Optimal Subarchitecture Extraction For BERT
This provides a more efficient and faster variant of BERT for natural language understanding tasks, though it is incremental as it builds on existing neural architecture search methods.
The paper tackles the problem of reducing the size and training time of BERT-large by extracting an optimal subarchitecture called Bort, which achieves a 5.5% effective size, 1.2% pretraining time compared to RoBERTa-large, and performance improvements of up to 31% on NLU benchmarks.
We extract an optimal subset of architectural parameters for the BERT architecture from Devlin et al. (2018) by applying recent breakthroughs in algorithms for neural architecture search. This optimal subset, which we refer to as "Bort", is demonstrably smaller, having an effective (that is, not counting the embedding layer) size of $5.5\%$ the original BERT-large architecture, and $16\%$ of the net size. Bort is also able to be pretrained in $288$ GPU hours, which is $1.2\%$ of the time required to pretrain the highest-performing BERT parametric architectural variant, RoBERTa-large (Liu et al., 2019), and about $33\%$ of that of the world-record, in GPU hours, required to train BERT-large on the same hardware. It is also $7.9$x faster on a CPU, as well as being better performing than other compressed variants of the architecture, and some of the non-compressed variants: it obtains performance improvements of between $0.3\%$ and $31\%$, absolute, with respect to BERT-large, on multiple public natural language understanding (NLU) benchmarks.