Hugging Face's logo

neuralmagic
/
Llama-3.2-11B-Vision-Instruct-FP8-dynamic

Text Generation
Safetensors
mllama
fp8
vllm
conversational
compressed-tensors
Model card Files Community
3
mgoin commited on
Commit
c0ee264
•
1 Parent(s): 4bcd61f

Upload folder using huggingface_hub

Browse files
Files changed (6) hide show
  1. config.json +245 -2
  2. model-00001-of-00003.safetensors +2 -2
  3. model-00002-of-00003.safetensors +2 -2
  4. model-00003-of-00003.safetensors +2 -2
  5. model.safetensors.index.json +0 -0
  6. recipe.yaml +6 -0
config.json CHANGED
@@ -37,8 +37,251 @@
37
  }
38
  },
39
  "format": "float-quantized",
40
- "global_compression_ratio": 1.2490120910036353,
41
- "ignore": [],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
  "kv_cache_scheme": null,
43
  "quant_method": "compressed-tensors",
44
  "quantization_status": "compressed"
 
37
  }
38
  },
39
  "format": "float-quantized",
40
+ "global_compression_ratio": 1.1510108015292029,
41
+ "ignore": [
42
+ "vision_model.transformer.layers.0.self_attn.q_proj",
43
+ "vision_model.transformer.layers.0.self_attn.k_proj",
44
+ "vision_model.transformer.layers.0.self_attn.v_proj",
45
+ "vision_model.transformer.layers.0.self_attn.o_proj",
46
+ "vision_model.transformer.layers.0.mlp.fc1",
47
+ "vision_model.transformer.layers.0.mlp.fc2",
48
+ "vision_model.transformer.layers.1.self_attn.q_proj",
49
+ "vision_model.transformer.layers.1.self_attn.k_proj",
50
+ "vision_model.transformer.layers.1.self_attn.v_proj",
51
+ "vision_model.transformer.layers.1.self_attn.o_proj",
52
+ "vision_model.transformer.layers.1.mlp.fc1",
53
+ "vision_model.transformer.layers.1.mlp.fc2",
54
+ "vision_model.transformer.layers.2.self_attn.q_proj",
55
+ "vision_model.transformer.layers.2.self_attn.k_proj",
56
+ "vision_model.transformer.layers.2.self_attn.v_proj",
57
+ "vision_model.transformer.layers.2.self_attn.o_proj",
58
+ "vision_model.transformer.layers.2.mlp.fc1",
59
+ "vision_model.transformer.layers.2.mlp.fc2",
60
+ "vision_model.transformer.layers.3.self_attn.q_proj",
61
+ "vision_model.transformer.layers.3.self_attn.k_proj",
62
+ "vision_model.transformer.layers.3.self_attn.v_proj",
63
+ "vision_model.transformer.layers.3.self_attn.o_proj",
64
+ "vision_model.transformer.layers.3.mlp.fc1",
65
+ "vision_model.transformer.layers.3.mlp.fc2",
66
+ "vision_model.transformer.layers.4.self_attn.q_proj",
67
+ "vision_model.transformer.layers.4.self_attn.k_proj",
68
+ "vision_model.transformer.layers.4.self_attn.v_proj",
69
+ "vision_model.transformer.layers.4.self_attn.o_proj",
70
+ "vision_model.transformer.layers.4.mlp.fc1",
71
+ "vision_model.transformer.layers.4.mlp.fc2",
72
+ "vision_model.transformer.layers.5.self_attn.q_proj",
73
+ "vision_model.transformer.layers.5.self_attn.k_proj",
74
+ "vision_model.transformer.layers.5.self_attn.v_proj",
75
+ "vision_model.transformer.layers.5.self_attn.o_proj",
76
+ "vision_model.transformer.layers.5.mlp.fc1",
77
+ "vision_model.transformer.layers.5.mlp.fc2",
78
+ "vision_model.transformer.layers.6.self_attn.q_proj",
79
+ "vision_model.transformer.layers.6.self_attn.k_proj",
80
+ "vision_model.transformer.layers.6.self_attn.v_proj",
81
+ "vision_model.transformer.layers.6.self_attn.o_proj",
82
+ "vision_model.transformer.layers.6.mlp.fc1",
83
