config.json

{
  "image_encoder_params": {
    "depth": 12,
    "embed_dim": 768,
    "global_attn_indexes": [
      2,
      5,
      8,
      11
    ],
    "img_size": 256,
    "mlp_ratio": 4,
    "num_heads": 12,
    "out_chans": 256,
    "patch_size": 8,
    "qkv_bias": true,
    "use_rel_pos": true,
    "window_size": 14
  },
  "mask_decoder_params": {
    "iou_head_depth": 3,
    "iou_head_hidden_dim": 256,
    "num_multimask_outputs": 3,
    "transformer_dim": 256
  },
  "prompt_encoder_params": {
    "embed_dim": 256,
    "image_embedding_size": [
      32,
      32
    ],
    "input_image_size": [
      256,
      256
    ],
    "mask_in_chans": 16
  },
  "transformer_params": {
    "depth": 2,
    "embedding_dim": 256,
    "mlp_dim": 2048,
    "num_heads": 8
  }
}