VLM](https://github.com/Blaizzy/mlx-vlm)
38
+ ##### For more details see our [demonstration videos](https://youtube.com/xcreate) or visit [DeepSeek-V4-Flash](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash).
39
+
40
+ ## Disclaimer
41
+
42
+ We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.
config.json ADDED
@@ -0,0 +1,460 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "attention_bias": false,
3
+ "attention_dropout": 0.0,
4
+ "bos_token_id": 0,
5
+ "compress_ratios": [
6
+ 0,
7
+ 0,
8
+ 4,
9
+ 128,
10
+ 4,
11
+ 128,
12
+ 4,
13
+ 128,
14
+ 4,
15
+ 128,
16
+ 4,
17
+ 128,
18
+ 4,
19
+ 128,
20
+ 4,
21
+ 128,
22
+ 4,
23
+ 128,
24
+ 4,
25
+ 128,
26
+ 4,
27
+ 128,
28
+ 4,
29
+ 128,
30
+ 4,
31
+ 128,
32
+ 4,
33
+ 128,
34
+ 4,
35
+ 128,
36
+ 4,
37
+ 128,
38
+ 4,
39
+ 128,
40
+ 4,
41
+ 128,
42
+ 4,
43
+ 128,
44
+ 4,
45
+ 128,
46
+ 4,
47
+ 128,
48
+ 4,
49
+ 0,
50
+ 0,
51
+ 0
52
+ ],
53
+ "compress_rope_theta": 160000,
54
+ "dspark_block_size": 5,
55
+ "dspark_markov_rank": 256,
56
+ "dspark_noise_token_id": 128799,
57
+ "dspark_target_layer_ids": [
58
+ 40,
59
+ 41,
60
+ 42
61
+ ],
62
+ "eos_token_id": 1,
63
+ "expert_dtype": "fp4",
64
+ "hc_eps": 1e-06,
65
+ "hc_mult": 4,
66
+ "hc_sinkhorn_iters": 20,
67
+ "head_dim": 512,
68
+ "hidden_act": "silu",
69
+ "hidden_size": 4096,
70
+ "index_head_dim": 128,
71
+ "index_n_heads": 64,
72
+ "index_topk": 512,
73
+ "initializer_range": 0.02,
74
+ "max_position_embeddings": 1048576,
75
+ "model_type": "deepseek_v4_dspark",
76
+ "moe_intermediate_size": 2048,
77
+ "n_mtp_layers": 3,
78
+ "n_routed_experts": 256,
79
+ "n_shared_experts": 1,
80
+ "norm_topk_prob": true,
81
+ "num_attention_heads": 64,
82
+ "num_experts_per_tok": 6,
83
+ "num_hash_layers": 3,
84
+ "num_hidden_layers": 43,
85
+ "num_key_value_heads": 1,
86
+ "num_nextn_predict_layers": 1,
87
+ "o_groups": 8,
88
+ "o_lora_rank": 1024,
89
+ "q_lora_rank": 1024,
90
+ "qk_rope_head_dim": 64,
91
+ "quantization": {
92
+ "group_size": 64,
93
+ "bits": 8,
94
+ "mode": "affine",
95
+ "blocks.0.attn.wq_a": {
96
+ "group_size": 32,
97
+ "bits": 8,
98
+ "mode": "mxfp8"
99
+ },
100
+ "blocks.0.attn.wq_b": {
101
+ "group_size": 32,
102
+ "bits": 8,
103
+ "mode": "mxfp8"
104
+ },
105
+ "blocks.0.attn.wo_a": {
106
+ "group_size": 32,
107
+ "bits": 8,
108
+ "mode": "mxfp8"
109
+ },
110
+ "blocks.0.attn.wkv": {
111
+ "group_size": 32,
112
+ "bits": 8,
113
+ "mode": "mxfp8"
114
+ },
115
+ "blocks.0.attn.wo_b": {
116
+ "group_size": 32,
117
+ "bits": 8,
118
+ "mode": "mxfp8"
119
+ },
120
+ "blocks.0.ffn.shared_experts.gate_proj": {
121
+ "group_size": 32,
122
+ "bits": 8,
123
+ "mode": "mxfp8"
124
+ },
125
+ "blocks.0.ffn.shared_experts.up_proj": {
126
+ "group_size": 32,
127
+ "bits": 8,
128
+ "mode": "mxfp8"
129
+ },
130
+ "blocks.0.ffn.shared_experts.down_proj": {
131
+ "group_size": 32,
