Vega v2
Vega v2 is an AI model developed by Wuhan University, JD Explore Academy, Shanghai AI Lab, Nanyang Technological University, Washington University in St Louis, Chongqing University of Posts and Telecommunications and University of Sydney (China, Singapore, United States and Australia), first published in December 2022. It works in the language domain, on tasks such as language modeling, question answering and word sense disambiguation.
Training it took an estimated 7.8×10²² FLOP of compute. The model has 6,000,000,000 parameters. It was trained on roughly 6.4B datapoints. Training ran on 320 NVIDIA A100 for about 720 hours.
Access: Unreleased. Its weights are not openly released. The reference paper has 42 citations. Epoch AI rates the confidence of this record as confident.
- Organization
- Wuhan University, JD Explore Academy, Shanghai AI Lab, Nanyang Technological University, Washington University in St Louis, Chongqing University of Posts and Telecommunications, University of Sydney
- Country of organization
- China, Singapore, United States, Australia
- Domain
- Language
- Task
- Language modeling, Question answering, Word sense disambiguation
- Training compute
- 7.8×10²² FLOP
- Parameters
- 6,000,000,000
- Dataset size
- 6.4B
- Training hardware
- NVIDIA A100
- Chips used
- 320
- Training time
- 720 h
- Training power draw
- 255.9 kW
- Model accessibility
- Unreleased
- Open weights
- No
- Citations
- 42
- Epoch confidence
- Confident