Ir al contenido
Menú
Español

Stack recomendado

¿Qué estás construyendo? Customise a model

Una tecnología por posición, con el motivo de la sugerencia y qué usar en su lugar.

Base model

model

GPT

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

Why derivado del grafo

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

Fuente

Method

technique

Fine-tuning

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

Why derivado del grafo

LoRA used for Fine-tuning

Fuente

Efficient alternative

technique

LoRA

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

Why derivado del grafo

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

Fuente

Serving

technique

KV Cache

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

Why derivado del grafo

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

Fuente