Timezone: »
Attention based Transformer architecture has enabled significant advances in the field of natural language processing. In addition to new pre-training techniques, recent improvements crucially rely on working with a relatively larger embedding dimension for tokens. Unfortunately, this leads to models that are prohibitively large to be employed in the downstream tasks. In this paper we identify one of the important factors contributing to the large embedding size requirement. In particular, our analysis highlights that the scaling between the number of heads and the size of each head in the current architecture gives rise to a low-rank bottleneck in attention heads, causing this limitation. We further validate this in our experiments. As a solution we propose to set the head size of an attention unit to input sequence length, and independent of the number of heads, resulting in multi-head attention layers with provably more expressive power. We empirically show that this allows us to train models with a relatively smaller embedding dimension and with better performance scaling.
Author Information
Srinadh Bhojanapalli (Google AI)
Chulhee Yun (MIT)
Ankit Singh Rawat (Google)
Sashank Jakkam Reddi (Google)
Sanjiv Kumar (Google Research, NY)
More from the Same Authors
-
2023 : SpecTr: Fast Speculative Decoding via Optimal Transport »
Ziteng Sun · Ananda Suresh · Jae Ro · Ahmad Beirami · Himanshu Jain · Felix Xinnan Yu · Michael Riley · Sanjiv Kumar -
2023 Poster: On the Role of Attention in Prompt-tuning »
Samet Oymak · Ankit Singh Rawat · Mahdi Soltanolkotabi · Christos Thrampoulidis -
2023 Poster: A Statistical Perspective on Retrieval-Based Models »
Soumya Basu · Ankit Singh Rawat · Manzil Zaheer -
2023 Poster: Efficient Training of Language Models using Few-Shot Learning »
Sashank Jakkam Reddi · Sobhan Miryoosefi · Stefani Karp · Shankar Krishnan · Satyen Kale · Seungyeon Kim · Sanjiv Kumar -
2022 Poster: In defense of dual-encoders for neural ranking »
Aditya Menon · Sadeep Jayasumana · Ankit Singh Rawat · Seungyeon Kim · Sashank Jakkam Reddi · Sanjiv Kumar -
2022 Spotlight: In defense of dual-encoders for neural ranking »
Aditya Menon · Sadeep Jayasumana · Ankit Singh Rawat · Seungyeon Kim · Sashank Jakkam Reddi · Sanjiv Kumar -
2022 Poster: Private Adaptive Optimization with Side information »
Tian Li · Manzil Zaheer · Sashank Jakkam Reddi · Virginia Smith -
2022 Poster: Robust Training of Neural Networks Using Scale Invariant Architectures »
Zhiyuan Li · Srinadh Bhojanapalli · Manzil Zaheer · Sashank Jakkam Reddi · Sanjiv Kumar -
2022 Spotlight: Private Adaptive Optimization with Side information »
Tian Li · Manzil Zaheer · Sashank Jakkam Reddi · Virginia Smith -
2022 Oral: Robust Training of Neural Networks Using Scale Invariant Architectures »
Zhiyuan Li · Srinadh Bhojanapalli · Manzil Zaheer · Sashank Jakkam Reddi · Sanjiv Kumar -
2021 Poster: A statistical perspective on distillation »
Aditya Menon · Ankit Singh Rawat · Sashank Jakkam Reddi · Seungyeon Kim · Sanjiv Kumar -
2021 Poster: Disentangling Sampling and Labeling Bias for Learning in Large-output Spaces »
Ankit Singh Rawat · Aditya Menon · Wittawat Jitkrittum · Sadeep Jayasumana · Felix Xinnan Yu · Sashank Jakkam Reddi · Sanjiv Kumar -
2021 Spotlight: A statistical perspective on distillation »
Aditya Menon · Ankit Singh Rawat · Sashank Jakkam Reddi · Seungyeon Kim · Sanjiv Kumar -
2021 Spotlight: Disentangling Sampling and Labeling Bias for Learning in Large-output Spaces »
Ankit Singh Rawat · Aditya Menon · Wittawat Jitkrittum · Sadeep Jayasumana · Felix Xinnan Yu · Sashank Jakkam Reddi · Sanjiv Kumar -
2021 Poster: Federated Composite Optimization »
Honglin Yuan · Manzil Zaheer · Sashank Jakkam Reddi -
2021 Spotlight: Federated Composite Optimization »
Honglin Yuan · Manzil Zaheer · Sashank Jakkam Reddi -
2020 Poster: Does label smoothing mitigate label noise? »
Michal Lukasik · Srinadh Bhojanapalli · Aditya Menon · Sanjiv Kumar -
2020 Poster: Accelerating Large-Scale Inference with Anisotropic Vector Quantization »
Ruiqi Guo · Philip Sun · Erik Lindgren · Quan Geng · David Simcha · Felix Chern · Sanjiv Kumar -
2020 Poster: SCAFFOLD: Stochastic Controlled Averaging for Federated Learning »
Sai Praneeth Reddy Karimireddy · Satyen Kale · Mehryar Mohri · Sashank Jakkam Reddi · Sebastian Stich · Ananda Theertha Suresh -
2020 Poster: Federated Learning with Only Positive Labels »
Felix Xinnan Yu · Ankit Singh Rawat · Aditya Menon · Sanjiv Kumar -
2019 : Structured matrices for efficient deep learning »
Sanjiv Kumar -
2019 Poster: Escaping Saddle Points with Adaptive Gradient Methods »
Matthew Staib · Sashank Jakkam Reddi · Satyen Kale · Sanjiv Kumar · Suvrit Sra -
2019 Oral: Escaping Saddle Points with Adaptive Gradient Methods »
Matthew Staib · Sashank Jakkam Reddi · Satyen Kale · Sanjiv Kumar · Suvrit Sra -
2019 Poster: Learning a Compressed Sensing Measurement Matrix via Gradient Unrolling »
Shanshan Wu · Alexandros Dimakis · Sujay Sanghavi · Felix Xinnan Yu · Daniel Holtmann-Rice · Dmitry Storcheus · Afshin Rostamizadeh · Sanjiv Kumar -
2019 Oral: Learning a Compressed Sensing Measurement Matrix via Gradient Unrolling »
Shanshan Wu · Alexandros Dimakis · Sujay Sanghavi · Felix Xinnan Yu · Daniel Holtmann-Rice · Dmitry Storcheus · Afshin Rostamizadeh · Sanjiv Kumar -
2018 Poster: Loss Decomposition for Fast Learning in Large Output Spaces »
En-Hsu Yen · Satyen Kale · Felix Xinnan Yu · Daniel Holtmann-Rice · Sanjiv Kumar · Pradeep Ravikumar -
2018 Oral: Loss Decomposition for Fast Learning in Large Output Spaces »
En-Hsu Yen · Satyen Kale · Felix Xinnan Yu · Daniel Holtmann-Rice · Sanjiv Kumar · Pradeep Ravikumar -
2017 Poster: Stochastic Generative Hashing »
Bo Dai · Ruiqi Guo · Sanjiv Kumar · Niao He · Le Song -
2017 Talk: Stochastic Generative Hashing »
Bo Dai · Ruiqi Guo · Sanjiv Kumar · Niao He · Le Song -
2017 Poster: Distributed Mean Estimation with Limited Communication »
Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Brendan McMahan -
2017 Talk: Distributed Mean Estimation with Limited Communication »
Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Brendan McMahan