Update config.json
Browse files- config.json +2 -10
config.json
CHANGED
|
@@ -13,7 +13,6 @@
|
|
| 13 |
"freeze_mm_mlp_adapter": true,
|
| 14 |
"hidden_act": "silu",
|
| 15 |
"hidden_size": 3584,
|
| 16 |
-
"history_compression": "196token",
|
| 17 |
"ignore_index": -100,
|
| 18 |
"image_aspect_ratio": "anyres_max_9",
|
| 19 |
"image_crop_resolution": null,
|
|
@@ -183,14 +182,11 @@
|
|
| 183 |
"mm_use_im_start_end": false,
|
| 184 |
"mm_vision_select_feature": "patch",
|
| 185 |
"mm_vision_select_layer": -2,
|
| 186 |
-
"mm_vision_tower": "/
|
| 187 |
-
"mm_vision_tower_lr": 5e-06,
|
| 188 |
"model_type": "llava",
|
| 189 |
"multimodal_projector_bias": true,
|
| 190 |
"num_attention_heads": 28,
|
| 191 |
-
"num_future_steps": 4,
|
| 192 |
"num_hidden_layers": 28,
|
| 193 |
-
"num_history": 8,
|
| 194 |
"num_key_value_heads": 4,
|
| 195 |
"pos_skipping_range": 4096,
|
| 196 |
"pretrain_mm_mlp_adapter": null,
|
|
@@ -199,7 +195,6 @@
|
|
| 199 |
"rope_scaling": null,
|
| 200 |
"rope_theta": 1000000.0,
|
| 201 |
"sliding_window": 131072,
|
| 202 |
-
"stop_weight": 1.0,
|
| 203 |
"text_config": {
|
| 204 |
"attention_bias": false,
|
| 205 |
"attention_dropout": 0.0,
|
|
@@ -230,7 +225,6 @@
|
|
| 230 |
"transformers_version": "4.51.0",
|
| 231 |
"tune_mm_mlp_adapter": false,
|
| 232 |
"use_cache": true,
|
| 233 |
-
"use_memory_tokens": true,
|
| 234 |
"use_mm_proj": true,
|
| 235 |
"use_pos_skipping": false,
|
| 236 |
"use_sliding_window": false,
|
|
@@ -254,9 +248,7 @@
|
|
| 254 |
},
|
| 255 |
"vision_feature_layer": -2,
|
| 256 |
"vision_feature_select_strategy": "default",
|
| 257 |
-
"vision_tower": "/
|
| 258 |
"vision_tower_pretrained": null,
|
| 259 |
"vocab_size": 152064,
|
| 260 |
-
"voxel_range": 5.0,
|
| 261 |
-
"voxel_size": 0.5
|
| 262 |
}
|
|
|
|
| 13 |
"freeze_mm_mlp_adapter": true,
|
| 14 |
"hidden_act": "silu",
|
| 15 |
"hidden_size": 3584,
|
|
|
|
| 16 |
"ignore_index": -100,
|
| 17 |
"image_aspect_ratio": "anyres_max_9",
|
| 18 |
"image_crop_resolution": null,
|
|
|
|
| 182 |
"mm_use_im_start_end": false,
|
| 183 |
"mm_vision_select_feature": "patch",
|
| 184 |
"mm_vision_select_layer": -2,
|
| 185 |
+
"mm_vision_tower": "google/siglip-so400m-patch14-384",
|
|
|
|
| 186 |
"model_type": "llava",
|
| 187 |
"multimodal_projector_bias": true,
|
| 188 |
"num_attention_heads": 28,
|
|
|
|
| 189 |
"num_hidden_layers": 28,
|
|
|
|
| 190 |
"num_key_value_heads": 4,
|
| 191 |
"pos_skipping_range": 4096,
|
| 192 |
"pretrain_mm_mlp_adapter": null,
|
|
|
|
| 195 |
"rope_scaling": null,
|
| 196 |
"rope_theta": 1000000.0,
|
| 197 |
"sliding_window": 131072,
|
|
|
|
| 198 |
"text_config": {
|
| 199 |
"attention_bias": false,
|
| 200 |
"attention_dropout": 0.0,
|
|
|
|
| 225 |
"transformers_version": "4.51.0",
|
| 226 |
"tune_mm_mlp_adapter": false,
|
| 227 |
"use_cache": true,
|
|
|
|
| 228 |
"use_mm_proj": true,
|
| 229 |
"use_pos_skipping": false,
|
| 230 |
"use_sliding_window": false,
|
|
|
|
| 248 |
},
|
| 249 |
"vision_feature_layer": -2,
|
| 250 |
"vision_feature_select_strategy": "default",
|
| 251 |
+
"vision_tower": "google/siglip-so400m-patch14-384",
|
| 252 |
"vision_tower_pretrained": null,
|
| 253 |
"vocab_size": 152064,
|
|
|
|
|
|
|
| 254 |
}
|