REVIEW 2 cited by
Speeding up Deep Model Training by Sharing Weights and Then Unsharing
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
We propose a simple and efficient approach for training the BERT model. Our approach exploits the special structure of BERT that contains a stack of repeated modules (i.e., transformer encoders). Our proposed approach first trains BERT with the weights shared across all the repeated modules till some point. This is for learning the commonly shared component of weights across all repeated layers. We then stop weight sharing and continue training until convergence. We present theoretic insights for training by sharing weights then unsharing with analysis for simplified models. Empirical experiments on the BERT model show that our method yields better performance of trained models, and significantly reduces the number of training iterations.
Forward citations
Cited by 2 Pith papers
-
Parallel Scaling Law for Language Models
Running P parallel streams of a language model with learned prefixes gives performance comparable to multiplying parameters by about k log P + 1, with k around 0.33 to 0.39.
-
A multilevel approach to accelerate the training of Transformers
A multilevel scheme that alternates fine transformer training with two half-depth coarse models reaches the single-level training loss with 44 percent fewer FLOPs on one small language-model setup.
Discussion (0). Continue with ORCID to comment.