tfm.nlp.layers.FourierTransformLayer

Fourier Transform layer.

tfm.nlp.layers.FourierTransformLayer(
    use_fft: bool = False, name: str = 'fourier_transform', **kwargs
)

Applies 2D Fourier Transform over final two dimensions of query inputs - typically the sequence and hidden dimensions.

Args
`use_fft`	Whether to use Fast Fourier Transform (True) or the Discrete Fourier Transform (DFT) matrix (False) to compute the Fourier Transform. See _pick_fourier_transform() for recommendations on when to use FFT or DFT.
`name`	Name for layer.
`**kwargs`	Keyword arguments.

Methods

call(
    query: tf.Tensor, value: tf.Tensor, **kwargs
) -> tf.Tensor

Applies layer to query.

Args
`query`	Batch of input embeddings, typically of shape [batch_size, max_seq_length, hidden_dim].
`value`	Unused. Included to match attention layer API.
`**kwargs`	Optional arguments to catch unused attention keyword arguments.

Returns
Real part of discrete Fourier Transform of `query` inputs with shape [batch_size, max_seq_length, hidden_dim].