Hugging Face's logo

lmms-lab
/
llava-onevision-qwen2-7b-ov

Text Generation
Transformers
Safetensors
English
Chinese
llava
multimodal
conversational
Eval Results
Inference Endpoints
Model card Files Community
10
PY007 commited on
Commit
240d262
•
1 Parent(s): 0b66d17

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +28 -298
config.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "_name_or_path": "Qwen/Qwen2-7B-Instruct",
3
  "architectures": [
4
  "LlavaQwenForCausalLM"
5
  ],
@@ -8,6 +8,7 @@
8
  "eos_token_id": 151645,
9
  "hidden_act": "silu",
10
  "hidden_size": 3584,
 
11
  "image_aspect_ratio": "anyres_max_9",
12
  "image_crop_resolution": null,
13
  "image_grid_pinpoints": [
@@ -27,401 +28,129 @@
27
  384,
28
  1536
29
  ],
30
- [
31
- 768,
32
- 768
33
- ],
34
  [
35
  384,
36
  1920
37
  ],
38
- [
39
- 768,
40
- 768
41
- ],
42
  [
43
  384,
44
  2304
45
  ],
46
  [
47
  768,
48
- 1152
49
- ],
50
- [
51
- 384,
52
- 2688
53
- ],
54
- [
55
- 768,
56
- 1152
57
- ],
58
- [
59
- 384,
60
- 3072
61
  ],
62
  [
63
  768,
64
- 1536
65
- ],
66
- [
67
- 384,
68
- 3456
69
  ],
70
  [
71
  768,
72
- 1536
73
- ],
74
- [
75
- 1152,
76
  1152
77
  ],
78
  [
79
  768,
80
- 1920
81
- ],
82
- [
83
- 1152,
84
- 1152
85
  ],
86
  [
87
  768,
88
  1920
89
  ],
90
- [
91
- 1152,
92
- 1152
93
- ],
94
  [
95
  768,
96
  2304
97
  ],
98
  [
99
  1152,
100
- 1536
101
- ],
102
- [
103
- 768,
104
- 2304
105
- ],
106
- [
107
- 1152,
108
- 1536
109
- ],
110
- [
111
- 768,
112
- 2688
113
  ],
114
  [
115
  1152,
116
- 1536
117
- ],
118
- [
119
- 768,
120
- 2688
121
  ],
122
  [
123
  1152,
124
- 1920
125
- ],
126
- [
127
- 768,
128
- 3072
129
  ],
130
  [
131
  1152,
132
- 1920
133
- ],
134
- [
135
- 1536,
136
  1536
137
  ],
138
- [
139
- 768,
140
- 3072
141
- ],
142
  [
143
  1152,
144
  1920
145
  ],
146
- [
147
- 1536,
148
- 1536
149
- ],
150
- [
151
- 768,
152
- 3456
153
- ],
154
- [
155
- 1152,
156
- 2304
157
- ],
158
- [
159
- 1536,
160
- 1536
161
- ],
162
- [
163
- 768,
164
- 3456
165
- ],
166
  [
167
  1152,
168
  2304
169
  ],
170
  [
171
  1536,
172
- 1536
173
- ],
174
- [
175
- 768,
176
- 3840
177
- ],
178
- [
179
- 1152,
180
- 2304
181
  ],
182
  [
183
  1536,
184
- 1920
185
- ],
186
- [
187
- 768,
188
- 3840
189
- ],
190
- [
191
- 1152,
192
- 2688
193
  ],
194
  [
195
  1536,
196
- 1920
197
- ],
198
- [
199
- 768,
200
- 4224
201
- ],
202
- [
203
- 1152,
204
- 2688
205
  ],
206
  [
207
  1536,
208
- 1920
209
- ],
210
- [
211
- 768,
212
- 4224
213
- ],
214
- [
215
- 1152,
216
- 2688
217
  ],
218
  [
219
  1536,
220
  1920
221
  ],
222
- [
