YOLOX-X

Training compute

6.3×10²⁰ FLOP

Parameters

99.1M

Published

Aug 6, 2021

YOLOX-X is an AI model developed by Megvii Inc (China), first published in August 2021. It works in the vision domain, on tasks such as object detection.

Training it took an estimated 6.3×10²⁰ FLOP of compute (estimation method: operation counting). The model has 99,100,000 parameters. It was trained on roughly 2.5M datapoints. Training ran on 8 NVIDIA V100.

Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 5,755 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: Megvii Inc
Country of organization: China
Domain: Vision
Task: Object detection
Training compute: 6.3×10²⁰ FLOP
Compute estimation method: Operation counting
Parameters: 99,100,000
Dataset size: 2.5M
Training hardware: NVIDIA V100
Chips used: 8
Training power draw: 4.8 kW
Numerical format: FP16
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Citations: 5,755
Epoch confidence: Likely