132
+ "bits": 8,
133
+ "mode": "mxfp8"
134
+ },
135
+ "blocks.0.main_proj": {
136
+ "group_size": 32,
137
+ "bits": 8,
138
+ "mode": "mxfp8"
139
+ },
140
+ "blocks.1.attn.wq_a": {
141
+ "group_size": 32,
142
+ "bits": 8,
143
+ "mode": "mxfp8"
144
+ },
145
+ "blocks.1.attn.wq_b": {
146
+ "group_size": 32,
147
+ "bits": 8,
148
+ "mode": "mxfp8"
149
+ },
150
+ "blocks.1.attn.wo_a": {
151
+ "group_size": 32,
152
+ "bits": 8,
153
+ "mode": "mxfp8"
154
+ },
155
+ "blocks.1.attn.wkv": {
156
+ "group_size": 32,
157
+ "bits": 8,
158
+ "mode": "mxfp8"
159
+ },
160
+ "blocks.1.attn.wo_b": {
161
+ "group_size": 32,
162
+ "bits": 8,
163
+ "mode": "mxfp8"
164
+ },
165
+ "blocks.1.ffn.shared_experts.gate_proj": {
166
+ "group_size": 32,
167
+ "bits": 8,
168
+ "mode": "mxfp8"
169
+ },
170
+ "blocks.1.ffn.shared_experts.up_proj": {
171
+ "group_size": 32,
172
+ "bits": 8,
173
+ "mode": "mxfp8"
174
+ },
175
+ "blocks.1.ffn.shared_experts.down_proj": {
176
+ "group_size": 32,
177
+ "bits": 8,
178
+ "mode": "mxfp8"
179
+ },
180
+ "blocks.2.attn.wq_a": {
181
+ "group_size": 32,
182
+ "bits": 8,
183
+ "mode": "mxfp8"
184
+ },
185
+ "blocks.2.attn.wq_b": {
186
+ "group_size": 32,
187
+ "bits": 8,
188
+ "mode": "mxfp8"
189
+ },
190
+ "blocks.2.attn.wo_a": {
191
+ "group_size": 32,
192
+ "bits": 8,
193
+ "mode": "mxfp8"
194
+ },
195
+ "blocks.2.attn.wkv": {
196
+ "group_size": 32,
197
+ "bits": 8,
198
+ "mode": "mxfp8"
199
+ },
200
+ "blocks.2.attn.wo_b": {
201
+ "group_size": 32,
202
+ "bits": 8,
203
+ "mode": "mxfp8"
204
+ },
205
+ "blocks.2.ffn.shared_experts.gate_proj": {
206
+ "group_size": 32,
207
+ "bits": 8,
208
+ "mode": "mxfp8"
209
+ },
210
+ "blocks.2.ffn.shared_experts.up_proj": {
211
+ "group_size": 32,
212
+ "bits": 8,
213
+ "mode": "mxfp8"
214
+ },
215
+ "blocks.2.ffn.shared_experts.down_proj": {
216
+ "group_size": 32,
217
+ "bits": 8,
218
+ "mode": "mxfp8"
219
+ },
220
+ "blocks.0.ffn.switch_mlp.gate_proj": {
221
+ "group_size": 32,
222
+ "bits": 4,
223
+ "mode": "mxfp4"
224
+ },
225
+ "blocks.0.ffn.switch_mlp.down_proj": {
226
+ "group_size": 32,
227
+ "bits": 4,
228
+ "mode": "mxfp4"
229
+ },
230
+ "blocks.0.ffn.switch_mlp.up_proj": {
231
+ "group_size": 32,
232
+ "bits": 4,
233
+ "mode": "mxfp4"
234
+ },
235
+ "blocks.1.ffn.switch_mlp.gate_proj": {
236
+ "group_size": 32,
237
+ "bits": 4,
238
+ "mode": "mxfp4"
239
+ },
240
+ "blocks.1.ffn.switch_mlp.down_proj": {
241
+ "group_size": 32,
242
+ "bits": 4,
243
+ "mode": "mxfp4"
244
+ },
245
+ "blocks.1.ffn.switch_mlp.up_proj": {
246
+ "group_size": 32,
247
+ "bits": 4,
248
+ "mode": "mxfp4"
249
+ },
250
+ "blocks.2.ffn.switch_mlp.gate_proj": {
251
+ "group_size": 32,
252
+ "bits": 4,
253
+ "mode": "mxfp4"
254
+ },
255
+ "blocks.2.ffn.switch_mlp.down_proj": {
256
+ "group_size": 32,
257
+ "bits": 4,
258
+ "mode": "mxfp4"
259
+ },
260
