CbhgVariant
Enums · AiDotNet.NeuralNetworks.Layers
Which reference implementation’s CBHG conventions a CbhgLayer follows.
Fields
| Field | Summary |
|---|---|
CbhgVariant ForwardTacotron |
ForwardTacotron (as-ideas/ForwardTacotron common_layers.CBHG): PyTorch padding of k / 2 with the sequence truncated to its length, max pooling over the previous frame, dropout after the pooling and the first projection, and an always-pres… |
CbhgVariant Tacotron |
Tacotron (keithito/tacotron): TensorFlow “same” padding, max pooling over the next frame, a highway input projection only when the widths differ, no dropout. |
int value__ |