Hugging Face's logo

jinghan23
/
lov1

Safetensors
llava
Model card Files Community
jinghan23 commited on
Commit
81c8989
·
verified ·
1 Parent(s): d80724f

Upload to Hugging Face - config.json

Browse files
Files changed (1) hide show
  1. config.json +217 -0
config.json ADDED
@@ -0,0 +1,217 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/project/mlmlongvideo/checkpoints/pure_text_extend_llava_qwen/checkpoint-200/",
3
+ "architectures": [
4
+ "LlavaQwenForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "eos_token_id": 151645,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 3584,
11
+ "ignore_index": -100,
12
+ "image_aspect_ratio": "anyres_max_9",
13
+ "image_crop_resolution": null,
14
+ "image_grid_pinpoints": [
15
+ [
16
+ 384,
17
+ 384
18
+ ],
19
+ [
20
+ 384,
21
+ 768
22
+ ],
23
+ [
24
+ 384,
25
+ 1152
26
+ ],
27
+ [
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,
157
+ 2304
158
+ ]
159
+ ],
160
+ "image_split_resolution": null,
161
+ "image_token_index": 151646,
162
+ "initializer_range": 0.02,
163
+ "intermediate_size": 18944,
164
+ "max_position_embeddings": 32768,
165
+ "max_window_layers": 28,
166
+ "mm_hidden_size": 1152,
167
+ "mm_newline_position": "one_token",
168
+ "mm_patch_merge_type": "spatial_unpad",
169
+ "mm_projector_lr": null,
170
+ "mm_projector_type": "mlp2x_gelu",
171
+ "mm_resampler_type": null,
172
+ "mm_spatial_pool_mode": "bilinear",
173
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
174
+ "mm_use_im_patch_token": false,
175
+ "mm_use_im_start_end": false,
176
+ "mm_vision_select_feature": "patch",
177
+ "mm_vision_select_layer": -2,
178
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
179
+ "mm_vision_tower_lr": 2e-06,
180
+ "model_type": "llava",
181
+ "num_attention_heads": 28,
182
+ "num_hidden_layers": 28,
183
+ "num_key_value_heads": 4,
184
+ "pos_skipping_range": 4096,
185
+ "projector_hidden_act": "gelu",
186
+ "rms_norm_eps": 1e-06,
187
+ "rope_scaling": null,
188
+ "rope_theta": 1000000000.0,
189
+ "sliding_window": 131072,
190
+ "text_config": {
191
+ "model_type": "llama",
192
+ "vocab_size": 152064
193
+ },
194
+ "tie_word_embeddings": false,
195
+ "tokenizer_model_max_length": 224000,
196
+ "tokenizer_padding_side": "right",
197
+ "torch_dtype": "bfloat16",
198
+ "transformers_version": "4.40.0.dev0",
199
+ "use_cache": true,
200
+ "use_mm_proj": true,
201
+ "use_pos_skipping": false,
202
+ "use_sliding_window": false,
203
+ "vision_config": {
204
+ "hidden_size": 1024,
205
+ "image_size": 336,
206
+ "intermediate_size": 4096,
207
+ "model_type": "clip_vision_model",
208
+ "num_attention_heads": 16,
209
+ "num_hidden_layers": 24,
210
+ "patch_size": 14,
211
+ "projection_dim": 768,
212
+ "vocab_size": 32000
213
+ },
214
+ "vision_feature_layer": -2,
215
+ "vision_feature_select_strategy": "default",
216
+ "vision_tower_pretrained": null
217
+ }