+ "vision_model.transformer.layers.6.mlp.fc2",
84
+ "vision_model.transformer.layers.7.self_attn.q_proj",
85
+ "vision_model.transformer.layers.7.self_attn.k_proj",
86
+ "vision_model.transformer.layers.7.self_attn.v_proj",
87
+ "vision_model.transformer.layers.7.self_attn.o_proj",
88
+ "vision_model.transformer.layers.7.mlp.fc1",
89
+ "vision_model.transformer.layers.7.mlp.fc2",
90
+ "vision_model.transformer.layers.8.self_attn.q_proj",
91
+ "vision_model.transformer.layers.8.self_attn.k_proj",
92
+ "vision_model.transformer.layers.8.self_attn.v_proj",
93
+ "vision_model.transformer.layers.8.self_attn.o_proj",
94
+ "vision_model.transformer.layers.8.mlp.fc1",
95
+ "vision_model.transformer.layers.8.mlp.fc2",
96
+ "vision_model.transformer.layers.9.self_attn.q_proj",
97
+ "vision_model.transformer.layers.9.self_attn.k_proj",
98
+ "vision_model.transformer.layers.9.self_attn.v_proj",
99
+ "vision_model.transformer.layers.9.self_attn.o_proj",
100
+ "vision_model.transformer.layers.9.mlp.fc1",
101
+ "vision_model.transformer.layers.9.mlp.fc2",
102
+ "vision_model.transformer.layers.10.self_attn.q_proj",
103
+ "vision_model.transformer.layers.10.self_attn.k_proj",
104
+ "vision_model.transformer.layers.10.self_attn.v_proj",
105
+ "vision_model.transformer.layers.10.self_attn.o_proj",
106
+ "vision_model.transformer.layers.10.mlp.fc1",
107
+ "vision_model.transformer.layers.10.mlp.fc2",
108
+ "vision_model.transformer.layers.11.self_attn.q_proj",
109
+ "vision_model.transformer.layers.11.self_attn.k_proj",
110
+ "vision_model.transformer.layers.11.self_attn.v_proj",
111
+ "vision_model.transformer.layers.11.self_attn.o_proj",
112
+ "vision_model.transformer.layers.11.mlp.fc1",
113
+ "vision_model.transformer.layers.11.mlp.fc2",
114
+ "vision_model.transformer.layers.12.self_attn.q_proj",
115
+ "vision_model.transformer.layers.12.self_attn.k_proj",
116
+ "vision_model.transformer.layers.12.self_attn.v_proj",
117
+ "vision_model.transformer.layers.12.self_attn.o_proj",
118
+ "vision_model.transformer.layers.12.mlp.fc1",
119
+ "vision_model.transformer.layers.12.mlp.fc2",
120
+ "vision_model.transformer.layers.13.self_attn.q_proj",
121
+ "vision_model.transformer.layers.13.self_attn.k_proj",
122
+ "vision_model.transformer.layers.13.self_attn.v_proj",
123
+ "vision_model.transformer.layers.13.self_attn.o_proj",
124
+ "vision_model.transformer.layers.13.mlp.fc1",
125
+ "vision_model.transformer.layers.13.mlp.fc2",
126
+ "vision_model.transformer.layers.14.self_attn.q_proj",
127
+ "vision_model.transformer.layers.14.self_attn.k_proj",
128
+ "vision_model.transformer.layers.14.self_attn.v_proj",
129
+ "vision_model.transformer.layers.14.self_attn.o_proj",
130
+ "vision_model.transformer.layers.14.mlp.fc1",
131
+ "vision_model.transformer.layers.14.mlp.fc2",
132
+ "vision_model.transformer.layers.15.self_attn.q_proj",
133
+ "vision_model.transformer.layers.15.self_attn.k_proj",
134
+ "vision_model.transformer.layers.15.self_attn.v_proj",
135
+ "vision_model.transformer.layers.15.self_attn.o_proj",
136
+ "vision_model.transformer.layers.15.mlp.fc1",
137
+ "vision_model.transformer.layers.15.mlp.fc2",
138
+ "vision_model.transformer.layers.16.self_attn.q_proj",
139
+ "vision_model.transformer.layers.16.self_attn.k_proj",
140
+ "vision_model.transformer.layers.16.self_attn.v_proj",
