What is Parameter-Efficient Fine-Tuning (PEFT)? Definition | AI Understanding

Ma'anarsa

Hanyoyin da ke daidaita samfura ta hanyar horar da ƙaramin juzu'i na ƙarin sigogi.

Sharuɗɗa masu alaƙa

Kyakkyawan-Tuning

Ci gaba da horarwa akan ƙayyadaddun bayanai na yanki don daidaita samfurin da aka riga aka horar zuwa takamaiman aiki.

LoRA (Ƙarancin Matsayi)

Hanyar daidaitawa mai inganci mai inganci wacce ke ƙara ƙananan matrix adaftar.

Matakan horarwa da aka yi amfani da su bayan horo na farko, kamar kunna koyarwa, inganta fifiko, da daidaita aminci.

Tunatarwa Umarni

Kyakkyawan daidaita samfurin akan nau'i-nau'i na amsa umarni-don inganta ɗawainiya mai biyo baya.

Hyperparameter na horarwa yana sarrafa nawa sigogi ke canza kowane mataki na sabuntawa.

DPO (Haɓaka fifiko kai tsaye)

Hanyar horarwa wacce ke daidaita ƙira kai tsaye akan nau'ikan zaɓin zaɓi ba tare da buƙatar samfurin lada daban ba.

Ƙara koyo a cikin jagororinmu na kyauta

Fine-Tuning Hard Parameter Sharing in Multi-Task Networks Gumbel-Softmax and Reparameterization Rejection Sampling Fine-Tuning

Duba kuma

Parameter Perplexity Overfitting Pipeline Open-Source Model Precision OCR (Optical Character Recognition)Pretraining

Nemo cikakken AI ƙamus Bincika duk jagororin AI kyauta