XLNet
XLNet is an AI model developed by Carnegie Mellon University (CMU) and Google Brain (United States), first published in June 2019. It works in the language domain, on tasks such as language modeling/generation, question answering and sentiment classification. It counts among the frontier models: the systems trained with the most compute of their moment.
Training it took an estimated 6.2×10²¹ FLOP of compute (estimation method: hardware,operation counting). The model has 340,000,000 parameters. It was trained on roughly 32.9B datapoints. Training ran on Google TPU v3. The compute alone is estimated at $14K in 2023 dollars.
Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 9,353 citations. Epoch AI rates the confidence of this record as confident.
- Organization
- Carnegie Mellon University (CMU), Google Brain
- Country of organization
- United States
- Domain
- Language
- Task
- Language modeling/generation, Question answering, Sentiment classification
- Training compute
- 6.2×10²¹ FLOP
- Compute estimation method
- Hardware, Operation counting
- Parameters
- 340,000,000
- Dataset size
- 32.9B
- Training hardware
- Google TPU v3
- Training cost (2023 USD)
- $14K
- Numerical format
- FP32
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Citations
- 9,353
- Epoch confidence
- Confident