Hugging Face's logo

Journey9ni
/
Spatial_LLM

llava
Model card Files Community
Journey9ni commited on
Commit
18390d6
·
verified ·
1 Parent(s): 115409c

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +231 -0
  2. mm_projector.bin +3 -0
  3. trainer_state.json +0 -0
config.json ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "checkpoints/LLaVA-Video-7B-Qwen2",
3
+ "add_faster_video": false,
4
+ "add_time_instruction": true,
5
+ "architectures": [
6
+ "LlavaQwenForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "bos_token_id": 151643,
10
+ "eos_token_id": 151645,
11
+ "faster_token_stride": 10,
12
+ "force_sample": true,
13
+ "freeze_mm_mlp_adapter": false,
14
+ "freeze_mm_vision_resampler": false,
15
+ "fusion_block": "mlp2x_gelu",
16
+ "hidden_act": "silu",
17
+ "hidden_size": 3584,
18
+ "ignore_index": -100,
19
+ "image_aspect_ratio": "anyres_max_9",
20
+ "image_crop_resolution": null,
21
+ "image_grid_pinpoints": [
22
+ [
23
+ 384,
24
+ 384
25
+ ],
26
+ [
27
+ 384,
28
+ 768
29
+ ],
30
+ [
31
+ 384,
32
+ 1152
33
+ ],
34
+ [
35
+ 384,
36
+ 1536
37
+ ],
38
+ [
39
+ 384,
40
+ 1920
41
+ ],
42
+ [
43
+ 384,
44
+ 2304
45
+ ],
46
+ [
47
+ 768,
48
+ 384
49
+ ],
50
+ [
51
+ 768,
52
+ 768
53
+ ],
54
+ [
55
+ 768,
56
+ 1152
57
+ ],
58
+ [
59
+ 768,
60
+ 1536
61
+ ],
62
+ [
63
+ 768,
64
+ 1920
65
+ ],
66
+ [
67
+ 768,
68
+ 2304
69
+ ],
70
+ [
71
+ 1152,
72
+ 384
73
+ ],
74
+ [
75
+ 1152,
76
+ 768
77
+ ],
78
+ [
79
+ 1152,
80
+ 1152
81
+ ],
82
+ [
83
+ 1152,
84
+ 1536
85
+ ],
86
+ [
87
+ 1152,
88
+ 1920
89
+ ],
90
+ [
91
+ 1152,
92
+ 2304
93
+ ],
94
+ [
95
+ 1536,
96
+ 384
97
+ ],
98
+ [
99
+ 1536,
100
+ 768
101
+ ],
102
+ [
103
+ 1536,
104
+ 1152
105
+ ],
106
+ [
107
+ 1536,
108
+ 1536
109
+ ],
110
+ [
111
+ 1536,
112
+ 1920
113
+ ],
114
+ [
115
+ 1536,
116
+ 2304
117
+ ],
118
+ [
119
+ 1920,
120
+ 384
121
+ ],
122
+ [
123
+ 1920,
124
+ 768
125
+ ],
126
+ [
127
+ 1920,
128
+ 1152
129
+ ],
130
+ [
131
+ 1920,
132
+ 1536
133
+ ],
134
+ [
135
+ 1920,
136
+ 1920
137
+ ],
138
+ [
139
+ 1920,
140
+ 2304
141
+ ],
142
+ [
143
+ 2304,
144
+ 384
145
+ ],
146
+ [
147
+ 2304,
148
+ 768
149
+ ],
150
+ [
151
+ 2304,
152
+ 1152
153
+ ],
154
+ [
155
+ 2304,
156
+ 1536
157
+ ],
158
+ [
159
+ 2304,
160
+ 1920
161
+ ],
162
+ [
163
+ 2304,
164
+ 2304
165
+ ]
166
+ ],
167
+ "image_split_resolution": null,
168
+ "image_token_index": 151646,
169
+ "initializer_range": 0.02,
170
+ "intermediate_size": 18944,
171
+ "max_position_embeddings": 32768,
172
+ "max_window_layers": 28,
173
+ "mm_hidden_size": 1152,
174
+ "mm_newline_position": "grid",
175
+ "mm_patch_merge_type": "spatial_unpad",
176
+ "mm_projector_lr": null,
177
+ "mm_projector_type": "mlp2x_gelu",
178
+ "mm_resampler_type": null,
179
+ "mm_spatial_pool_mode": "bilinear",
180
+ "mm_spatial_pool_stride": 2,
181
+ "mm_spatial_tower": "cut3r",
182
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
183
+ "mm_use_im_patch_token": false,
184
+ "mm_use_im_start_end": false,
185
+ "mm_vision_select_feature": "patch",
186
+ "mm_vision_select_layer": -2,
187
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
188
+ "mm_vision_tower_lr": null,
189
+ "model_type": "llava",
190
+ "num_attention_heads": 28,
191
+ "num_hidden_layers": 28,
192
+ "num_key_value_heads": 4,
193
+ "pos_skipping_range": 4096,
194
+ "projector_hidden_act": "gelu",
195
+ "rms_norm_eps": 1e-06,
196
+ "rope_scaling": null,
197
+ "rope_theta": 1000000.0,
198
+ "sliding_window": 131072,
199
+ "spatial_tower": "cut3r",
200
+ "text_config": {
201
+ "model_type": "llama"
202
+ },
203
+ "tie_word_embeddings": false,
204
+ "tokenizer_model_max_length": 32768,
205
+ "tokenizer_padding_side": "right",
206
+ "torch_dtype": "bfloat16",
207
+ "transformers_version": "4.40.0.dev0",
208
+ "tune_fusion_block": true,
209
+ "tune_mm_mlp_adapter": true,
210
+ "tune_mm_vision_resampler": false,
211
+ "tune_spatial_tower": false,
212
+ "unfreeze_mm_vision_tower": false,
213
+ "use_cache": true,
214
+ "use_mm_proj": true,
215
+ "use_pos_skipping": false,
216
+ "use_sliding_window": false,
217
+ "vision_config": {
218
+ "hidden_size": 1024,
219
+ "image_size": 336,
220
+ "intermediate_size": 4096,
221
+ "model_type": "clip_vision_model",
222
+ "num_attention_heads": 16,
223
+ "num_hidden_layers": 24,
224
+ "patch_size": 14,
225
+ "projection_dim": 768,
226
+ "vocab_size": 32000
227
+ },
228
+ "vision_feature_layer": -2,
229
+ "vision_feature_select_strategy": "default",
230
+ "vision_tower_pretrained": null
231
+ }
mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d66f1f0aa2989a4f7b73fdb1b53b670f41d0eebcae595bd300ffa019b885a40
3
+ size 65174652
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff