eDiff-I

Training compute

5.5×10¹⁹ FLOP

Parameters

9.1B

Published

Nov 2, 2022

eDiff-I is an AI model developed by NVIDIA (United States), first published in November 2022. It works in the image generation domain, on tasks such as image generation and text-to-image.

Training it took an estimated 5.5×10¹⁹ FLOP of compute (estimation method: operation counting). The model has 9,100,000,000 parameters. It was trained on roughly 1.6T datapoints. Training ran on NVIDIA A100.

Access: Unreleased. Its weights are not openly released. The reference paper has 1,046 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: NVIDIA
Country of organization: United States
Domain: Image generation
Task: Image generation, Text-to-image
Training compute: 5.5×10¹⁹ FLOP
Compute estimation method: Operation counting
Parameters: 9,100,000,000
Dataset size: 1.6T
Training hardware: NVIDIA A100
Model accessibility: Unreleased
Open weights: No
Citations: 1,046
Epoch confidence: Likely