lgatr.layers.dropout.GradeDropout

class lgatr.layers.dropout.GradeDropout(p=0.0)

Bases: Module

Grade-wise dropout on multivectors (and an optional scalar stream).

Parameters:

p (float) – Dropout probability.

forward(multivectors, scalars=None)

Apply dropout to multivectors and (optionally) scalars.

Parameters:

• multivectors (Tensor) – Multivector inputs of shape (..., 16).

• scalars (Tensor | None) – Optional scalar inputs of shape (..., s_channels). If None, scalar dropout is skipped and outputs_s is None.

Return type:

tuple[Tensor, Tensor | None]

Returns:

• outputs_mv – Multivectors after dropout, shape (..., 16).

• outputs_s – Scalars after dropout, shape (..., s_channels), or None if scalars is None.