Accelerating NLP Tasks on FPGA with Compressed BERT and a Hardware-Oriented Early Exit Method.

Binjing Li Siyuan Lu Keli Xie Zhongfeng Wang

Published in: ISVLSI (2022)

Keyphrases

computational cost
preprocessing
similarity measure
high accuracy
significant improvement
pairwise
natural language
dynamic programming
high precision
evolutionary algorithm
cost function
probabilistic model
natural language processing
real time
low cost
wordnet
detection method
data acquisition
data compression