{
  "act_layer": "gelu",
  "attn_drop": 0.0,
  "attn_layer": "flex",
  "block_mask": "block",
  "depth": 21,
  "drop_path": 0.0,
  "embed_dim": 832,
  "ffn_layer": "mlp",
  "ffn_ratio": 2.57,
  "head_drop": 0.0,
  "img_size": 224,
  "in_channels": 3,
  "ls_init_values": 1e-06,
  "mask_token_grans": [
    32,
    64,
    96,
    128,
    192,
    256
  ],
  "mat_dims": [
    416,
    624,
    832
  ],
  "max_latent_tokens": 256,
  "norm_layer": "layernorm",
  "num_classes": 1000,
  "num_heads": 13,
  "patch_size": 16,
  "process_token_init": "learned",
  "proj_bias": true,
  "proj_drop": 0.0,
  "qkv_bias": true,
  "rope_theta": 10000,
  "use_embed_ffn": true,
  "use_output_ffn": false
}