mBART-50

Training compute

1.5×10²² FLOP

Parameters

610M

Published

Aug 2, 2020

mBART-50 is an AI model developed by Facebook AI (United States), first published in August 2020. It works in the language domain, on tasks such as translation.

Training it took an estimated 1.5×10²² FLOP of compute (estimation method: hardware). The model has 610,000,000 parameters. Training ran on 256 NVIDIA V100 for about 420 hours.

Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 552 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Facebook AI
Country of organization: United States
Domain: Language
Task: Translation
Training compute: 1.5×10²² FLOP
Compute estimation method: Hardware
Parameters: 610,000,000
Training hardware: NVIDIA V100
Chips used: 256
Training time: 420 h
Chip-hours: 107.5K
Training power draw: 156.5 kW
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Citations: 552
Epoch confidence: Confident