Create config.json
Browse files- config.json +66 -24
config.json
CHANGED
|
@@ -1,28 +1,70 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
-
"
|
| 4 |
],
|
| 5 |
-
"
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
"max_position_embeddings": 16384,
|
| 12 |
-
"model_type": "llama",
|
| 13 |
-
"num_attention_heads": 16,
|
| 14 |
-
"num_hidden_layers": 24,
|
| 15 |
-
"num_key_value_heads": 16,
|
| 16 |
-
"pretraining_tp": 1,
|
| 17 |
-
"rms_norm_eps": 1e-06,
|
| 18 |
-
"rope_scaling": {
|
| 19 |
-
"factor": 4.0,
|
| 20 |
-
"type": "linear"
|
| 21 |
},
|
| 22 |
-
"
|
| 23 |
-
"
|
| 24 |
-
"
|
| 25 |
-
"
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
{
|
| 2 |
+
"aligner_config": {
|
| 3 |
+
"cls": "MlpProjector",
|
| 4 |
+
"model_type": "aligner",
|
| 5 |
+
"params": {
|
| 6 |
+
"depth": 2,
|
| 7 |
+
"input_dim": 1024,
|
| 8 |
+
"n_embed": 4096,
|
| 9 |
+
"projector_type": "low_high_hybrid_split_mlp_gelu"
|
| 10 |
+
}
|
| 11 |
+
},
|
| 12 |
"architectures": [
|
| 13 |
+
"MultiModalityCausalLM"
|
| 14 |
],
|
| 15 |
+
"language_config": {
|
| 16 |
+
"max_position_embeddings": 16384,
|
| 17 |
+
"model_type": "llama",
|
| 18 |
+
"num_hidden_layers": 30,
|
| 19 |
+
"torch_dtype": "float16",
|
| 20 |
+
"vocab_size": 102400
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
},
|
| 22 |
+
"model_type": "multi_modality",
|
| 23 |
+
"torch_dtype": "float16",
|
| 24 |
+
"transformers_version": "4.38.2",
|
| 25 |
+
"vision_config": {
|
| 26 |
+
"cls": "HybridVisionTower",
|
| 27 |
+
"model_type": "vision",
|
| 28 |
+
"params": {
|
| 29 |
+
"concat_type": "tuple",
|
| 30 |
+
"freeze_high": true,
|
| 31 |
+
"freeze_low": true,
|
| 32 |
+
"high_res_cfg": {
|
| 33 |
+
"ckpt_path": "",
|
| 34 |
+
"image_size": 1024,
|
| 35 |
+
"model_name": "sam_b_downsample",
|
| 36 |
+
"output_dim": 1024,
|
| 37 |
+
"pixel_mean": [
|
| 38 |
+
0.48145466,
|
| 39 |
+
0.4578275,
|
| 40 |
+
0.40821073
|
| 41 |
+
],
|
| 42 |
+
"pixel_std": [
|
| 43 |
+
0.26862954,
|
| 44 |
+
0.26130258,
|
| 45 |
+
0.27577711
|
| 46 |
+
],
|
| 47 |
+
"select_feature": "same",
|
| 48 |
+
"select_layer": -1
|
| 49 |
+
},
|
| 50 |
+
"low_res_cfg": {
|
| 51 |
+
"ckpt_path": "",
|
| 52 |
+
"image_size": 384,
|
| 53 |
+
"model_name": "siglip_large_patch16_384",
|
| 54 |
+
"output_dim": 1024,
|
| 55 |
+
"pixel_mean": [
|
| 56 |
+
0.5,
|
| 57 |
+
0.5,
|
| 58 |
+
0.5
|
| 59 |
+
],
|
| 60 |
+
"pixel_std": [
|
| 61 |
+
0.5,
|
| 62 |
+
0.5,
|
| 63 |
+
0.5
|
| 64 |
+
],
|
| 65 |
+
"select_feature": "same",
|
| 66 |
+
"select_layer": -1
|
| 67 |
+
}
|
| 68 |
+
}
|
| 69 |
+
}
|
| 70 |
+
}
|