Package io.github.kirstenali.deepj.models

Class TextGenerator

java.lang.Object

io.github.kirstenali.deepj.models.TextGenerator

public final class TextGenerator
extends Object

Autoregressive text generation for any decoder-only transformer model. Supports temperature sampling and optional top-k filtering.

Method Summary

Modifier and Type	Method	Description
static String	generate(DeepSeekModel model, Tokenizer tok, TransformerConfig cfg, String prompt, int maxNewTokens, float temperature, int topK, long seed)
static String	generate(GPTModel model, Tokenizer tok, GPTConfig cfg, String prompt, int maxNewTokens, float temperature, int topK, long seed)
static String	generate(LlamaModel model, Tokenizer tok, TransformerConfig cfg, String prompt, int maxNewTokens, float temperature, int topK, long seed)
static String	generate(Function<int[],Tensor> forwarder, int maxSeqLen, Tokenizer tok, String prompt, int maxNewTokens, float temperature, int topK, long seed)	Generate text with an explicit maxSeqLen — useful when no config is available.
static String	generate(Function<int[],Tensor> forwarder, TransformerConfig cfg, Tokenizer tok, String prompt, int maxNewTokens, float temperature, int topK, long seed)	Generate text using any model that maps int[] ids → [seqLen × vocabSize] logits.

Methods inherited from class java.lang.Object

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

Method Details

generate

public static String generate(GPTModel model,
 Tokenizer tok,
 GPTConfig cfg,
 String prompt,
 int maxNewTokens,
 float temperature,
 int topK,
 long seed)

generate

public static String generate(LlamaModel model,
 Tokenizer tok,
 TransformerConfig cfg,
 String prompt,
 int maxNewTokens,
 float temperature,
 int topK,
 long seed)

generate

public static String generate(DeepSeekModel model,
 Tokenizer tok,
 TransformerConfig cfg,
 String prompt,
 int maxNewTokens,
 float temperature,
 int topK,
 long seed)

generate

public static String generate(Function<int[],Tensor> forwarder,
 TransformerConfig cfg,
 Tokenizer tok,
 String prompt,
 int maxNewTokens,
 float temperature,
 int topK,
 long seed)

Generate text using any model that maps int[] ids → [seqLen × vocabSize] logits.

generate

public static String generate(Function<int[],Tensor> forwarder,
 int maxSeqLen,
 Tokenizer tok,
 String prompt,
 int maxNewTokens,
 float temperature,
 int topK,
 long seed)

Generate text with an explicit maxSeqLen — useful when no config is available.