141
+ "vision_model.transformer.layers.16.self_attn.o_proj",
142
+ "vision_model.transformer.layers.16.mlp.fc1",
143
+ "vision_model.transformer.layers.16.mlp.fc2",
144
+ "vision_model.transformer.layers.17.self_attn.q_proj",
145
+ "vision_model.transformer.layers.17.self_attn.k_proj",
146
+ "vision_model.transformer.layers.17.self_attn.v_proj",
147
+ "vision_model.transformer.layers.17.self_attn.o_proj",
148
+ "vision_model.transformer.layers.17.mlp.fc1",
149
+ "vision_model.transformer.layers.17.mlp.fc2",
150
+ "vision_model.transformer.layers.18.self_attn.q_proj",
151
+ "vision_model.transformer.layers.18.self_attn.k_proj",
152
+ "vision_model.transformer.layers.18.self_attn.v_proj",
153
+ "vision_model.transformer.layers.18.self_attn.o_proj",
154
+ "vision_model.transformer.layers.18.mlp.fc1",
155
+ "vision_model.transformer.layers.18.mlp.fc2",
156
+ "vision_model.transformer.layers.19.self_attn.q_proj",
157
+ "vision_model.transformer.layers.19.self_attn.k_proj",
158
+ "vision_model.transformer.layers.19.self_attn.v_proj",
159
+ "vision_model.transformer.layers.19.self_attn.o_proj",
160
+ "vision_model.transformer.layers.19.mlp.fc1",
161
+ "vision_model.transformer.layers.19.mlp.fc2",
162
+ "vision_model.transformer.layers.20.self_attn.q_proj",
163
+ "vision_model.transformer.layers.20.self_attn.k_proj",
164
+ "vision_model.transformer.layers.20.self_attn.v_proj",
165
+ "vision_model.transformer.layers.20.self_attn.o_proj",
166
+ "vision_model.transformer.layers.20.mlp.fc1",
167
+ "vision_model.transformer.layers.20.mlp.fc2",
168
+ "vision_model.transformer.layers.21.self_attn.q_proj",
169
+ "vision_model.transformer.layers.21.self_attn.k_proj",
170
+ "vision_model.transformer.layers.21.self_attn.v_proj",
171
+ "vision_model.transformer.layers.21.self_attn.o_proj",
172
+ "vision_model.transformer.layers.21.mlp.fc1",
173
+ "vision_model.transformer.layers.21.mlp.fc2",
174
+ "vision_model.transformer.layers.22.self_attn.q_proj",
175
+ "vision_model.transformer.layers.22.self_attn.k_proj",
176
+ "vision_model.transformer.layers.22.self_attn.v_proj",
177
+ "vision_model.transformer.layers.22.self_attn.o_proj",
178
+ "vision_model.transformer.layers.22.mlp.fc1",
179
+ "vision_model.transformer.layers.22.mlp.fc2",
180
+ "vision_model.transformer.layers.23.self_attn.q_proj",
181
+ "vision_model.transformer.layers.23.self_attn.k_proj",
182
+ "vision_model.transformer.layers.23.self_attn.v_proj",
183
+ "vision_model.transformer.layers.23.self_attn.o_proj",
184
+ "vision_model.transformer.layers.23.mlp.fc1",
185
+ "vision_model.transformer.layers.23.mlp.fc2",
186
+ "vision_model.transformer.layers.24.self_attn.q_proj",
187
+ "vision_model.transformer.layers.24.self_attn.k_proj",
188
+ "vision_model.transformer.layers.24.self_attn.v_proj",
189
+ "vision_model.transformer.layers.24.self_attn.o_proj",
190
+ "vision_model.transformer.layers.24.mlp.fc1",
191
+ "vision_model.transformer.layers.24.mlp.fc2",
192
+ "vision_model.transformer.layers.25.self_attn.q_proj",
193
+ "vision_model.transformer.layers.25.self_attn.k_proj",
194
+ "vision_model.transformer.layers.25.self_attn.v_proj",
195
+ "vision_model.transformer.layers.25.self_attn.o_proj",
196
+ "vision_model.transformer.layers.25.mlp.fc1",
197
