{
  "decoder_type": "pos_free",
  "distill_dim": ["EC_1280", "EP_1280", "DC_1280", "DP_1280"],
  "distill_location": "encoder-decoder-separate",
  "distill_mode": "dinov3_fancy",
  "embedding_dim": 32,
  "hypocnn_cfg": {
    "hid_dim": 256,
    "in_dim": 256,
    "ks": "3",
    "out_bias": 0.0,
    "out_dim": 3,
    "strds": "4"
  },
  "hyponet_cfg": {
    "depth": 3,
    "force_act": true,
    "hidden_dim": "256_256",
    "in_dim": 2,
    "out_bias": 0.0,
    "out_dim": 256,
    "pe_dim": 64,
    "strides": "1_1_1",
    "use_pe": true
  },
  "is_patch_mode": true,
  "mod_idxs": "1",
  "n_groups": null,
  "tokenizer_cfg": {
    "decoder_dim": 1280,
    "dim": 1024,
    "fixed_posemb": false,
    "frozen_posemb": false,
    "input_size": 256,
    "mae_init": false,
    "mask_dim": 1280,
    "mask_ratio": 0.0,
    "modern": true,
    "patch_size": 16,
    "use_decoder": false,
    "use_transcoder": false
  },
  "transformer_decoder_cfg": {
    "attention": false,
    "depth": 4,
    "dim": 1280,
    "ff_dim": 4096,
    "head_dim": 80,
    "n_head": 16
  },
  "transformer_encoder_cfg": {
    "dim": 1024,
    "img_size": 512,
    "mae_init": false,
    "name": "vitl_16_norm_only_rope_fixed_ft",
    "use_registers": false
  },
  "transformer_transcoder_cfg": {
    "dim": -1
  },
  "use_global_token": true,
  "use_hypocnn": true
}