Exploring the Benefit of Activation Sparsity in Pre-training

• URL: http://arxiv.org/abs/2410.03440v1
• Date: Fri, 4 Oct 2024 13:53:33 GMT
• Title: Exploring the Benefit of Activation Sparsity in Pre-training
• Authors: Zhengyan Zhang, Chaojun Xiao, Qiujieli Qin, Yankai Lin, Zhiyuan Zeng, Xu Han, Zhiyuan Liu, Ruobing Xie, Maosong Sun, Jie Zhou,
• Abstract summary: We study how activation properties change during pre-training. We propose Switchable Sparse-Dense Learning (SSD) SSD achieves comparable performance with identical model size and reduces pre-training costs.
• Score: 117.25661020250658
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Pre-trained Transformers inherently possess the characteristic of sparse activation, where only a small fraction of the neurons are activated for each token. While sparse activation has been explored through post-training methods, its potential in pre-training remains untapped. In this work, we first study how activation properties change during pre-training. Our examination reveals that Transformers exhibit sparse activation throughout the majority of the pre-training process while the activation correlation keeps evolving as training progresses. Leveraging this observation, we propose Switchable Sparse-Dense Learning (SSD). SSD adaptively switches between the Mixtures-of-Experts (MoE) based sparse training and the conventional dense training during the pre-training process, leveraging the efficiency of sparse training and avoiding the static activation correlation of sparse training. Compared to dense training, SSD achieves comparable performance with identical model size and reduces pre-training costs. Moreover, the models trained with SSD can be directly used as MoE models for sparse inference and achieve the same performance as dense models with up to $2\times$ faster inference speed. Codes are available at https://github.com/thunlp/moefication.

Related papers

• Mixed Sparsity Training: Achieving 4$\times$ FLOP Reduction for Transformer Pretraining [32.925150708409205]
  Mixed Sparsity Training (MST) is an efficient pretraining method that can reduce about $75%$ of Floating Point Operations (FLOPs) while maintaining performance. Our experiment on GPT-2 showcases a FLOP reduction of $4times$ without compromising performance.
  arXiv  Detail & Related papers  (2024-08-21T16:13:16Z)
• Enhancing Neural Training via a Correlated Dynamics Model [2.9302545029880394]
  Correlation Mode Decomposition (CMD) is an algorithm that clusters the parameter space into groups, that display synchronized behavior across epochs. We introduce an efficient CMD variant, designed to run concurrently with training. Our experiments indicate that CMD surpasses the state-of-the-art method for compactly modeled dynamics on image classification.
  arXiv  Detail & Related papers  (2023-12-20T18:22:49Z)
• Latent State Models of Training Dynamics [51.88132043461152]
  We train models with different random seeds and compute a variety of metrics throughout training. We then fit a hidden Markov model (HMM) over the resulting sequences of metrics. We use the HMM representation to study phase transitions and identify latent "detour" states that slow down convergence.
  arXiv  Detail & Related papers  (2023-08-18T13:20:08Z)
• Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening [51.34904967046097]
  Selective Synaptic Dampening (SSD) is a fast, performant, and does not require long-term storage of the training data. We present a novel two-step, post hoc, retrain-free approach to machine unlearning which is fast, performant, and does not require long-term storage of the training data.
  arXiv  Detail & Related papers  (2023-08-15T11:30:45Z)
• Task-Customized Self-Supervised Pre-training with Scalable Dynamic Routing [76.78772372631623]
  A common practice for self-supervised pre-training is to use as much data as possible. For a specific downstream task, however, involving irrelevant data in pre-training may degenerate the downstream performance. It is burdensome and infeasible to use different downstream-task-customized datasets in pre-training for different tasks.
  arXiv  Detail & Related papers  (2022-05-26T10:49:43Z)
• Training for temporal sparsity in deep neural networks, application in video processing [0.30458514384586394]
  Activation sparsity improves compute efficiency and resource utilization in sparsity-aware neural network accelerators. We introduce a new layer (called Delta Activation Layer) to promote temporal sparsity of activations during training. We report an almost 3x improvement of activation sparsity, with recoverable loss of model accuracy after longer training.
  arXiv  Detail & Related papers  (2021-07-15T13:17:11Z)
• AC/DC: Alternating Compressed/DeCompressed Training of Deep Neural Networks [78.62086125399831]
  We present a general approach called Alternating Compressed/DeCompressed (AC/DC) training of deep neural networks (DNNs) AC/DC outperforms existing sparse training methods in accuracy at similar computational budgets. An important property of AC/DC is that it allows co-training of dense and sparse models, yielding accurate sparse-dense model pairs at the end of the training process.
  arXiv  Detail & Related papers  (2021-06-23T13:23:00Z)
• FracTrain: Fractionally Squeezing Bit Savings Both Temporally and Spatially for Efficient DNN Training [81.85361544720885]
  We propose FracTrain that integrates progressive fractional quantization which gradually increases the precision of activations, weights, and gradients. FracTrain reduces computational cost and hardware-quantified energy/latency of DNN training while achieving a comparable or better (-0.12%+1.87%) accuracy.
  arXiv  Detail & Related papers  (2020-12-24T05:24:10Z)
• Regularized Evolutionary Population-Based Training [11.624954122221562]
  This paper presents an algorithm called Population-Based Training (EPBT) that interleaves the training of a DNN's weights with the metalearning of loss functions. EPBT results in faster, more accurate learning on image classification benchmarks.
  arXiv  Detail & Related papers  (2020-02-11T06:28:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.