+ "vision_model.transformer.layers.25.mlp.fc2",
198
+ "vision_model.transformer.layers.26.self_attn.q_proj",
199
+ "vision_model.transformer.layers.26.self_attn.k_proj",
200
+ "vision_model.transformer.layers.26.self_attn.v_proj",
201
+ "vision_model.transformer.layers.26.self_attn.o_proj",
202
+ "vision_model.transformer.layers.26.mlp.fc1",
203
+ "vision_model.transformer.layers.26.mlp.fc2",
204
+ "vision_model.transformer.layers.27.self_attn.q_proj",
205
+ "vision_model.transformer.layers.27.self_attn.k_proj",
206
+ "vision_model.transformer.layers.27.self_attn.v_proj",
207
+ "vision_model.transformer.layers.27.self_attn.o_proj",
208
+ "vision_model.transformer.layers.27.mlp.fc1",
209
+ "vision_model.transformer.layers.27.mlp.fc2",
210
+ "vision_model.transformer.layers.28.self_attn.q_proj",
211
+ "vision_model.transformer.layers.28.self_attn.k_proj",
212
+ "vision_model.transformer.layers.28.self_attn.v_proj",
213
+ "vision_model.transformer.layers.28.self_attn.o_proj",
214
+ "vision_model.transformer.layers.28.mlp.fc1",
215
+ "vision_model.transformer.layers.28.mlp.fc2",
216
+ "vision_model.transformer.layers.29.self_attn.q_proj",
217
+ "vision_model.transformer.layers.29.self_attn.k_proj",
218
+ "vision_model.transformer.layers.29.self_attn.v_proj",
219
+ "vision_model.transformer.layers.29.self_attn.o_proj",
220
+ "vision_model.transformer.layers.29.mlp.fc1",
221
+ "vision_model.transformer.layers.29.mlp.fc2",
222
+ "vision_model.transformer.layers.30.self_attn.q_proj",
223
+ "vision_model.transformer.layers.30.self_attn.k_proj",
224
+ "vision_model.transformer.layers.30.self_attn.v_proj",
225
+ "vision_model.transformer.layers.30.self_attn.o_proj",
226
+ "vision_model.transformer.layers.30.mlp.fc1",
227
+ "vision_model.transformer.layers.30.mlp.fc2",
228
+ "vision_model.transformer.layers.31.self_attn.q_proj",
229
+ "vision_model.transformer.layers.31.self_attn.k_proj",
230
+ "vision_model.transformer.layers.31.self_attn.v_proj",
231
+ "vision_model.transformer.layers.31.self_attn.o_proj",
232
+ "vision_model.transformer.layers.31.mlp.fc1",
233
+ "vision_model.transformer.layers.31.mlp.fc2",
234
+ "vision_model.global_transformer.layers.0.self_attn.q_proj",
235
+ "vision_model.global_transformer.layers.0.self_attn.k_proj",
236
+ "vision_model.global_transformer.layers.0.self_attn.v_proj",
237
+ "vision_model.global_transformer.layers.0.self_attn.o_proj",
238
+ "vision_model.global_transformer.layers.0.mlp.fc1",
239
+ "vision_model.global_transformer.layers.0.mlp.fc2",
240
+ "vision_model.global_transformer.layers.1.self_attn.q_proj",
241
+ "vision_model.global_transformer.layers.1.self_attn.k_proj",
242
+ "vision_model.global_transformer.layers.1.self_attn.v_proj",
243
+ "vision_model.global_transformer.layers.1.self_attn.o_proj",
244
+ "vision_model.global_transformer.layers.1.mlp.fc1",
245
+ "vision_model.global_transformer.layers.1.mlp.fc2",
246
+ "vision_model.global_transformer.layers.2.self_attn.q_proj",
247
+ "vision_model.global_transformer.layers.2.self_attn.k_proj",
248
+ "vision_model.global_transformer.layers.2.self_attn.v_proj",
249
+ "vision_model.global_transformer.layers.2.self_attn.o_proj",
250
+ "vision_model.global_transformer.layers.2.mlp.fc1",
251
+ "vision_model.global_transformer.layers.2.mlp.fc2",
252
