Package io.github.kirstenali.deepj.transformer

Class LlamaTransformerBuilder

java.lang.Object

io.github.kirstenali.deepj.transformer.LlamaTransformerBuilder

public final class LlamaTransformerBuilder
extends Object

Builder for Llama-style transformer stacks.

Assembles a TransformerStack of LlamaTransformerBlocks (RMSNorm + RoPE attention + SwiGLU).

See Also:

• GPTTransformerBuilder
• DeepSeekTransformerBuilder

Constructor Summary

Constructors

Constructor	Description
LlamaTransformerBuilder()

Method Summary

Modifier and Type	Method	Description
TransformerStack	build()
LlamaTransformerBuilder	dFF(int dFF)
LlamaTransformerBuilder	dModel(int dModel)
LlamaTransformerBuilder	maxSeqLen(int maxSeqLen)	Maximum sequence length for the RoPE table.
LlamaTransformerBuilder	nHeads(int nHeads)
LlamaTransformerBuilder	nLayers(int nLayers)
LlamaTransformerBuilder	random(Random rnd)
LlamaTransformerBuilder	seed(long seed)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

LlamaTransformerBuilder

public LlamaTransformerBuilder()

Method Details

dModel

public LlamaTransformerBuilder dModel(int dModel)

nHeads

public LlamaTransformerBuilder nHeads(int nHeads)

dFF

public LlamaTransformerBuilder dFF(int dFF)

nLayers

public LlamaTransformerBuilder nLayers(int nLayers)

maxSeqLen

public LlamaTransformerBuilder maxSeqLen(int maxSeqLen)

Maximum sequence length for the RoPE table. Required.

seed

public LlamaTransformerBuilder seed(long seed)

random

public LlamaTransformerBuilder random(Random rnd)

build

public TransformerStack build()