Papers › Layer-wise Pruning of Transformer Attention Heads for Efficient Language Modeling

Layer-wise Pruning of Transformer Attention Heads for Efficient Language Modeling

7 Oct 20212021 18th International SoC Design Conference (ISOCC) 2021 11arXiv:2110.03252archive 2025-07-28

Kyuhong Shim, Iksoo Choi, Wonyong Sung, Jungwook Choi

While Transformer-based models have shown impressive language modeling performance, the large computation cost is often prohibitive for practical use. Attention head pruning, which removes unnecessary attention heads in the multihead attention, is a promising technique to solve this problem. However, it does not evenly reduce the overall load because the heavy feedforward module is not affected by head pruning. In this paper, we apply layer-wise attention head pruning on All-attention Transformer so that the entire computation and the number of parameters can be reduced proportionally to the number of pruned heads. While the architecture has the potential to fully utilize head pruning, we propose three training methods that are especially helpful to minimize performance degradation and stabilize the pruning process. Our pruned model shows consistently lower perplexity within a comparable parameter size than Transformer-XL on WikiText-103 language modeling benchmark.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

aiha-lab/Attention-Head-Pruning officialmentioned on GitHubpytorchGPL-3.0 report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Language ModelingLanguage Modelling

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

Absolute Position EncodingsAdamAdaptive Input RepresentationsAdaptive SoftmaxAttentionBPECosine AnnealingDense ConnectionsDropoutLabel SmoothingLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionPosition-Wise Feed-Forward LayerPruningReLUResidual ConnectionSoftmaxTransformerTransformer-XLVariational Dropout

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections