Very Deep VAEs (ImageNet-64)

Training compute

1.8×10²¹ FLOP

Parameters

125M

Published

Mar 16, 2021

Very Deep VAEs (ImageNet-64) is an AI model developed by OpenAI (United States), first published in March 2021. It works in the image generation domain, on tasks such as image generation and image representation.

Training it took an estimated 1.8×10²¹ FLOP of compute (estimation method: hardware). The model has 125,000,000 parameters. Training ran on 32 NVIDIA V100 for about 420 hours.

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

Full record

Organization: OpenAI
Country of organization: United States
Domain: Image generation
Task: Image generation, Image representation
Training compute: 1.8×10²¹ FLOP
Compute estimation method: Hardware
Parameters: 125,000,000
Training hardware: NVIDIA V100
Chips used: 32
Training time: 420 h
Training power draw: 19.5 kW
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Confident