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

Incazelo

Izindlela ezijwayelanisa amamodeli ngokuqeqesha isethi encane yamapharamitha angeziwe.

Imigomo ehlobene

Ukuqhubeka nokuqeqeshwa kudatha eqondene nesizinda ukuze kulungiswe imodeli eqeqeshwe ngaphambilini emsebenzini othile.

I-LoRA (Ukujwayela Kwezinga Eliphansi)

Indlela yokushuna esebenza kahle ngepharamitha eyengeza amatrices e-adaptha esezingeni eliphansi.

Ngemva kokuqeqeshwa

Izinyathelo zokuqeqesha zisetshenziswa ngemva kokuqeqeshwa kusengaphambili, njengokushuna iziyalezo, ukwenza kahle okuthandwayo, nokushuna kokuphepha.

I-Instruction Tuning

Ukulungisa kahle imodeli kumapheya emiyalelo nezimpendulo ukuze uthuthukise umsebenzi olandelayo.

Izinga lokufunda

I-hyperparameter yokuqeqesha elawula ukuthi mangakanani amapharamitha ashintsha isinyathelo ngasinye sokubuyekeza.

I-DPO (Direct Preference Optimization)

Indlela yokuqeqesha eshuna kahle imodeli ngokuqondile kumapheya athandwayo ngaphandle kokudinga imodeli yomklomelo ehlukile.

Funda kabanzi kumhlahlandlela wethu wamahhala

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

Bhekafuthi

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

Phequlula Uhla lwamagama olugcwele lwe-AI Hlola yonke imihlahlandlela ye-AI yamahhala