AWD-FWM (WT2)

Training compute

7.1×10¹⁷ FLOP

Parameters

37M

Published

Nov 16, 2020

AWD-FWM (WT2) is an AI model developed by IDSIA and Microsoft Research (Switzerland and United States), first published in November 2020. It works in the language domain, on tasks such as language modeling.

Training it took an estimated 7.1×10¹⁷ FLOP of compute (estimation method: operation counting). The model has 37,000,000 parameters. It was trained on roughly 2M datapoints.

Access: Unreleased. Its weights are not openly released. The reference paper has 64 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: IDSIA, Microsoft Research
Country of organization: Switzerland, United States
Domain: Language
Task: Language modeling
Training compute: 7.1×10¹⁷ FLOP
Compute estimation method: Operation counting
Parameters: 37,000,000
Dataset size: 2M
Model accessibility: Unreleased
Open weights: No
Citations: 64
Epoch confidence: Confident