Upload config.json with huggingface_hub
Browse files- config.json +151 -0
config.json
ADDED
|
@@ -0,0 +1,151 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "Spec-Vision-instruct",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"SpecVisionForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "configuration_spec_vision.SpecVisionConfig",
|
| 9 |
+
"AutoModelForCausalLM": "modeling_spec_vision.SpecVisionForCausalLM"
|
| 10 |
+
},
|
| 11 |
+
"bos_token_id": 1,
|
| 12 |
+
"embd_layer": {
|
| 13 |
+
"embedding_cls": "image",
|
| 14 |
+
"hd_transform_order": "sub_glb",
|
| 15 |
+
"projection_cls": "mlp",
|
| 16 |
+
"use_hd_transform": true,
|
| 17 |
+
"with_learnable_separator": true
|
| 18 |
+
},
|
| 19 |
+
"embd_pdrop": 0.0,
|
| 20 |
+
"eos_token_id": 2,
|
| 21 |
+
"hidden_act": "silu",
|
| 22 |
+
"hidden_size": 3072,
|
| 23 |
+
"img_processor": {
|
| 24 |
+
"image_dim_out": 1024,
|
| 25 |
+
"model_name": "openai/clip-vit-large-patch14-336",
|
| 26 |
+
"name": "clip_vision_model",
|
| 27 |
+
"num_img_tokens": 144
|
| 28 |
+
},
|
| 29 |
+
"initializer_range": 0.02,
|
| 30 |
+
"intermediate_size": 8192,
|
| 31 |
+
"max_position_embeddings": 131072,
|
| 32 |
+
"model_type": "spec_vision",
|
| 33 |
+
"num_attention_heads": 32,
|
| 34 |
+
"num_hidden_layers": 32,
|
| 35 |
+
"num_key_value_heads": 32,
|
| 36 |
+
"original_max_position_embeddings": 4096,
|
| 37 |
+
"pad_token_id": 32000,
|
| 38 |
+
"resid_pdrop": 0.0,
|
| 39 |
+
"rms_norm_eps": 1e-05,
|
| 40 |
+
"rope_scaling": {
|
| 41 |
+
"long_factor": [
|
| 42 |
+
1.0800000429153442,
|
| 43 |
+
1.1100000143051147,
|
| 44 |
+
1.1399999856948853,
|
| 45 |
+
1.340000033378601,
|
| 46 |
+
1.5899999141693115,
|
| 47 |
+
1.600000023841858,
|
| 48 |
+
1.6200000047683716,
|
| 49 |
+
2.620000123977661,
|
| 50 |
+
3.2300000190734863,
|
| 51 |
+
3.2300000190734863,
|
| 52 |
+
4.789999961853027,
|
| 53 |
+
7.400000095367432,
|
| 54 |
+
7.700000286102295,
|
| 55 |
+
9.09000015258789,
|
| 56 |
+
12.199999809265137,
|
| 57 |
+
17.670000076293945,
|
| 58 |
+
24.46000099182129,
|
| 59 |
+
28.57000160217285,
|
| 60 |
+
30.420001983642578,
|
| 61 |
+
30.840002059936523,
|
| 62 |
+
32.590003967285156,
|
| 63 |
+
32.93000411987305,
|
| 64 |
+
42.320003509521484,
|
| 65 |
+
44.96000289916992,
|
| 66 |
+
50.340003967285156,
|
| 67 |
+
50.45000457763672,
|
| 68 |
+
57.55000305175781,
|
| 69 |
+
57.93000411987305,
|
| 70 |
+
58.21000289916992,
|
| 71 |
+
60.1400032043457,
|
| 72 |
+
62.61000442504883,
|
| 73 |
+
62.62000274658203,
|
| 74 |
+
62.71000289916992,
|
| 75 |
+
63.1400032043457,
|
| 76 |
+
63.1400032043457,
|
| 77 |
+
63.77000427246094,
|
| 78 |
+
63.93000411987305,
|
| 79 |
+
63.96000289916992,
|
| 80 |
+
63.970001220703125,
|
| 81 |
+
64.02999877929688,
|
| 82 |
+
64.06999969482422,
|
| 83 |
+
64.08000183105469,
|
| 84 |
+
64.12000274658203,
|
| 85 |
+
64.41000366210938,
|
| 86 |
+
64.4800033569336,
|
| 87 |
+
64.51000213623047,
|
| 88 |
+
64.52999877929688,
|
| 89 |
+
64.83999633789062
|
| 90 |
+
],
|
| 91 |
+
"short_factor": [
|
| 92 |
+
1.08,
|
| 93 |
+
1.1,
|
| 94 |
+
1.1300000000000001,
|
| 95 |
+
1.2800000000000002,
|
| 96 |
+
1.3100000000000003,
|
| 97 |
+
1.4500000000000004,
|
| 98 |
+
1.4500000000000004,
|
| 99 |
+
1.9500000000000008,
|
| 100 |
+
2.030000000000001,
|
| 101 |
+
2.4299999999999926,
|
| 102 |
+
2.5699999999999896,
|
| 103 |
+
2.9499999999999815,
|
| 104 |
+
3.729999999999965,
|
| 105 |
+
3.869999999999962,
|
| 106 |
+
4.189999999999955,
|
| 107 |
+
4.43999999999995,
|
| 108 |
+
4.6399999999999455,
|
| 109 |
+
4.979999999999938,
|
| 110 |
+
5.159999999999934,
|
| 111 |
+
5.279999999999932,
|
| 112 |
+
5.759999999999922,
|
| 113 |
+
5.889999999999919,
|
| 114 |
+
5.889999999999919,
|
| 115 |
+
5.969999999999917,
|
| 116 |
+
6.089999999999915,
|
| 117 |
+
6.2799999999999105,
|
| 118 |
+
6.7699999999999,
|
| 119 |
+
6.8899999999998975,
|
| 120 |
+
7.109999999999893,
|
| 121 |
+
7.129999999999892,
|
| 122 |
+
7.179999999999891,
|
| 123 |
+
7.289999999999889,
|
| 124 |
+
7.339999999999888,
|
| 125 |
+
7.559999999999883,
|
| 126 |
+
7.619999999999882,
|
| 127 |
+
7.69999999999988,
|
| 128 |
+
7.879999999999876,
|
| 129 |
+
7.879999999999876,
|
| 130 |
+
7.879999999999876,
|
| 131 |
+
7.939999999999875,
|
| 132 |
+
7.949999999999875,
|
| 133 |
+
7.979999999999874,
|
| 134 |
+
8.19999999999987,
|
| 135 |
+
8.439999999999864,
|
| 136 |
+
8.469999999999864,
|
| 137 |
+
8.589999999999861,
|
| 138 |
+
8.809999999999857,
|
| 139 |
+
8.999999999999853
|
| 140 |
+
],
|
| 141 |
+
"type": "su"
|
| 142 |
+
},
|
| 143 |
+
"rope_theta": 10000.0,
|
| 144 |
+
"sliding_window": 262144,
|
| 145 |
+
"tie_word_embeddings": false,
|
| 146 |
+
"torch_dtype": "bfloat16",
|
| 147 |
+
"transformers_version": "4.38.1",
|
| 148 |
+
"use_cache": true,
|
| 149 |
+
"vocab_size": 32064,
|
| 150 |
+
"_attn_implementation": "flash_attention_2"
|
| 151 |
+
}
|