config.json

{
  "architectures": [
    "Sam2HieraDetModel"
  ],
  "blocks_per_stage": [
    1,
    2,
    7,
    2
  ],
  "dtype": "float32",
  "embed_dim_per_stage": [
    12,
    24,
    48,
    96
  ],
  "global_attention_blocks": [
    5,
    7,
    9
  ],
  "hidden_act": "gelu",
  "hidden_size": 12,
  "image_size": 128,
  "initializer_range": 0.02,
  "layer_norm_eps": 1e-06,
  "mlp_ratio": 4.0,
  "model_type": "sam2_hiera_det_model",
  "num_attention_heads": 1,
  "num_attention_heads_per_stage": [
    1,
    2,
    4,
    8
  ],
  "num_channels": 3,
  "num_query_pool_stages": 3,
  "patch_kernel_size": 7,
  "patch_padding": 3,
  "patch_stride": 4,
  "query_stride": [
    2,
    2
  ],
  "transformers_version": "5.16.0.dev0",
  "window_positional_embedding_background_size": [
    7,
    7
  ],
  "window_size_per_stage": [
    8,
    4,
    14,
    7
  ]
}