Documentation
¶
Overview ¶
Package slstm implements a Sentence-State LSTM graph neural network.
Reference: "Sentence-State LSTM for Text Representation" by Zhang et al, 2018. (https://arxiv.org/pdf/1805.02474.pdf)
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type HyperLinear3 ¶
type HyperLinear3 struct { W nn.Param `spago:"type:weights"` U nn.Param `spago:"type:weights"` B nn.Param `spago:"type:biases"` }
HyperLinear3 groups multiple params for an affine transformation.
type HyperLinear4 ¶
type HyperLinear4 struct { W nn.Param `spago:"type:weights"` U nn.Param `spago:"type:weights"` V nn.Param `spago:"type:weights"` B nn.Param `spago:"type:biases"` }
HyperLinear4 groups multiple params for an affine transformation.
type Model ¶
type Model struct { nn.BaseModel Config Config InputGate *HyperLinear4 `spago:"type:params"` LeftCellGate *HyperLinear4 `spago:"type:params"` RightCellGate *HyperLinear4 `spago:"type:params"` CellGate *HyperLinear4 `spago:"type:params"` SentCellGate *HyperLinear4 `spago:"type:params"` OutputGate *HyperLinear4 `spago:"type:params"` InputActivation *HyperLinear4 `spago:"type:params"` NonLocalSentCellGate *HyperLinear3 `spago:"type:params"` NonLocalInputGate *HyperLinear3 `spago:"type:params"` NonLocalSentOutputGate *HyperLinear3 `spago:"type:params"` StartH nn.Param `spago:"type:weights"` EndH nn.Param `spago:"type:weights"` InitValue nn.Param `spago:"type:weights"` Support Support `spago:"scope:processor"` }
Model contains the serializable parameters.
type Support ¶ added in v0.2.0
type Support struct { // Shared among all nodes at the same step ViPrevG ag.Node VlPrevG ag.Node VrPrevG ag.Node VfPrevG ag.Node VsPrevG ag.Node VoPrevG ag.Node VuPrevG ag.Node // contains filtered or unexported fields }
Support contains nodes used during the forward step.
Click to show internal directories.
Click to hide internal directories.