MOSS-Moon-003

MOSS-Moon-003 ni modeli ya lugha iliyoundwa na Fudan University (China), iliyotolewa 19 Aprili 2023.

Uzito wake (weights) uko wazi: mtu yeyote anaweza kuipakua, kuiendesha na kuitumia hata kibiashara.

Ina vigezo (parameters) bilioni 16.

Naweza kuiendesha mwenyewe?

Ndiyo, kwenye kompyuta yenye uwezo mkubwa au kifaa chenye uwezo mkubwa zaidi.

  1. SimuRAM ya GB 6–8
  2. Kompyuta mpakatoRAM ya GB 16
  3. Kompyuta yenye uwezo mkubwaGB 24 hadi 64 (kadi ya michoro au Mac yenye uwezo mkubwa)
  4. Sevazaidi ya GB 64

Inahitaji takriban GB 10.6 za kumbukumbu (RAM) ikibanwa hadi biti 4.

Inahitaji kadi ya michoro (GPU) yenye kumbukumbu kubwa au kompyuta ya Mac yenye kumbukumbu ya GB 32–64.

Modeli nyingine za Fudan University

  1. InternVL2_5-26BChinese University of Hong Kong (CUHK), Fudan University, Nanjing University, SenseTime, Shanghai AI Lab, Shanghai Jiao Tong University na Tsinghua UniversityDes 2024Data mseto, Lugha
    Wazi25.5BKompyuta yenye uwezo mkubwa
  2. InternVL2_5-38BChinese University of Hong Kong (CUHK), Fudan University, Nanjing University, SenseTime, Shanghai AI Lab, Shanghai Jiao Tong University na Tsinghua UniversityDes 2024Data mseto, Lugha
    Wazi38.4BKompyuta yenye uwezo mkubwa
  3. InternVL2_5-78BChinese University of Hong Kong (CUHK), Fudan University, Nanjing University, SenseTime, Shanghai AI Lab, Shanghai Jiao Tong University na Tsinghua UniversityDes 2024Data mseto, Lugha
    Wazi, kwa masharti78.4BSeva
  4. ProtRNAFudan University na Shanghai AI LabSep 2024Biolojia
    Wazi650M
  5. InternVL1.5Chinese University of Hong Kong (CUHK), Fudan University, Nanjing University, SenseTime, Shanghai AI Lab na Tsinghua UniversityApr 2024Data mseto, Utambuzi wa picha
    Wazi25.5BKompyuta yenye uwezo mkubwa
  6. GroundingGPTByteDance na Fudan UniversityMar 2024Data mseto, Lugha
    Wazi7BKompyuta mpakato