UniRep

Training compute

2.2×10¹⁹ FLOP

Parameters

18.2M

Published

Mar 26, 2019

UniRep is an AI model developed by Harvard University (United States), first published in March 2019. It works in the biology domain, on tasks such as proteins and protein or nucleotide language model (plm/nlm).

Training it took an estimated 2.2×10¹⁹ FLOP of compute (estimation method: hardware). The model has 18,200,000 parameters. Training ran on 4 NVIDIA Tesla K80 for about 588 hours.

Access: Open weights (non-commercial). Its weights are openly available. The reference paper has 989 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: Harvard University
Country of organization: United States
Domain: Biology
Task: Proteins, Protein or nucleotide language model (pLM/nLM)
Training compute: 2.2×10¹⁹ FLOP
Compute estimation method: Hardware
Parameters: 18,200,000
Training hardware: NVIDIA Tesla K80
Chips used: 4
Training time: 588 h
Chip-hours: 2.4K
Training power draw: 2.5 kW
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Citations: 989
Epoch confidence: Likely