ProtBERT-BFD
ProtBERT-BFD is an AI model developed by Technical University of Munich, NVIDIA, Seoul National University, Google, Oak Ridge National Laboratory and Med AI Technology (Germany, United States, South Korea and China), first published in May 2021. It works in the biology domain, on tasks such as proteins and protein or nucleotide language model (plm/nlm).
Training it took an estimated 3.9×10²² FLOP of compute (estimation method: operation counting). The model has 420,000,000 parameters. It was trained on roughly 59B datapoints. Training ran on 1,024 Google TPU v3. The compute alone is estimated at $45K in 2023 dollars.
Access: Open weights (unrestricted). Its weights are openly available. Epoch AI rates the confidence of this record as confident.
- Organization
- Technical University of Munich, NVIDIA, Seoul National University, Google, Oak Ridge National Laboratory, Med AI Technology
- Country of organization
- Germany, United States, South Korea, China
- Domain
- Biology
- Task
- Proteins, Protein or nucleotide language model (pLM/nLM)
- Training compute
- 3.9×10²² FLOP
- Compute estimation method
- Operation counting
- Parameters
- 420,000,000
- Dataset size
- 59B
- Training hardware
- Google TPU v3
- Chips used
- 1,024
- Training power draw
- 933.1 kW
- Training cost (2023 USD)
- $45K
- Numerical format
- FP32
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Epoch confidence
- Confident