+ "blocks.2.ffn.switch_mlp.up_proj": {
261
+ "group_size": 32,
262
+ "bits": 4,
263
+ "mode": "mxfp4"
264
+ }
265
+ },
266
+ "quantization_config": {
267
+ "group_size": 64,
268
+ "bits": 8,
269
+ "mode": "affine",
270
+ "blocks.0.attn.wq_a": {
271
+ "group_size": 32,
272
+ "bits": 8,
273
+ "mode": "mxfp8"
274
+ },
275
+ "blocks.0.attn.wq_b": {
276
+ "group_size": 32,
277
+ "bits": 8,
278
+ "mode": "mxfp8"
279
+ },
280
+ "blocks.0.attn.wo_a": {
281
+ "group_size": 32,
282
+ "bits": 8,
283
+ "mode": "mxfp8"
284
+ },
285
+ "blocks.0.attn.wkv": {
286
+ "group_size": 32,
287
+ "bits": 8,
288
+ "mode": "mxfp8"
289
+ },
290
+ "blocks.0.attn.wo_b": {
291
+ "group_size": 32,
292
+ "bits": 8,
293
+ "mode": "mxfp8"
294
+ },
295
+ "blocks.0.ffn.shared_experts.gate_proj": {
296
+ "group_size": 32,
297
+ "bits": 8,
298
+ "mode": "mxfp8"
299
+ },
300
+ "blocks.0.ffn.shared_experts.up_proj": {
301
+ "group_size": 32,
302
+ "bits": 8,
303
+ "mode": "mxfp8"
304
+ },
305
+ "blocks.0.ffn.shared_experts.down_proj": {
306
+ "group_size": 32,
307
+ "bits": 8,
308
+ "mode": "mxfp8"
309
+ },
310
+ "blocks.0.main_proj": {
311
+ "group_size": 32,
312
+ "bits": 8,
313
+ "mode": "mxfp8"
314
+ },
315
+ "blocks.1.attn.wq_a": {
316
+ "group_size": 32,
317
+ "bits": 8,
318
+ "mode": "mxfp8"
319
+ },
320
+ "blocks.1.attn.wq_b": {
321
+ "group_size": 32,
322
+ "bits": 8,
323
+ "mode": "mxfp8"
324
+ },
325
+ "blocks.1.attn.wo_a": {
326
+ "group_size": 32,
327
+ "bits": 8,
328
+ "mode": "mxfp8"
329
+ },
330
+ "blocks.1.attn.wkv": {
331
+ "group_size": 32,
332
+ "bits": 8,
333
+ "mode": "mxfp8"
334
+ },
335
+ "blocks.1.attn.wo_b": {
336
+ "group_size": 32,
337
+ "bits": 8,
338
+ "mode": "mxfp8"
339
+ },
340
+ "blocks.1.ffn.shared_experts.gate_proj": {
341
+ "group_size": 32,
342
+ "bits": 8,
343
+ "mode": "mxfp8"
344
+ },
345
+ "blocks.1.ffn.shared_experts.up_proj": {
346
+ "group_size": 32,
347
+ "bits": 8,
348
+ "mode": "mxfp8"
349
+ },
350
+ "blocks.1.ffn.shared_experts.down_proj": {
351
+ "group_size": 32,
352
+ "bits": 8,
353
+ "mode": "mxfp8"
354
+ },
355
+ "blocks.2.attn.wq_a": {
356
+ "group_size": 32,
357
+ "bits": 8,
358
+ "mode": "mxfp8"
359
+ },
360
+ "blocks.2.attn.wq_b": {
361
+ "group_size": 32,
362
+ "bits": 8,
363
+ "mode": "mxfp8"
364
+ },
365
+ "blocks.2.attn.wo_a": {
366
+ "group_size": 32,
367
+ "bits": 8,
368
+ "mode": "mxfp8"
369
+ },
370
+ "blocks.2.attn.wkv": {
371
+ "group_size": 32,
372
+ "bits": 8,
373
+ "mode": "mxfp8"
374
+ },
375
+ "blocks.2.attn.wo_b": {
376
+ "group_size": 32,
377
+ "bits": 8,
378
+ "mode": "mxfp8"
379
+ },
380
+ "blocks.2.ffn.shared_experts.gate_proj": {
381
+ "group_size": 32,
382
+ "bits": 8,
383
+ "mode": "mxfp8"
384
+ },
385
+ "blocks.2.ffn.shared_experts.up_proj": {
386
+ "group_size": 32,
387
+ "bits": 8,
388
+ "mode": "mxfp8"
389
+ },
390
+ "blocks.2.ffn.shared_experts.down_proj": {
