Upload config
Browse files- config.json +78 -0
config.json
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"InternVLForConditionalGeneration"
|
4 |
+
],
|
5 |
+
"downsample_ratio": 0.5,
|
6 |
+
"image_seq_length": 256,
|
7 |
+
"image_token_index": 151667,
|
8 |
+
"model_type": "internvl",
|
9 |
+
"projector_hidden_act": "gelu",
|
10 |
+
"text_config": {
|
11 |
+
"architectures": [
|
12 |
+
"Qwen2ForCausalLM"
|
13 |
+
],
|
14 |
+
"attention_dropout": 0.0,
|
15 |
+
"bos_token_id": 151643,
|
16 |
+
"eos_token_id": 151645,
|
17 |
+
"hidden_act": "silu",
|
18 |
+
"hidden_size": 896,
|
19 |
+
"initializer_range": 0.02,
|
20 |
+
"intermediate_size": 4864,
|
21 |
+
"max_position_embeddings": 32768,
|
22 |
+
"max_window_layers": 70,
|
23 |
+
"model_type": "qwen2",
|
24 |
+
"num_attention_heads": 14,
|
25 |
+
"num_hidden_layers": 24,
|
26 |
+
"num_key_value_heads": 2,
|
27 |
+
"rms_norm_eps": 1e-06,
|
28 |
+
"rope_scaling": {
|
29 |
+
"factor": 2.0,
|
30 |
+
"rope_type": "dynamic",
|
31 |
+
"type": "dynamic"
|
32 |
+
},
|
33 |
+
"rope_theta": 1000000.0,
|
34 |
+
"sliding_window": null,
|
35 |
+
"torch_dtype": "bfloat16",
|
36 |
+
"use_cache": true,
|
37 |
+
"use_sliding_window": false,
|
38 |
+
"vocab_size": 151674
|
39 |
+
},
|
40 |
+
"transformers_version": "4.52.0.dev0",
|
41 |
+
"vision_config": {
|
42 |
+
"architectures": [
|
43 |
+
"InternVisionModel"
|
44 |
+
],
|
45 |
+
"attention_bias": true,
|
46 |
+
"attention_dropout": 0.0,
|
47 |
+
"dropout": 0.0,
|
48 |
+
"hidden_act": "gelu",
|
49 |
+
"hidden_dropout_prob": 0.0,
|
50 |
+
"hidden_size": 1024,
|
51 |
+
"image_size": [
|
52 |
+
448,
|
53 |
+
448
|
54 |
+
],
|
55 |
+
"initializer_factor": 0.1,
|
56 |
+
"initializer_range": 1e-10,
|
57 |
+
"intermediate_size": 4096,
|
58 |
+
"layer_norm_eps": 1e-06,
|
59 |
+
"layer_scale_init_value": 0.1,
|
60 |
+
"model_type": "internvl_vision",
|
61 |
+
"norm_type": "layer_norm",
|
62 |
+
"num_attention_heads": 16,
|
63 |
+
"num_channels": 3,
|
64 |
+
"num_hidden_layers": 24,
|
65 |
+
"patch_size": [
|
66 |
+
14,
|
67 |
+
14
|
68 |
+
],
|
69 |
+
"projection_dropout": 0.0,
|
70 |
+
"torch_dtype": "bfloat16",
|
71 |
+
"use_absolute_position_embeddings": true,
|
72 |
+
"use_mask_token": false,
|
73 |
+
"use_mean_pooling": true,
|
74 |
+
"use_qk_norm": false
|
75 |
+
},
|
76 |
+
"vision_feature_layer": -1,
|
77 |
+
"vision_feature_select_strategy": "default"
|
78 |
+
}
|