Seq2Seq LSTM

Training compute

5.6×10¹⁹ FLOP

Parameters

1.9B

Published

Sep 10, 2014

Seq2Seq LSTM is an AI model developed by Google (United States), first published in September 2014. It works in the language domain, on tasks such as translation. It counts among the frontier models: the systems trained with the most compute of their moment.

Training it took an estimated 5.6×10¹⁹ FLOP of compute (estimation method: operation counting,hardware,third-party estimation). The model has 1,920,000,000 parameters. It was trained on roughly 870M datapoints.

The reference paper has 22,025 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Google
Country of organization: United States
Domain: Language
Task: Translation
Training compute: 5.6×10¹⁹ FLOP
Compute estimation method: Operation counting, Hardware, Third-party estimation
Parameters: 1,920,000,000
Dataset size: 870M
Training time: 240 h
Citations: 22,025
Epoch confidence: Confident