Halittu › Tsarin wucin gadi
Manyan samfuran harshe
Hanyoyin sadarwa na jijiyoyi na Transformer da aka horar don hasashen token na gaba a manyan tarin rubutu, sannan sau da yawa a gyara su da ra’ayin ɗan Adam.
Hanyoyin sadarwa na jijiyoyi na Transformer da aka horar don hasashen token na gaba a manyan tarin rubutu, sannan sau da yawa a gyara su da ra’ayin ɗan Adam. Suna amfani da self-attention a kan dukan taga mahallin. Takarda ɗaya ta 2020 ta ba da rahoton girman parameters biliyan 175. Suna samar da harshe mai santsi, don haka fitarwa mai kama da ta ɗan Adam gwaji ne mara kyau na ƙwarewar ciki.
A cikin bishiyoyi
Tushe
- Vaswani A et al. (2017). Attention is all you need. NeurIPS. arXiv:1706.03762.
- Brown TB et al. (2020). Language models are few-shot learners. NeurIPS 33. arXiv:2005.14165 (175B parameters).