Masked Autoencoders ViT-H

Training compute

4.6×10²⁰ FLOP

Parameters

632M

Published

Nov 11, 2021

Masked Autoencoders ViT-H is an AI model developed by Facebook AI Research (United States and France), first published in November 2021. It works in the vision domain, on tasks such as semantic segmentation, image classification and image generation.

Training it took an estimated 4.6×10²⁰ FLOP of compute (estimation method: hardware,operation counting). The model has 632,000,000 parameters. It was trained on roughly 328M datapoints.

Access: Open weights (non-commercial). Its weights are openly available. It is built on top of ViT-Huge/14. The reference paper has 11,449 citations. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: Facebook AI Research
Country of organization: United States, France
Domain: Vision
Task: Semantic segmentation, Image classification, Image generation
Training compute: 4.6×10²⁰ FLOP
Compute estimation method: Hardware, Operation counting
Parameters: 632,000,000
Dataset size: 328M
Training time: 69 h
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Base model: ViT-Huge/14
Citations: 11,449
Epoch confidence: Speculative