Hugging Face's logo

mkd-dev
/
keural-vision-encoder-poc

Image Feature Extraction
Transformers
Safetensors
English
keural_vision
vision
vision-encoder
image-text
contrastive-learning
custom-architecture
vlm
Model card Files
xet
 Community
keural-vision-encoder-poc
File size: 2,848 Bytes
96faa83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
{
  "architectures": [
    "KeuralVisionModel"
  ],
  "default_token_budget": 256,
  "depth": 12,
  "dtype": "float32",
  "embed_dim": 384,
  "keural_config": {
    "data": {
      "captions_path": "data/indexes/cc3m_wds_train.tsv",
      "dataset_name": "cc3m+cc12m",
      "extra_sources": [
        {
          "captions_path": "data/indexes/cc12m_wds_train.tsv",
          "images_root": "data/cc12m"
        }
      ],
      "image_buckets": [
        [
          224,
          224
        ],
        [
          256,
          256
        ],
        [
          320,
          320
        ]
      ],
      "images_root": "data/cc3m",
      "max_caption_length": 64,
      "num_workers": 4,
      "pixel_mean": [
        0.485,
        0.456,
        0.406
      ],
      "pixel_std": [
        0.229,
        0.224,
        0.225
      ],
      "text_encoder_name": "sentence-transformers/all-MiniLM-L6-v2"
    },
    "loss": {
      "hct_alpha": 0.5,
      "lambda_hct": 1.0,
      "lambda_sal": 1.0,
      "learnable_bias_init": -10.0,
      "learnable_temperature_init": 2.3,
      "primary": "sigmoid_siglip",
      "saliency_beta": 0.01,
      "saliency_lambda_smooth": 4.0
    },
    "name": "keural-tiny-poc",
    "saliency": {
      "hidden_channels": 96
    },
    "spatial": {
      "depth": 12,
      "embed_dim": 384,
      "head_dim": 64,
      "mlp_ratio": 4,
      "norm_eps": 1e-06,
      "num_heads": 6,
      "rope_base": 10000.0,
      "use_pool_token": true
    },
    "stem": {
      "activation": "gelu",
      "channels": [
        64,
        128,
        256,
        384
      ],
      "kernel_size": 3
    },
    "tokenizer": {
      "budget_floors": {
        "detail": 0,
        "global": 4,
        "region": 8
      },
      "budget_split": {
        "detail": 0.7,
        "global": 0.05,
        "region": 0.25
      },
      "default_token_budget": 256,
      "gumbel_temperature_end": 0.1,
      "gumbel_temperature_start": 1.0,
      "max_token_budget": 1024,
      "nms_radius": {
        "detail": 1,
        "region": 2
      },
      "roi_align_size": 2,
      "saliency_smooth_kernel": {
        "detail": 3,
        "region": 5
      },
      "temperature_anneal_steps": 10000,
      "warmup_steps_fixed_grid": 5000
    },
    "training": {
      "batch_size_per_device": 256,
      "betas": [
        0.9,
        0.95
      ],
      "gradient_accumulation": 4,
      "gradient_checkpointing": false,
      "log_every": 50,
      "lr": 0.001,
      "lr_schedule": "cosine_with_warmup",
      "max_steps": 75000,
      "optimizer": "adamw",
      "precision": "bfloat16",
      "save_every": 1000,
      "seed": 42,
      "warmup_steps": 2000,
      "weight_decay": 0.05
    }
  },
  "max_token_budget": 1024,
  "model_type": "keural_vision",
  "num_heads": 6,
  "transformers_version": "4.57.2"
}