Hugging Face's logo

lmms-lab
/
LLaVA-Video-7B-Qwen2-Video-Only

Text Generation
Transformers
Safetensors
English
llava
multimodal
conversational
Eval Results
Inference Endpoints
Model card Files Community
1
ZhangYuanhan commited on
Commit
5450c99
•
1 Parent(s): fcfba4a

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +208 -0
config.json ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/tiktok-mm-2/aiic/users/wujinming/code/work_dir/OV-Qwen2-7B-AM9",
3
+ "add_faster_video": false,
4
+ "architectures": [
5
+ "LlavaQwenForCausalLM"
6
+ ],
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 151643,
9
+ "delay_load": false,
10
+ "eos_token_id": 151645,
11
+ "freeze_mm_mlp_adapter": false,
12
+ "freeze_mm_vision_resampler": false,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 3584,
15
+ "image_aspect_ratio": "anyres_max_9",
16
+ "image_crop_resolution": 224,
17
+ "image_grid_pinpoints": [
18
+ [
19
+ 384,
20
+ 384
21
+ ],
22
+ [
23
+ 384,
24
+ 768
25
+ ],
26
+ [
27
+ 384,
28
+ 1152
29
+ ],
30
+ [
31
+ 384,
32
+ 1536
33
+ ],
34
+ [
35
+ 384,
36
+ 1920
37
+ ],
38
+ [
39
+ 384,
40
+ 2304
41
+ ],
42
+ [
43
+ 768,
44
+ 384
45
+ ],
46
+ [
47
+ 768,
48
+ 768
49
+ ],
50
+ [
51
+ 768,
52
+ 1152
53
+ ],
54
+ [
55
+ 768,
56
+ 1536
57
+ ],
58
+ [
59
+ 768,
60
+ 1920
61
+ ],
62
+ [
63
+ 768,
64
+ 2304
65
+ ],
66
+ [
67
+ 1152,
68
+ 384
69
+ ],
70
+ [
71
+ 1152,
72
+ 768
73
+ ],
74
+ [
75
+ 1152,
76
+ 1152
77
+ ],
78
+ [
79
+ 1152,
80
+ 1536
81
+ ],
82
+ [
83
+ 1152,
84
+ 1920
85
+ ],
86
+ [
87
+ 1152,
88
+ 2304
89
+ ],
90
+ [
91
+ 1536,
92
+ 384
93
+ ],
94
+ [
95
+ 1536,
96
+ 768
97
+ ],
98
+ [
99
+ 1536,
100
+ 1152
101
+ ],
102
+ [
103
+ 1536,
104
+ 1536
105
+ ],
106
+ [
107
+ 1536,
108
+ 1920
109
+ ],
110
+ [
111
+ 1536,
112
+ 2304
113
+ ],
114
+ [
115
+ 1920,
116
+ 384
117
+ ],
118
+ [
119
+ 1920,
120
+ 768
121
+ ],
122
+ [
123
+ 1920,
124
+ 1152
125
+ ],
126
+ [
127
+ 1920,
128
+ 1536
129
+ ],
130
+ [
131
+ 1920,
132
+ 1920
133
+ ],
134
+ [
135
+ 1920,
136
+ 2304
137
+ ],
138
+ [
139
+ 2304,
140
+ 384
141
+ ],
142
+ [
143
+ 2304,
144
+ 768
145
+ ],
146
+ [
147
+ 2304,
148
+ 1152
149
+ ],
150
+ [
151
+ 2304,
152
+ 1536
153
+ ],
154
+ [
155
+ 2304,
156
+ 1920
157
+ ],
158
+ [
159
+ 2304,
160
+ 2304
161
+ ]
162
+ ],
163
+ "image_split_resolution": 224,
164
+ "initializer_range": 0.02,
165
+ "intermediate_size": 18944,
166
+ "max_position_embeddings": 32768,
167
+ "max_window_layers": 28,
168
+ "mm_hidden_size": 1152,
169
+ "mm_newline_position": "grid",
170
+ "mm_patch_merge_type": "spatial_unpad",
171
+ "mm_pooling_position": "after",
172
+ "mm_projector_lr": null,
173
+ "mm_projector_type": "mlp2x_gelu",
174
+ "mm_resampler_type": null,
175
+ "mm_spatial_pool_mode": "average",
176
+ "mm_spatial_pool_stride": 2,
177
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
178
+ "mm_use_im_patch_token": false,
179
+ "mm_use_im_start_end": false,
180
+ "mm_vision_select_feature": "patch",
181
+ "mm_vision_select_layer": -2,
182
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
183
+ "mm_vision_tower_lr": null,
184
+ "modality_max_length": "None",
185
+ "model_type": "qwen2",
186
+ "num_attention_heads": 28,
187
+ "num_hidden_layers": 28,
188
+ "num_key_value_heads": 4,
189
+ "pos_skipping_range": 4096,
190
+ "rms_norm_eps": 1e-06,
191
+ "rope_scaling": null,
192
+ "rope_theta": 1000000.0,
193
+ "sliding_window": 131072,
194
+ "tie_word_embeddings": false,
195
+ "tokenizer_model_max_length": 32768,
196
+ "tokenizer_padding_side": "right",
197
+ "torch_dtype": "bfloat16",
198
+ "transformers_version": "4.39.2",
199
+ "tune_mm_mlp_adapter": false,
200
+ "tune_mm_vision_resampler": false,
201
+ "unfreeze_mm_vision_tower": false,
202
+ "use_cache": true,
203
+ "use_mm_proj": true,
204
+ "use_pos_skipping": false,
205
+ "use_sliding_window": false,
206
+ "vision_tower_pretrained": null,
207
+ "vocab_size": 152064
208
+ }