# Exploiting Student Parallelism for Efficient GPU Inference of BERT-like Models in Online Services

arXiv: https://arxiv.org/abs/2408.12526 · HTML: https://mlsanity.com/paper/2408.12526/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 33 of its 61 references that are in the map.
- First posted on arXiv 22 Aug 2024.

## Builds on (most cited first, 30 of 33)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://mlsanity.com/paper/2006.03654/) (2020)
- [Big Bird: Transformers for Longer Sequences](https://mlsanity.com/paper/2007.14062/) (2020)
- [Are Sixteen Heads Really Better than One?](https://mlsanity.com/paper/1905.10650/) (2019)
- [Once-for-All: Train One Network and Specialize it for Efficient Deployment](https://mlsanity.com/paper/1908.09791/) (2019)
- [Patient Knowledge Distillation for BERT Model Compression](https://mlsanity.com/paper/1908.09355/) (2019)
- [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://mlsanity.com/paper/2004.02984/) (2020)
- [Q-BERT: Hessian Based Ultra Low Precision Quantization of BERT](https://mlsanity.com/paper/1909.05840/) (2019)
- [DeeBERT: Dynamic Early Exiting for Accelerating BERT Inference](https://mlsanity.com/paper/2004.12993/) (2020)
- [Q8BERT: Quantized 8Bit BERT](https://mlsanity.com/paper/1910.06188/) (2019)
- [BERT Loses Patience: Fast and Robust Inference with Early Exit](https://mlsanity.com/paper/2006.04152/) (2020)
- [I-BERT: Integer-only BERT Quantization](https://mlsanity.com/paper/2101.01321/) (2021)
- [DynaBERT: Dynamic BERT with Adaptive Width and Depth](https://mlsanity.com/paper/2004.04037/) (2020)
- [FastBERT: a Self-distilling BERT with Adaptive Inference Time](https://mlsanity.com/paper/2004.02178/) (2020)
- [Proving the Lottery Ticket Hypothesis: Pruning is All You Need](https://mlsanity.com/paper/2002.00585/) (2020)
- [MLPerf Inference Benchmark](https://mlsanity.com/paper/1911.02549/) (2019)
- [Clipper: A Low-Latency Online Prediction Serving System](https://mlsanity.com/paper/1612.03079/) (2016)
- [Knowledge Distillation by On-the-Fly Native Ensemble](https://mlsanity.com/paper/1806.04606/) (2018)
- [Compressing BERT: Studying the Effects of Weight Pruning on Transfer Learning](https://mlsanity.com/paper/2002.08307/) (2020)
- [Learned Token Pruning for Transformers](https://mlsanity.com/paper/2107.00910/) (2021)
- [Encoder-Decoder Models Can Benefit from Pre-trained Masked Language Models in Grammatical Error Correction](https://mlsanity.com/paper/2005.00987/) (2020)
- [NAS-BERT: Task-Agnostic and Adaptive-Size BERT Compression with Neural Architecture Search](https://mlsanity.com/paper/2105.14444/) (2021)
- [Ensemble Knowledge Distillation for Learning Improved and Efficient Networks](https://mlsanity.com/paper/1909.08097/) (2019)
- [Model Compression with Two-stage Multi-teacher Knowledge Distillation for Web Question Answering System](https://mlsanity.com/paper/1910.08381/) (2019)
- [MoEBERT: from BERT to Mixture-of-Experts via Importance-Guided Adaptation](https://mlsanity.com/paper/2204.07675/) (2022)
- [Learning Deep ResNet Blocks Sequentially using Boosting Theory](https://mlsanity.com/paper/1706.04964/) (2017)
- [ByteTransformer: A High-Performance Transformer Boosted for Variable-Length Inputs](https://mlsanity.com/paper/2210.03052/) (2022)

## Abstract (arXiv)

Due to high accuracy, BERT-like models have been widely adopted by text mining and web searching. However, large BERT-like models suffer from inefficient online inference, facing the following two problems on GPUs: (1) their high accuracy relies on the large model depth, which linearly increases the sequential computation on GPUs; (2) stochastic and dynamic online workloads cause extra costs from batching and paddings. Therefore, we present sys for the real-world setting of GPU inference on online workloads. At its core, sys adopts stacking distillation and boosting ensemble, distilling the original deep model into a group of shallow but virtually stacked student models running in parallel. This enables sys to achieve a lower model depth (e.g., two layers) than the others and the lowest inference latency while maintaining accuracy. In addition, adaptive student pruning realizes dynamic student numbers according to changing online workloads. Especially for occasional workload bursts, it can temporarily decrease the student number with minimal accuracy loss to improve system throughput. We conduct comprehensive experiments to verify the effectiveness, whose results show that sys outperforms the baselines by 4.1×~ 1.6× in latency while maintaining accuracy and achieves up to 22.27× higher throughput for workload bursts.