paul-krug / pytorch-tcn

(Realtime) Temporal Convolutions in PyTorch
MIT License
55 stars 8 forks source link

Add embeddings #4

Closed paul-krug closed 6 months ago

paul-krug commented 6 months ago

add the option to condition every temporal block with an arbitrary number of embedding tensors as additional inputs to the TCN forward function. Global and local conditioning are possible. Different options how to feed the embeddings into the network are available. Updated unit tests accordingly.