Genie-SCOPe (bio)

Training compute

1.8×10²¹ FLOP

Parameters

4.1M

Published

Jan 29, 2023

Genie-SCOPe (bio) is an AI model developed by Columbia University (United States), first published in January 2023. It works in the biology domain, on tasks such as protein design.

Training it took an estimated 1.8×10²¹ FLOP of compute (estimation method: hardware). The model has 4,100,000 parameters. It was trained on roughly 1.8M datapoints. Training ran on 12 NVIDIA A100 for about 336 hours.

Access: Open weights (unrestricted). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Columbia University
Country of organization: United States
Domain: Biology
Task: Protein design
Training compute: 1.8×10²¹ FLOP
Compute estimation method: Hardware
Parameters: 4,100,000
Dataset size: 1.8M
Training hardware: NVIDIA A100
Chips used: 12
Training time: 336 h
Training power draw: 9.6 kW
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Confident