391
+ "group_size": 32,
392
+ "bits": 8,
393
+ "mode": "mxfp8"
394
+ },
395
+ "blocks.0.ffn.switch_mlp.gate_proj": {
396
+ "group_size": 32,
397
+ "bits": 4,
398
+ "mode": "mxfp4"
399
+ },
400
+ "blocks.0.ffn.switch_mlp.down_proj": {
401
+ "group_size": 32,
402
+ "bits": 4,
403
+ "mode": "mxfp4"
404
+ },
405
+ "blocks.0.ffn.switch_mlp.up_proj": {
406
+ "group_size": 32,
407
+ "bits": 4,
408
+ "mode": "mxfp4"
409
+ },
410
+ "blocks.1.ffn.switch_mlp.gate_proj": {
411
+ "group_size": 32,
412
+ "bits": 4,
413
+ "mode": "mxfp4"
414
+ },
415
+ "blocks.1.ffn.switch_mlp.down_proj": {
416
+ "group_size": 32,
417
+ "bits": 4,
418
+ "mode": "mxfp4"
419
+ },
420
+ "blocks.1.ffn.switch_mlp.up_proj": {
421
+ "group_size": 32,
422
+ "bits": 4,
423
+ "mode": "mxfp4"
424
+ },
425
+ "blocks.2.ffn.switch_mlp.gate_proj": {
426
+ "group_size": 32,
427
+ "bits": 4,
428
+ "mode": "mxfp4"
429
+ },
430
+ "blocks.2.ffn.switch_mlp.down_proj": {
431
+ "group_size": 32,
432
+ "bits": 4,
433
+ "mode": "mxfp4"
434
+ },
435
+ "blocks.2.ffn.switch_mlp.up_proj": {
436
+ "group_size": 32,
437
+ "bits": 4,
438
+ "mode": "mxfp4"
439
+ }
440
+ },
441
+ "rms_norm_eps": 1e-06,
442
+ "rope_scaling": {
443
+ "beta_fast": 32,
444
+ "beta_slow": 1,
445
+ "factor": 16,
446
+ "original_max_position_embeddings": 65536,
447
+ "type": "yarn"
448
+ },
449
+ "rope_theta": 10000,
450
+ "routed_scaling_factor": 1.5,
451
+ "scoring_func": "sqrtsoftplus",
452
+ "sliding_window": 128,
453
+ "swiglu_limit": 10.0,
454
+ "tie_word_embeddings": false,
455
+ "topk_method": "noaux_tc",
456
+ "torch_dtype": "bfloat16",
457
+ "transformers_version": "4.57.1",
458
+ "use_cache": true,
459
+ "vocab_size": 129280
460
+ }
generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 1,
5
+ "do_sample": true,
6
+ "temperature": 1.0,
7
+ "top_p": 1.0,
8
+ "transformers_version": "4.46.3"
9
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b85975ff649e50af534b48e90ea81514d9b1446fb72a1e42be00dd2c013a6666
3
+ size 12994906286
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_eos_token": false,
4
+ "bos_token": {
5
+ "__type": "AddedToken",
6
+ "content": "<｜begin▁of▁sentence｜>",
7
+ "lstrip": false,
8
+ "normalized": true,
9
+ "rstrip": false,
10
+ "single_word": false
11
+ },
12
+ "clean_up_tokenization_spaces": false,
13
+ "eos_token": {
14
+ "__type": "AddedToken",
15
+ "content": "<｜end▁of▁sentence｜>",
16
+ "lstrip": false,
17
+ "normalized": true,
18
+ "rstrip": false,
19
+ "single_word": false
20
+ },
21
+ "legacy": true,
22
+ "model_max_length": 1048576,
23
+ "pad_token": {
24
+ "__type": "AddedToken",
25
+ "content": "<｜end▁of▁sentence｜>",
26
+ "lstrip": false,
27
+ "normalized": true,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ },
31
+ "sp_model_kwargs": {},
32
+ "unk_token": null,
33
+ "tokenizer_class": "PreTrainedTokenizerFast"
34
+ }