BERT-Large-CAS (PTB+WT2+WT103)

Training compute

1.5×10²⁰ FLOP

Parameters

395M

Published

Apr 20, 2019

BERT-Large-CAS (PTB+WT2+WT103) 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/generation.

Training it took an estimated 1.5×10²⁰ FLOP of compute (estimation method: operation counting). The model has 395,000,000 parameters. It was trained on roughly 1.3B datapoints.

Access: Unreleased. Its weights are not openly released. The reference paper has 139 citations. 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/generation
Training compute: 1.5×10²⁰ FLOP
Compute estimation method: Operation counting
Parameters: 395,000,000
Dataset size: 1.3B
Numerical format: FP32
Model accessibility: Unreleased
Open weights: No
Citations: 139
Epoch confidence: Likely