MetNet

Training compute

9.5×10¹⁸ FLOP

Parameters

225M

Published

Mar 24, 2020

MetNet is an AI model developed by Google (United States), first published in March 2020. It works in the earth science domain, on tasks such as weather forecasting.

Training it took an estimated 9.5×10¹⁸ FLOP of compute (estimation method: operation counting). The model has 225,000,000 parameters. It was trained on roughly 7B datapoints. Training ran on 256 Google TPU v3.

Access: Unreleased. Its weights are not openly released. The reference paper has 345 citations. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: Google
Country of organization: United States
Domain: Earth science
Task: Weather forecasting
Training compute: 9.5×10¹⁸ FLOP
Compute estimation method: Operation counting
Parameters: 225,000,000
Dataset size: 7B
Training hardware: Google TPU v3
Chips used: 256
Training power draw: 235.4 kW
Model accessibility: Unreleased
Open weights: No
Citations: 345
Epoch confidence: Speculative