Stack recommandé
Que construisez-vous ? Customise a model
Une technologie par emplacement, avec la raison de la suggestion et par quoi la remplacer.
Base model
modelFamily of decoder-only Transformer language models trained with a generative objective.
Why déduit du graphe
The agents converse through a model backend, with the hosted models among the supported ones.
Alternatives
Method
techniqueContinue training a pre-trained model on task-specific data to adapt its behaviour.
Alternatives
Efficient alternative
techniqueFine-tuning by training small low-rank adapter matrices instead of all weights.
Why déduit du graphe
QLoRA adds 4-bit quantisation of the frozen base model on top of LoRA adapters.
Serving
techniqueCached attention keys and values that make generating one token at a time affordable.
Why déduit du graphe
Reusing a shared prefix means reusing the cached keys and values for it.