StyleGAN3-R

Training compute

2.4×10²¹ FLOP

Parameters

1.6M

Published

Jun 21, 2021

StyleGAN3-R is an AI model developed by NVIDIA and Aalto University (United States and Finland), first published in June 2021. It works in the image generation domain, on tasks such as image generation.

Training it took an estimated 2.4×10²¹ FLOP of compute (estimation method: hardware). The model has 1,580,000 parameters. It was trained on roughly 50M datapoints. Training ran on 8 NVIDIA V100 for about 2.2K hours.

Access: Open weights (non-commercial). Its weights are openly available. The reference paper has 2,031 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: NVIDIA, Aalto University
Country of organization: United States, Finland
Domain: Image generation
Task: Image generation
Training compute: 2.4×10²¹ FLOP
Compute estimation method: Hardware
Parameters: 1,580,000
Dataset size: 50M
Training hardware: NVIDIA V100
Chips used: 8
Training time: 2,248 h
Training power draw: 4.9 kW
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Citations: 2,031
Epoch confidence: Confident