{
  "model_type": "stellar",
  "library_name": "stellar",
  "code": "https://github.com/microsoft/STELLAR",
  "default": "stellar-b16",
  "models": {
    "stellar-b16": {
      "weights": "stellar-b16.safetensors",
      "category": "main",
      "backbone": "facebook/vit-mae-base",
      "image_size": 224,
      "patch_size": 16,
      "embed_dim": 768,
      "num_sparse_tokens": 16,
      "num_decoder_layers": 6,
      "spatial_temp": 0.06,
      "do_recon": true,
      "do_clustering": true,
      "do_cls": true,
      "num_clusters": 16384,
      "recon_type": "vq"
    },
    "stellar-l16": {
      "weights": "stellar-l16.safetensors",
      "category": "main",
      "backbone": "facebook/vit-mae-large",
      "image_size": 224,
      "patch_size": 16,
      "embed_dim": 1024,
      "num_sparse_tokens": 16,
      "num_decoder_layers": 6,
      "spatial_temp": 0.06,
      "do_recon": true,
      "do_clustering": true,
      "do_cls": true,
      "num_clusters": 16384,
      "recon_type": "vq"
    },
    "stellar-h16": {
      "weights": "stellar-h16.safetensors",
      "category": "main",
      "backbone": "facebook/vit-mae-huge",
      "image_size": 224,
      "patch_size": 14,
      "embed_dim": 1280,
      "num_sparse_tokens": 16,
      "num_decoder_layers": 6,
      "spatial_temp": 0.06,
      "do_recon": true,
      "do_clustering": true,
      "do_cls": true,
      "num_clusters": 16384,
      "recon_type": "vq"
    },
    "stellar-b8": {
      "weights": "stellar-b8.safetensors",
      "category": "ablation",
      "backbone": "facebook/vit-mae-base",
      "image_size": 224,
      "patch_size": 16,
      "embed_dim": 768,
      "num_sparse_tokens": 8,
      "num_decoder_layers": 6,
      "spatial_temp": 0.06,
      "do_recon": true,
      "do_clustering": true,
      "do_cls": true,
      "num_clusters": 16384,
      "recon_type": "vq"
    },
    "stellar-b24": {
      "weights": "stellar-b24.safetensors",
      "category": "ablation",
      "backbone": "facebook/vit-mae-base",
      "image_size": 224,
      "patch_size": 16,
      "embed_dim": 768,
      "num_sparse_tokens": 24,
      "num_decoder_layers": 6,
      "spatial_temp": 0.06,
      "do_recon": true,
      "do_clustering": true,
      "do_cls": true,
      "num_clusters": 16384,
      "recon_type": "vq"
    }
  }
}