223
- 768,
224
- 4608
225
- ],
226
- [
227
- 1152,
228
- 3072
229
- ],
230
- [
231
- 1536,
232
- 2304
233
- ],
234
- [
235
- 768,
236
- 4608
237
- ],
238
- [
239
- 1152,
240
- 3072
241
- ],
242
  [
243
  1536,
244
  2304
245
  ],
246
  [
247
  1920,
248
- 1920
249
- ],
250
- [
251
- 768,
252
- 4992
253
- ],
254
- [
255
- 1152,
256
- 3072
257
- ],
258
- [
259
- 1536,
260
- 2304
261
  ],
262
  [
263
  1920,
264
- 1920
265
- ],
266
- [
267
- 768,
268
- 4992
269
- ],
270
- [
271
- 1152,
272
- 3456
273
- ],
274
- [
275
- 1536,
276
- 2304
277
  ],
278
  [
279
  1920,
280
- 1920
281
- ],
282
- [
283
- 768,
284
- 5376
285
- ],
286
- [
287
- 1152,
288
- 3456
289
- ],
290
- [
291
- 1536,
292
- 2688
293
  ],
294
  [
295
  1920,
296
- 1920
297
- ],
298
- [
299
- 768,
300
- 5376
301
- ],
302
- [
303
- 1152,
304
- 3456
305
- ],
306
- [
307
- 1536,
308
- 2688
309
  ],
310
  [
311
  1920,
312
  1920
313
  ],
314
- [
315
- 768,
316
- 5760
317
- ],
318
- [
319
- 1152,
320
- 3840
321
- ],
322
- [
323
- 1536,
324
- 2688
325
- ],
326
  [
327
  1920,
328
  2304
329
  ],
330
  [
331
- 768,
332
- 5760
333
- ],
334
- [
335
- 1152,
336
- 3840
337
- ],
338
- [
339
- 1536,
340
- 2688
341
- ],
342
- [
343
- 1920,
344
- 2304
345
- ],
346
- [
347
- 768,
348
- 6144
349
- ],
350
- [
351
- 1152,
352
- 3840
353
- ],
354
- [
355
- 1536,
356
- 3072
357
- ],
358
- [
359
- 1920,
360
- 2304
361
- ],
362
- [
363
- 768,
364
- 6144
365
- ],
366
- [
367
- 1152,
368
- 4224
369
- ],
370
- [
371
- 1536,
372
- 3072
373
- ],
374
- [
375
- 1920,
376
- 2304
377
- ],
378
- [
379
- 768,
380
- 6528
381
- ],
382
- [
383
- 1152,
384
- 4224
385
- ],
386
- [
387
- 1536,
388
- 3072
389
- ],
390
- [
391
- 1920,
392
- 2304
393
- ],
394
- [
395
- 768,
396
- 6528
397
- ],
398
- [
399
- 1152,
400
- 4224
401
- ],
402
- [
403
- 1536,
404
- 3072
405
- ],
406
- [
407
- 1920,
408
- 2688
409
  ],
410
  [
411
- 768,
412
- 6912
413
  ],
414
  [
415
- 1152,
416
- 4608
417
  ],
418
  [
419
- 1536,
420
- 3456
421
  ],
422
  [
423
- 1920,
424
- 2688
425
  ],
426
  [
427
  2304,
@@ -438,6 +167,7 @@
438
  "mm_projector_lr": null,
439
  "mm_projector_type": "mlp2x_gelu",
440
  "mm_resampler_type": null,
 
441
  "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
442
  "mm_use_im_patch_token": false,
443
  "mm_use_im_start_end": false,
@@ -445,7 +175,7 @@
445
  "mm_vision_select_layer": -2,
446
  "mm_vision_tower": "google/siglip-so400m-patch14-384",
447
  "mm_vision_tower_lr": 2e-06,
448
- "model_type": "qwen2",
449
  "num_attention_heads": 28,
450
  "num_hidden_layers": 28,
451
  "num_key_value_heads": 4,
 
