Class TransformerBuilder
java.lang.Object
io.github.kirstenali.deepj.transformer.TransformerBuilder
Convenience builder for assembling transformer stacks.
deepj is transformer-oriented: this builder replaces generic network builders.
-
Constructor Summary
Constructors -
Method Summary
Modifier and TypeMethodDescriptionbuild()dFF(int dFF) dModel(int dModel) ffnActivation(Supplier<ActivationFunction> activationFactory) Activation used inside the FFN.nHeads(int nHeads) nLayers(int nLayers) seed(long seed)
-
Constructor Details
-
TransformerBuilder
public TransformerBuilder()
-
-
Method Details
-
dModel
-
nHeads
-
dFF
-
nLayers
-
ffnActivation
Activation used inside the FFN. Default: GELU. -
seed
-
random
-
build
-