{
  "architectures": [
    "Sam3TrackerModel"
  ],
  "dtype": "float32",
  "image_size": 224,
  "initializer_range": 0.02,
  "mask_decoder_config": {
    "attention_downsample_rate": 2,
    "dynamic_multimask_stability_delta": 0.05,
    "dynamic_multimask_stability_thresh": 0.98,
    "dynamic_multimask_via_stability": true,
    "hidden_act": "relu",
    "hidden_size": 32,
    "iou_head_depth": 3,
    "iou_head_hidden_dim": 32,
    "mlp_dim": 64,
    "model_type": "",
    "num_attention_heads": 4,
    "num_hidden_layers": 2,
    "num_multimask_outputs": 3
  },
  "mask_downsampler_embed_dim": 32,
  "memory_attention_feed_forward_hidden_size": 32,
  "memory_attention_hidden_size": 32,
  "memory_attention_num_layers": 1,
  "memory_encoder_hidden_size": 32,
  "memory_fuser_embed_dim": 32,
  "model_type": "sam3_tracker",
  "prompt_encoder_config": {
    "hidden_act": "gelu",
    "hidden_size": 32,
    "image_size": 128,
    "layer_norm_eps": 1e-06,
    "mask_input_channels": 8,
    "model_type": "",
    "num_point_embeddings": 4,
    "patch_size": 16,
    "scale": 1
  },
  "transformers_version": "5.16.0.dev0",
  "vision_config": {
    "backbone_config": {
      "_name_or_path": "",
      "architectures": null,
      "attention_dropout": 0.0,
      "chunk_size_feed_forward": 0,
      "dtype": null,
      "global_attn_indexes": [
        0,
        1
      ],
      "hidden_act": "gelu",
      "hidden_dropout": 0.0,
      "hidden_size": 32,
      "id2label": {
        "0": "LABEL_0",
        "1": "LABEL_1"
      },
      "image_size": 224,
      "initializer_range": 0.02,
      "intermediate_size": 64,
      "is_encoder_decoder": false,
      "label2id": {
        "LABEL_0": 0,
        "LABEL_1": 1
      },
      "layer_norm_eps": 1e-06,
      "layer_scale_init_value": null,
      "model_type": "sam3_vit_model",
      "num_attention_heads": 4,
      "num_channels": 3,
      "num_hidden_layers": 2,
      "output_attentions": false,
      "output_hidden_states": false,
      "patch_size": 14,
      "pretrain_image_size": 336,
      "problem_type": null,
      "return_dict": true,
      "rope_theta": 10000.0,
      "window_size": 8
    },
    "backbone_feature_sizes": [
      [
        32,
        32
      ],
      [
        16,
        16
      ],
      [
        8,
        8
      ]
    ],
    "fpn_hidden_size": 32,
    "hidden_act": "gelu",
    "initializer_range": 0.02,
    "layer_norm_eps": 1e-06,
    "model_type": "sam3_vision_model",
    "scale_factors": [
      2.0,
      1.0,
      0.5
    ]
  }
}