PointWiseFeedForward

class rectools.models.nn.transformers.net_blocks.PointWiseFeedForward(n_factors: int, n_factors_ff: int, dropout_rate: float, activation: Module, bias: bool = True)

Bases: Module

Feed-Forward network to introduce nonlinearity into the transformer model. This implementation is the one used by SASRec authors.

Parameters

• n_factors (int) – Latent embeddings size.

• n_factors_ff (int) – How many hidden units to use in the network.

• dropout_rate (float) – Probability of a hidden unit to be zeroed.

• activation (torch.nn.Module) – Activation function module.

• bias (bool, default True) – If True, add bias to linear layers.

Methods

forward(seqs)

Forward pass.

Attributes

forward(seqs: Tensor) → Tensor

Forward pass.

Parameters

seqs (torch.Tensor) – User sequences of item embeddings.

Returns

User sequence that passed through all layers.

Return type

torch.Tensor