config.json

{
  "model_type": "vit_b",
  "sfg_ckpt_path": null,
  "sfg_filter_num": 1,
  "sfg_filter_type": "conv2d",
  "sfg_intermediate_channels": 32,
  "x_channel_num": 1,
  "x_data_field": "depth_images",
  "x_encoder_ckpt_path": null,
  "x_lora_rank": 4,
  "x_norm_type": "mean-std"
}