Ir para o conteúdo
Menu
Português

Stack recomendado

O que está a construir? Customise a model

Uma tecnologia por posição, com o motivo da sugestão e o que usar em alternativa.

Base model

model

GPT

Family of decoder-only Transformer language models trained with a generative objective.

Why derivado do grafo

The agents converse through a model backend, with the hosted models among the supported ones.

Fonte

Method

technique

Fine-tuning

Continue training a pre-trained model on task-specific data to adapt its behaviour.

Why derivado do grafo

LoRA used for Fine-tuning

Fonte

Efficient alternative

technique

LoRA

Fine-tuning by training small low-rank adapter matrices instead of all weights.

Why derivado do grafo

QLoRA adds 4-bit quantisation of the frozen base model on top of LoRA adapters.

Fonte

Serving

technique

KV Cache

Cached attention keys and values that make generating one token at a time affordable.

Why derivado do grafo

Reusing a shared prefix means reusing the cached keys and values for it.

Fonte