kf_config.json

{
  "library_name": "kerasformers",
  "kerasformers_version": "1.2.5",
  "model_module": "kerasformers.models.tipsv2_dpt",
  "model_class": "Tipsv2DptDensePredict",
  "variant": "tipsv2-g14-dpt",
  "weights": "model.weights.h5",
  "schema_version": 2,
  "model_type": "tipsv2_dpt",
  "vision_config": {
    "image_size": 448,
    "patch_size": 14,
    "num_register_tokens": 1,
    "vision_hidden_dim": 1536,
    "vision_num_layers": 40,
    "vision_num_heads": 24,
    "vision_mlp_ratio": 4,
    "vision_use_swiglu_ffn": true,
    "vision_layerscale_value": 1.0,
    "vision_layer_norm_eps": 1e-06,
    "out_indices": [
      10,
      20,
      30,
      40
    ],
    "neck_hidden_sizes": [
      192,
      384,
      768,
      1536
    ],
    "reassemble_factors": [
      4,
      2,
      1,
      0.5
    ],
    "fusion_hidden_size": 256,
    "readout_activation": "gelu_tanh",
    "num_depth_bins": 256,
    "min_depth": 0.001,
    "max_depth": 10.0,
    "depth_decoder_activation": "relu",
    "num_labels": 150
  }
}