config.json

{
  "camera_head": true,
  "decode_chunk_size": 128,
  "decoder_depth": 2,
  "decoder_embed_dim": 384,
  "decoder_head_type": "linear",
  "decoder_init_values": null,
  "decoder_num_heads": 6,
  "depth_decoder_depth": null,
  "depth_decoder_embed_dim": null,
  "depth_decoder_num_heads": null,
  "depth_head_type": "conv",
  "drop_path": 0.1,
  "embed_dim": 768,
  "gradient_checkpointing_config": null,
  "img_size": 504,
  "inference_steps": 12,
  "k_sampler_beta_a": 2,
  "k_sampler_beta_b": 1,
  "k_sampling": "linspace",
  "load_patch_embed_weights": false,
  "min_steps": 8,
  "mlp_ratio": 4.0,
  "num_heads": 12,
  "num_register_tokens": 4,
  "num_steps": 16,
  "patch_embed": "dinov2_vitb14_reg",
  "patch_size": 14,
  "recurrence_mode": "gated",
  "stochastic_depth": 0.3,
  "stochastic_depth_mode": "random",
  "sync_stochastic_depth": true,
  "time_conditioning": "interval"
}