Skip to main content

Flatten

Flatten

Flattens dimensions from start_dim to end_dim (inclusive) into a single dimension. Commonly used to transition from convolutional layers to fully-connected layers.

Parameters:

  • start_dim: First dimension to flatten (default: 1, preserving batch)
  • end_dim: Last dimension to flatten (default: -1, all remaining dims)

Shape Contract:

  • Input: [*shape_in] arbitrary input shape
  • Output: [*shape_out] flattened output (depends on start_dim and end_dim)

Notes:

  • Default behavior: Flatten(1, -1) preserves batch dimension
  • Example: [32, 3, 224, 224] with default params gives [32, 150528]
  • No learnable parameters (pure reshaping operation)

Signature​

neuron Flatten(start_dim=Int(1), end_dim=BinOp { op: Sub, left: Int(0), right: Int(1) })

Ports​

Inputs:

  • default: [*shape_in]

Outputs:

  • default: [*shape_out]

Implementation​

"from core import structural/Flatten"
Source { source: "core", path: "structural/Flatten" }