WeNet (PTB)

Training compute

3.5×10¹⁸ FLOP

Parameters

23M

Published

Apr 8, 2019

WeNet (PTB) is an AI model developed by Amazon (United States), first published in April 2019. It works in the language domain, on tasks such as neural architecture search - nas and language modeling.

Training it took an estimated 3.5×10¹⁸ FLOP of compute (estimation method: hardware,operation counting). The model has 23,000,000 parameters. Training ran on 1 NVIDIA V100 for about 120 hours.

Access: Unreleased. Its weights are not openly released. Epoch AI rates the confidence of this record as likely.

Full record

Organization: Amazon
Country of organization: United States
Domain: Language
Task: Neural Architecture Search - NAS, Language modeling
Training compute: 3.5×10¹⁸ FLOP
Compute estimation method: Hardware, Operation counting
Parameters: 23,000,000
Training hardware: NVIDIA V100
Chips used: 1
Training time: 120 h
Training power draw: 339 W
Model accessibility: Unreleased
Open weights: No
Epoch confidence: Likely