What is Post-training? Definition | AI Understanding

Incazelo

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

Imigomo ehlobene

I-DPO (Direct Preference Optimization)

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

Ukuqeqeshwa okuphakathi

Isigaba sokuqeqesha esimaphakathi phakathi kokuqeqeshwa kwangaphambili kanye nokuqeqeshwa kwangemuva, okuvame ukusetshenziselwa amandla noma ukulungiswa kwesizinda.

I-Parameter-Efficient Fine-Tuning (PEFT)

Izindlela ezijwayelanisa amamodeli ngokuqeqesha isethi encane yamapharamitha angeziwe.

Ukuqeqeshwa kokuqala kwemodeli yezinga elikhulu kudatha ebanzi ngaphambi kokujwayela umfula.

Isethi Yokuhlola

Idathasethi ebanjiwe esetshenziselwa ukukala ikhwalithi yemodeli ngemva kokuqeqeshwa.

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

Funda kabanzi kumhlahlandlela wethu wamahhala

AI Training Pseudo-Labeling and Self-Training ELECTRA Pretraining RoBERTa Training Recipe

Bhekafuthi

Self-Play Watermarking RAG-Fusion Citations Multi-Query Retrieval Factuality Parent-Document Retrieval GSM8K

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