chmv2/CHMv2ForDepthEstimation/config.json

{
  "architectures": [
    "CHMv2ForDepthEstimation"
  ],
  "backbone_config": {
    "apply_layernorm": true,
    "attention_dropout": 0.0,
    "drop_path_rate": 0.0,
    "hidden_act": "gelu",
    "hidden_size": 32,
    "image_size": 32,
    "initializer_range": 0.02,
    "intermediate_size": 64,
    "key_bias": true,
    "layer_norm_eps": 1e-06,
    "layerscale_value": 1.0,
    "mlp_bias": true,
    "model_type": "dinov3_vit",
    "num_attention_heads": 4,
    "num_channels": 3,
    "num_hidden_layers": 2,
    "num_register_tokens": 0,
    "out_features": [
      "stage1",
      "stage2"
    ],
    "out_indices": [
      1,
      2
    ],
    "patch_size": 16,
    "pos_embed_jitter": null,
    "pos_embed_rescale": 2.0,
    "pos_embed_shift": null,
    "proj_bias": true,
    "query_bias": true,
    "reshape_hidden_states": true,
    "return_class_token": true,
    "rope_theta": 100.0,
    "stage_names": [
      "stem",
      "stage1",
      "stage2"
    ],
    "use_gated_mlp": false,
    "value_bias": true
  },
  "bins_strategy": "chmv2_mixlog",
  "dtype": "float32",
  "fusion_hidden_size": 16,
  "head_hidden_size": 16,
  "initializer_range": 0.02,
  "max_depth": 96.0,
  "min_depth": 0.001,
  "model_type": "chmv2",
  "norm_strategy": "chmv2_mixlog",
  "number_output_channels": 4,
  "patch_size": 16,
  "post_process_channels": [
    16,
    16
  ],
  "readout_type": "project",
  "reassemble_factors": [
    4,
    2
  ],
  "reassemble_hidden_size": 32,
  "transformers_version": "5.15.0.dev0"
}