+ "vision_model.global_transformer.layers.3.self_attn.q_proj",
253
+ "vision_model.global_transformer.layers.3.self_attn.k_proj",
254
+ "vision_model.global_transformer.layers.3.self_attn.v_proj",
255
+ "vision_model.global_transformer.layers.3.self_attn.o_proj",
256
+ "vision_model.global_transformer.layers.3.mlp.fc1",
257
+ "vision_model.global_transformer.layers.3.mlp.fc2",
258
+ "vision_model.global_transformer.layers.4.self_attn.q_proj",
259
+ "vision_model.global_transformer.layers.4.self_attn.k_proj",
260
+ "vision_model.global_transformer.layers.4.self_attn.v_proj",
261
+ "vision_model.global_transformer.layers.4.self_attn.o_proj",
262
+ "vision_model.global_transformer.layers.4.mlp.fc1",
263
+ "vision_model.global_transformer.layers.4.mlp.fc2",
264
+ "vision_model.global_transformer.layers.5.self_attn.q_proj",
265
+ "vision_model.global_transformer.layers.5.self_attn.k_proj",
266
+ "vision_model.global_transformer.layers.5.self_attn.v_proj",
267
+ "vision_model.global_transformer.layers.5.self_attn.o_proj",
268
+ "vision_model.global_transformer.layers.5.mlp.fc1",
269
+ "vision_model.global_transformer.layers.5.mlp.fc2",
270
+ "vision_model.global_transformer.layers.6.self_attn.q_proj",
271
+ "vision_model.global_transformer.layers.6.self_attn.k_proj",
272
+ "vision_model.global_transformer.layers.6.self_attn.v_proj",
273
+ "vision_model.global_transformer.layers.6.self_attn.o_proj",
274
+ "vision_model.global_transformer.layers.6.mlp.fc1",
275
+ "vision_model.global_transformer.layers.6.mlp.fc2",
276
+ "vision_model.global_transformer.layers.7.self_attn.q_proj",
277
+ "vision_model.global_transformer.layers.7.self_attn.k_proj",
278
+ "vision_model.global_transformer.layers.7.self_attn.v_proj",
279
+ "vision_model.global_transformer.layers.7.self_attn.o_proj",
280
+ "vision_model.global_transformer.layers.7.mlp.fc1",
281
+ "vision_model.global_transformer.layers.7.mlp.fc2",
282
+ "language_model.lm_head",
283
+ "multi_modal_projector"
284
+ ],
285
  "kv_cache_scheme": null,
286
  "quant_method": "compressed-tensors",
287
  "quantization_status": "compressed"
model-00001-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:838fc9fcd71b2ba08200bdc11968f6c57c35a56024ce3ecd6cb6818240c27848
3
- size 4988647178
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:317264170b96bc034493d33a86f54561bab33f17f5a8c3f6bfa24538783a8c88
3
+ size 4985205270
model-00002-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a2ff2bfb86394322607ebd274b1299844f6c395253c042d259d48521515cc8ad
3
- size 4960037248
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb639cf527cb15143d59bce368fdb2d2f38dbf696f3587859573c7e8008214f4
3
+ size 4993616092
model-00003-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4f483cab3840fe24ec23bd7cc3eb3b26ca0dd858b324ead50b19469182b897fa
3
- size 1329195940
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce6dcedf02f212dae7dae1f6f461b003debc071176c7dbaa02527b87c8dcf834
3
+ size 2641064860
model.safetensors.index.json CHANGED
The diff for this file is too large to render. See raw diff
 
recipe.yaml ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ DEFAULT_stage:
2
+ DEFAULT_modifiers:
3
+ QuantizationModifier:
4
+ ignore: ['re:.*lm_head', 're:multi_modal_projector.*', 're:vision_model.*']
5
+ targets: Linear
6
+ scheme: FP8_DYNAMIC