align_vision_model/AlignVisionModel/config.json

{
  "architectures": [
    "AlignVisionModel"
  ],
  "batch_norm_eps": 0.001,
  "batch_norm_momentum": 0.99,
  "depth_coefficient": 3.1,
  "depth_divisor": 8,
  "depthwise_padding": [],
  "drop_connect_rate": 0.2,
  "dtype": "float32",
  "expand_ratios": [
    1,
    6,
    6
  ],
  "hidden_act": "gelu",
  "hidden_dim": 64,
  "image_size": 600,
  "in_channels": [
    32,
    16,
    24
  ],
  "initializer_range": 0.02,
  "kernel_sizes": [
    3,
    3,
    5
  ],
  "model_type": "align_vision_model",
  "num_block_repeats": [
    1,
    1,
    2
  ],
  "num_channels": 3,
  "num_hidden_layers": 16,
  "out_channels": [
    16,
    24,
    30
  ],
  "pooling_type": "mean",
  "squeeze_expansion_ratio": 0.25,
  "strides": [
    1,
    1,
    2
  ],
  "transformers_version": "5.15.0.dev0",
  "width_coefficient": 2.0
}