Text Generation
Transformers
Safetensors
English
llava
multimodal
conversational
Eval Results
Inference Endpoints
ZhangYuanhan commited on
Commit
725314e
1 Parent(s): 23a51fa

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +222 -0
config.json ADDED
@@ -0,0 +1,222 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/vl-research/checkpoints/onevision/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-72B-Instruct-final_to_ov_am9_july31v4/checkpoint-9000",
3
+ "add_faster_video": false,
4
+ "add_time_instruction": true,
5
+ "vocab_size": 152064,
6
+ "architectures": [
7
+ "LlavaQwenForCausalLM"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "bos_token_id": 151643,
11
+ "eos_token_id": 151645,
12
+ "faster_token_stride": 10,
13
+ "force_sample": true,
14
+ "hidden_act": "silu",
15
+ "hidden_size": 8192,
16
+ "ignore_index": -100,
17
+ "image_aspect_ratio": "anyres_max_9",
18
+ "image_crop_resolution": null,
19
+ "image_grid_pinpoints": [
20
+ [
21
+ 384,
22
+ 384
23
+ ],
24
+ [
25
+ 384,
26
+ 768
27
+ ],
28
+ [
29
+ 384,
30
+ 1152
31
+ ],
32
+ [
33
+ 384,
34
+ 1536
35
+ ],
36
+ [
37
+ 384,
38
+ 1920
39
+ ],
40
+ [
41
+ 384,
42
+ 2304
43
+ ],
44
+ [
45
+ 768,
46
+ 384
47
+ ],
48
+ [
49
+ 768,
50
+ 768
51
+ ],
52
+ [
53
+ 768,
54
+ 1152
55
+ ],
56
+ [
57
+ 768,
58
+ 1536
59
+ ],
60
+ [
61
+ 768,
62
+ 1920
63
+ ],
64
+ [
65
+ 768,
66
+ 2304
67
+ ],
68
+ [
69
+ 1152,
70
+ 384
71
+ ],
72
+ [
73
+ 1152,
74
+ 768
75
+ ],
76
+ [
77
+ 1152,
78
+ 1152
79
+ ],
80
+ [
81
+ 1152,
82
+ 1536
83
+ ],
84
+ [
85
+ 1152,
86
+ 1920
87
+ ],
88
+ [
89
+ 1152,
90
+ 2304
91
+ ],
92
+ [
93
+ 1536,
94
+ 384
95
+ ],
96
+ [
97
+ 1536,
98
+ 768
99
+ ],
100
+ [
101
+ 1536,
102
+ 1152
103
+ ],
104
+ [
105
+ 1536,
106
+ 1536
107
+ ],
108
+ [
109
+ 1536,
110
+ 1920
111
+ ],
112
+ [
113
+ 1536,
114
+ 2304
115
+ ],
116
+ [
117
+ 1920,
118
+ 384
119
+ ],
120
+ [
121
+ 1920,
122
+ 768
123
+ ],
124
+ [
125
+ 1920,
126
+ 1152
127
+ ],
128
+ [
129
+ 1920,
130
+ 1536
131
+ ],
132
+ [
133
+ 1920,
134
+ 1920
135
+ ],
136
+ [
137
+ 1920,
138
+ 2304
139
+ ],
140
+ [
141
+ 2304,
142
+ 384
143
+ ],
144
+ [
145
+ 2304,
146
+ 768
147
+ ],
148
+ [
149
+ 2304,
150
+ 1152
151
+ ],
152
+ [
153
+ 2304,
154
+ 1536
155
+ ],
156
+ [
157
+ 2304,
158
+ 1920
159
+ ],
160
+ [
161
+ 2304,
162
+ 2304
163
+ ]
164
+ ],
165
+ "image_split_resolution": null,
166
+ "image_token_index": 151646,
167
+ "initializer_range": 0.02,
168
+ "intermediate_size": 29568,
169
+ "max_position_embeddings": 32768,
170
+ "max_window_layers": 80,
171
+ "mm_hidden_size": 1152,
172
+ "mm_newline_position": "grid",
173
+ "mm_patch_merge_type": "spatial_unpad",
174
+ "mm_projector_lr": null,
175
+ "mm_projector_type": "mlp2x_gelu",
176
+ "mm_resampler_type": null,
177
+ "mm_spatial_pool_mode": "bilinear",
178
+ "mm_spatial_pool_stride": 2,
179
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
180
+ "mm_use_im_patch_token": false,
181
+ "mm_use_im_start_end": false,
182
+ "mm_vision_select_feature": "patch",
183
+ "mm_vision_select_layer": -2,
184
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
185
+ "mm_vision_tower_lr": 2e-06,
186
+ "model_type": "llava",
187
+ "num_attention_heads": 64,
188
+ "num_hidden_layers": 80,
189
+ "num_key_value_heads": 8,
190
+ "pos_skipping_range": 4096,
191
+ "projector_hidden_act": "gelu",
192
+ "rms_norm_eps": 1e-06,
193
+ "rope_scaling": null,
194
+ "rope_theta": 1000000.0,
195
+ "sliding_window": 131072,
196
+ "text_config": {
197
+ "model_type": "llama"
198
+ },
199
+ "tie_word_embeddings": false,
200
+ "tokenizer_model_max_length": 32768,
201
+ "tokenizer_padding_side": "right",
202
+ "torch_dtype": "bfloat16",
203
+ "transformers_version": "4.40.0.dev0",
204
+ "use_cache": true,
205
+ "use_mm_proj": true,
206
+ "use_pos_skipping": false,
207
+ "use_sliding_window": false,
208
+ "vision_config": {
209
+ "hidden_size": 1024,
210
+ "image_size": 336,
211
+ "intermediate_size": 4096,
212
+ "model_type": "clip_vision_model",
213
+ "num_attention_heads": 16,
214
+ "num_hidden_layers": 24,
215
+ "patch_size": 14,
216
+ "projection_dim": 768,
217
+ "vocab_size": 32000
218
+ },
219
+ "vision_feature_layer": -2,
220
+ "vision_feature_select_strategy": "default",
221
+ "vision_tower_pretrained": null
222
+ }