Hugging Face's logo

AI-Safeguard
/
Ivy-VL-llava

Visual Question Answering
Transformers
Safetensors
English
Chinese
qwen2
text-generation
multimodal
llava
text-generation-inference
Inference Endpoints
Model card Files Community
2
Ivy1997 commited on
Commit
f5f2a99
•
1 Parent(s): 679a364

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +197 -202
config.json CHANGED
@@ -1,203 +1,198 @@
1
  {
2
- "_name_or_path": "/data/oss_bucket_0/linxi/logs/TBStars_008_Omni/s3_siglip_qwen2.5_3b_bs1_128xA100_20241130125751/",
3
- "add_faster_video": false,
4
- "add_time_instruction": false,
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": false,
13
- "hidden_act": "silu",
14
- "hidden_size": 2048,
15
- "image_aspect_ratio": "anyres_max_9",
16
- "image_crop_resolution": null,
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": null,
164
- "initializer_range": 0.02,
165
- "intermediate_size": 11008,
166
- "max_position_embeddings": 32768,
167
- "max_window_layers": 70,
168
- "mm_hidden_size": 1152,
169
- "mm_newline_position": "grid",
170
- "mm_patch_merge_type": "spatial_unpad",
171
- "mm_projector_lr": null,
172
- "mm_projector_type": "mlp2x_gelu",
173
- "mm_resampler_type": null,
174
- "mm_spatial_pool_mode": "bilinear",
175
- "mm_spatial_pool_stride": null,
176
- "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
177
- "mm_use_im_patch_token": false,
178
- "mm_use_im_start_end": false,
179
- "mm_vision_select_feature": "patch",
180
- "mm_vision_select_layer": -2,
181
- "mm_vision_tower": "/data/oss_bucket_0/linxi/pre-trained/google/siglip-so400m-patch14-384",
182
- "mm_vision_tower_lr": 2e-06,
183
- "model_type": "qwen2",
184
- "num_attention_heads": 16,
185
- "num_hidden_layers": 36,
186
- "num_key_value_heads": 2,
187
- "pos_skipping_range": 4096,
188
- "rms_norm_eps": 1e-06,
189
- "rope_scaling": null,
190
- "rope_theta": 1000000.0,
191
- "sliding_window": 32768,
192
- "tie_word_embeddings": true,
193
- "tokenizer_model_max_length": 16384,
194
- "tokenizer_padding_side": "right",
195
- "torch_dtype": "bfloat16",
196
- "transformers_version": "4.40.2",
197
- "use_cache": true,
198
- "use_mm_proj": true,
199
- "use_pos_skipping": false,
200
- "use_sliding_window": false,
201
- "vision_tower_pretrained": null,
202
- "vocab_size": 151936
203
- }
 
1
  {
2
+ "architectures": [
3
+ "LlavaQwenForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": 151643,
7
+ "eos_token_id": 151645,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 2048,
10
+ "image_aspect_ratio": "anyres_max_9",
11
+ "image_crop_resolution": null,
12
+ "image_grid_pinpoints": [
13
+ [
14
+ 384,
15
+ 384
16
+ ],
17
+ [
18
+ 384,
19
+ 768
20
+ ],
21
+ [
22
+ 384,
23
+ 1152
24
+ ],
25
+ [
26
+ 384,
27
+ 1536
28
+ ],
29
+ [
30
+ 384,
31
+ 1920
32
+ ],
33
+ [
34
+ 384,
35
+ 2304
36
+ ],
37
+ [
38
+ 768,
39
+ 384
40
+ ],
41
+ [
42
+ 768,
43
+ 768
44
+ ],
45
+ [
46
+ 768,
47
+ 1152
48
+ ],
49
+ [
50
+ 768,
51
+ 1536
52
+ ],
53
+ [
54
+ 768,
55
+ 1920
56
+ ],
57
+ [
58
+ 768,
59
+ 2304
60
+ ],
61
+ [
62
+ 1152,
63
+ 384
64
+ ],
65
+ [
66
+ 1152,
67
+ 768
68
+ ],
69
+ [
70
+ 1152,
71
+ 1152
72
+ ],
73
+ [
74
+ 1152,
75
+ 1536
76
+ ],
77
+ [
78
+ 1152,
79
+ 1920
80
+ ],
81
+ [
82
+ 1152,
83
+ 2304
84
+ ],
85
+ [
86
+ 1536,
87
+ 384
88
+ ],
89
+ [
90
+ 1536,
91
+ 768
92
+ ],
93
+ [
94
+ 1536,
95
+ 1152
96
+ ],
97
+ [
98
+ 1536,
99
+ 1536
100
+ ],
101
+ [
102
+ 1536,
103
+ 1920
104
+ ],
105
+ [
106
+ 1536,
107
+ 2304
108
+ ],
109
+ [
110
+ 1920,
111
+ 384
112
+ ],
113
+ [
114
+ 1920,
115
+ 768
116
+ ],
117
+ [
118
+ 1920,
119
+ 1152
120
+ ],
121
+ [
122
+ 1920,
123
+ 1536
124
+ ],
125
+ [
126
+ 1920,
127
+ 1920
128
+ ],
129
+ [
130
+ 1920,
131
+ 2304
132
+ ],
133
+ [
134
+ 2304,
135
+ 384
136
+ ],
137
+ [
138
+ 2304,
139
+ 768
140
+ ],
141
+ [
142
+ 2304,
143
+ 1152
144
+ ],
145
+ [
146
+ 2304,
147
+ 1536
148
+ ],
149
+ [
150
+ 2304,
151
+ 1920
152
+ ],
153
+ [
154
+ 2304,
155
+ 2304
156
+ ]
157
+ ],
158
+ "image_split_resolution": null,
159
+ "initializer_range": 0.02,
160
+ "intermediate_size": 11008,
161
+ "max_position_embeddings": 32768,
162
+ "max_window_layers": 70,
163
+ "mm_hidden_size": 1152,
164
+ "mm_newline_position": "grid",
165
+ "mm_patch_merge_type": "spatial_unpad",
166
+ "mm_projector_lr": null,
167
+ "mm_projector_type": "mlp2x_gelu",
168
+ "mm_resampler_type": null,
169
+ "mm_spatial_pool_mode": "bilinear",
170
+ "mm_spatial_pool_stride": null,
171
+ "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
172
+ "mm_use_im_patch_token": false,
173
+ "mm_use_im_start_end": false,
174
+ "mm_vision_select_feature": "patch",
175
+ "mm_vision_select_layer": -2,
176
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
177
+ "mm_vision_tower_lr": 2e-06,
178
+ "model_type": "qwen2",
179
+ "num_attention_heads": 16,
180
+ "num_hidden_layers": 36,
181
+ "num_key_value_heads": 2,
182
+ "pos_skipping_range": 4096,
183
+ "rms_norm_eps": 1e-06,
184
+ "rope_scaling": null,
185
+ "rope_theta": 1000000.0,
186
+ "sliding_window": 32768,
187
+ "tie_word_embeddings": true,
188
+ "tokenizer_model_max_length": 16384,
189
+ "tokenizer_padding_side": "right",
190
+ "torch_dtype": "bfloat16",
191
+ "transformers_version": "4.40.2",
192
+ "use_cache": true,
193
+ "use_mm_proj": true,
194
+ "use_pos_skipping": false,
195
+ "use_sliding_window": false,
196
+ "vision_tower_pretrained": null,
197
+ "vocab_size": 151936
198
+ }