1
  {
2
+ "_name_or_path": "/mnt/bn/vl-research/checkpoints/onevision/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-mid_to_final_next_2p4m_am4",
3
  "architectures": [
4
  "LlavaQwenForCausalLM"
5
  ],
 
8
  "eos_token_id": 151645,
9
  "hidden_act": "silu",
10
  "hidden_size": 3584,
11
+ "image_token_index": 151646,
12
  "image_aspect_ratio": "anyres_max_9",
13
  "image_crop_resolution": null,
14
  "image_grid_pinpoints": [
 
28
  384,
29
  1536
30
  ],
 
 
 
 
31
  [
32
  384,
33
  1920
34
  ],
 
 
 
 
35
  [
36
  384,
37
  2304
38
  ],
39
  [
40
  768,
41
+ 384
 
 
 
 
 
 
 
 
 
 
 
 
42
  ],
43
  [
44
  768,
45
+ 768
 
 
 
 
46
  ],
47
  [
48
  768,
 
 
 
 
49
  1152
50
  ],
51
  [
52
  768,
53
+ 1536
 
 
 
 
54
  ],
55
  [
56
  768,
57
  1920
58
  ],
 
 
 
 
59
  [
60
  768,
61
  2304
62
  ],
63
  [
64
  1152,
65
+ 384
 
 
 
 
 
 
 
 
 
 
 
 
66
  ],
67
  [
68
  1152,
69
+ 768
 
 
 
 
70
  ],
71
  [
72
  1152,
73
+ 1152
 
 
 
 
74
  ],
75
  [
76
  1152,
 
 
 
 
77
  1536
78
  ],
 
 
 
 
79
  [
80
  1152,
81
  1920
82
  ],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83
  [
84
  1152,
85
  2304
86
  ],
87
  [
88
  1536,
89
+ 384
 
 
 
 
 
 
 
 
90
  ],
91
  [
92
  1536,
93
+ 768
 
 
 
 
 
 
 
 
94
  ],
95
  [
96
  1536,
97
+ 1152
 
 
 
 
 
 
 
 
98
  ],
99
  [
100
  1536,
101
+ 1536
 
 
 
 
 
 
 
 
102
  ],
103
  [
104
  1536,
105
  1920
106
  ],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
107
  [
108
  1536,
109
  2304
110
  ],
111
  [
112
  1920,
113
+ 384
 
 
 
 
 
 
 
 
 
 
 
 
114
  ],
115
  [
116
  1920,
117
+ 768
 
 
 
 
 
 
 
 
 
 
 
 
118
  ],
119
  [
120
  1920,
121
+ 1152
 
 
 
 
 
 
 
 
 
 
 
 
122
  ],
123
  [
124
  1920,
125
+ 1536
 
 
 
 
 
 
 
 
 
 
 
 
126
  ],
127
  [
128
  1920,
129
  1920
130
  ],
 
 
 
 
 
 
 
 
 
 
 
 
131
  [
132
  1920,
133
  2304
134
  ],
135
  [
136
+ 2304,
137
+ 384
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
138
  ],
139
  [
140
+ 2304,
141
+ 768
142
  ],
143
  [
144
+ 2304,
145
+ 1152
146
  ],
147
  [
148
+ 2304,
149
+ 1536
150
  ],
151
  [
152
+ 2304,
153
+ 1920
154
  ],
155
  [
156
  2304,
 
167
  "mm_projector_lr": null,
168
  "mm_projector_type": "mlp2x_gelu",
169
  "mm_resampler_type": null,
170
+ "mm_spatial_pool_mode": "bilinear",
171
  "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
172
  "mm_use_im_patch_token": false,
173
  "mm_use_im_start_end": false,
 
175
  "mm_vision_select_layer": -2,
176
  "mm_vision_tower": "google/siglip-so400m-patch14-384",
177
  "mm_vision_tower_lr": 2e-06,
178
+ "model_type": "llava",
179
  "num_attention_heads": 28,
180
  "num_hidden_layers": 28,
181
  "num_key_value_heads": 4,