Hugging Face's logo

llava-hf
/
llava-onevision-qwen2-72b-ov-hf

Image-Text-to-Text
Safetensors
English
Chinese
llava_onevision
vision
conversational
Model card Files Community
1
RaushanTurganbay HF staff commited on
Commit
6d456f0
•
1 Parent(s): 2465bad

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +195 -0
config.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlavaNextForConditionalGeneration"
4
+ ],
5
+ "ignore_index": -100,
6
+ "image_grid_pinpoints": [
7
+ [
8
+ 384,
9
+ 384
10
+ ],
11
+ [
12
+ 384,
13
+ 768
14
+ ],
15
+ [
16
+ 384,
17
+ 1152
18
+ ],
19
+ [
20
+ 384,
21
+ 1536
22
+ ],
23
+ [
24
+ 384,
25
+ 1920
26
+ ],
27
+ [
28
+ 384,
29
+ 2304
30
+ ],
31
+ [
32
+ 768,
33
+ 384
34
+ ],
35
+ [
36
+ 768,
37
+ 768
38
+ ],
39
+ [
40
+ 768,
41
+ 1152
42
+ ],
43
+ [
44
+ 768,
45
+ 1536
46
+ ],
47
+ [
48
+ 768,
49
+ 1920
50
+ ],
51
+ [
52
+ 768,
53
+ 2304
54
+ ],
55
+ [
56
+ 1152,
57
+ 384
58
+ ],
59
+ [
60
+ 1152,
61
+ 768
62
+ ],
63
+ [
64
+ 1152,
65
+ 1152
66
+ ],
67
+ [
68
+ 1152,
69
+ 1536
70
+ ],
71
+ [
72
+ 1152,
73
+ 1920
74
+ ],
75
+ [
76
+ 1152,
77
+ 2304
78
+ ],
79
+ [
80
+ 1536,
81
+ 384
82
+ ],
83
+ [
84
+ 1536,
85
+ 768
86
+ ],
87
+ [
88
+ 1536,
89
+ 1152
90
+ ],
91
+ [
92
+ 1536,
93
+ 1536
94
+ ],
95
+ [
96
+ 1536,
97
+ 1920
98
+ ],
99
+ [
100
+ 1536,
101
+ 2304
102
+ ],
103
+ [
104
+ 1920,
105
+ 384
106
+ ],
107
+ [
108
+ 1920,
109
+ 768
110
+ ],
111
+ [
112
+ 1920,
113
+ 1152
114
+ ],
115
+ [
116
+ 1920,
117
+ 1536
118
+ ],
119
+ [
120
+ 1920,
121
+ 1920
122
+ ],
123
+ [
124
+ 1920,
125
+ 2304
126
+ ],
127
+ [
128
+ 2304,
129
+ 384
130
+ ],
131
+ [
132
+ 2304,
133
+ 768
134
+ ],
135
+ [
136
+ 2304,
137
+ 1152
138
+ ],
139
+ [
140
+ 2304,
141
+ 1536
142
+ ],
143
+ [
144
+ 2304,
145
+ 1920
146
+ ],
147
+ [
148
+ 2304,
149
+ 2304
150
+ ]
151
+ ],
152
+ "image_token_index": 151646,
153
+ "model_type": "llava_next",
154
+ "projector_hidden_act": "gelu",
155
+ "text_config": {
156
+ "_name_or_path": "Qwen/Qwen2-72B-Instruct",
157
+ "architectures": [
158
+ "Qwen2ForCausalLM"
159
+ ],
160
+ "bos_token_id": 151643,
161
+ "eos_token_id": 151645,
162
+ "hidden_size": 8192,
163
+ "intermediate_size": 29568,
164
+ "max_position_embeddings": 32768,
165
+ "max_window_layers": 80,
166
+ "model_type": "qwen2",
167
+ "num_attention_heads": 64,
168
+ "num_hidden_layers": 80,
169
+ "num_key_value_heads": 8,
170
+ "rope_theta": 1000000.0,
171
+ "sliding_window": null,
172
+ "torch_dtype": "bfloat16",
173
+ "use_sliding_window": false,
174
+ "vocab_size": 152128
175
+ },
176
+ "tie_word_embeddings": false,
177
+ "torch_dtype": "float16",
178
+ "transformers_version": "4.45.0.dev0",
179
+ "use_image_newline_parameter": true,
180
+ "vision_aspect_ratio": "anyres_max_9",
181
+ "vision_config": {
182
+ "hidden_act": "gelu_pytorch_tanh",
183
+ "hidden_size": 1152,
184
+ "image_size": 384,
185
+ "intermediate_size": 4304,
186
+ "layer_norm_eps": 1e-06,
187
+ "model_type": "siglip_vision_model",
188
+ "num_attention_heads": 16,
189
+ "num_hidden_layers": 26,
190
+ "patch_size": 14,
191
+ "vision_use_head": false
192
+ },
193
+ "vision_feature_layer": -1,
194
+ "vision_feature_select_strategy": "full"
195
+ }