Package io.github.kirstenali.deepj.nn.transformer

Interface CausalLM

All Superinterfaces:

Trainable

All Known Implementing Classes:

DecoderOnlyModel, DeepJLoop, DeepJNova, DeepJOrbit, DeepJOrigin, DeepJPrism

public interface CausalLM
extends Trainable

Method Summary

Modifier and Type	Method	Description
Tensor	forward(int[] inputIds)
default Tensor	forwardBatch(int[][] inputIds)
float	gradClipNorm()
default GenerationSession	openGeneration()
default GenerationSession	openGeneration(ContextWindowMode mode)
default boolean	supportsBatchedForward()

Methods inherited from interface io.github.kirstenali.deepj.training.Trainable

parameters, zeroGrad

Method Details

forward

Tensor forward(int[] inputIds)

supportsBatchedForward

default boolean supportsBatchedForward()

forwardBatch

default Tensor forwardBatch(int[][] inputIds)

openGeneration

default GenerationSession openGeneration()

openGeneration

default GenerationSession openGeneration(ContextWindowMode mode)

gradClipNorm

float gradClipNorm()