config.json

{
  "decoder_args": {
    "attn_bias_for_inference_enabled": false,
    "attn_drop": 0.0,
    "attn_implementation": "flash_attention",
    "decoder_type": "fast3r",
    "depth": 24,
    "drop": 0.0,
    "embed_dim": 1024,
    "enc_embed_dim": 1024,
    "mlp_ratio": 4.0,
    "num_heads": 16,
    "qkv_bias": true,
    "random_image_idx_embedding": true
  },
  "encoder_args": {
    "attn_implementation": "flash_attention",
    "depth": 24,
    "embed_dim": 1024,
    "encoder_type": "croco",
    "img_size": 512,
    "mlp_ratio": 4,
    "num_heads": 16,
    "patch_embed_cls": "PatchEmbedDust3R",
    "patch_size": 16,
    "pos_embed": "RoPE100"
  },
  "head_args": {
    "conf_mode": [
      "exp",
      1,
      Infinity
    ],
    "depth_mode": [
      "exp",
      -Infinity,
      Infinity
    ],
    "head_type": "dpt",
    "landscape_only": false,
    "output_mode": "pts3d",
    "patch_size": 16,
    "with_local_head": true
  }
}