Navigating Extremes: Dynamic Sparsity in Large Output Space

• URL: http://arxiv.org/abs/2411.03171v2
• Date: Wed, 06 Nov 2024 17:19:10 GMT
• Title: Navigating Extremes: Dynamic Sparsity in Large Output Space
• Authors: Nasib Ullah, Erik Schultheis, Mike Lasby, Yani Ioannou, Rohit Babbar,
• Abstract summary: Dynamic Sparse Training (DST) has emerged as an alternative to post-training pruning for generating efficient models. We leverage recent advances in semi-structured sparse training to apply DST in the domain of classification with large output spaces. We find that poor gradient flow from the sparse classifier to the dense text encoder make it difficult to learn good input representations.
• Score: 5.231219025536679
• License:
• Abstract: In recent years, Dynamic Sparse Training (DST) has emerged as an alternative to post-training pruning for generating efficient models. In principle, DST allows for a more memory efficient training process, as it maintains sparsity throughout the entire training run. However, current DST implementations fail to capitalize on this in practice. Because sparse matrix multiplication is much less efficient than dense matrix multiplication on GPUs, most implementations simulate sparsity by masking weights. In this paper, we leverage recent advances in semi-structured sparse training to apply DST in the domain of classification with large output spaces, where memory-efficiency is paramount. With a label space of possibly millions of candidates, the classification layer alone will consume several gigabytes of memory. Switching from a dense to a fixed fan-in sparse layer updated with sparse evolutionary training (SET); however, severely hampers training convergence, especially at the largest label spaces. We find that poor gradient flow from the sparse classifier to the dense text encoder make it difficult to learn good input representations. By employing an intermediate layer or adding an auxiliary training objective, we recover most of the generalisation performance of the dense model. Overall, we demonstrate the applicability and practical benefits of DST in a challenging domain -- characterized by a highly skewed label distribution that differs substantially from typical DST benchmark datasets -- which enables end-to-end training with millions of labels on commodity hardware.

Related papers

• Sparser Training for On-Device Recommendation Systems [50.74019319100728]
  We propose SparseRec, a lightweight embedding method based on Dynamic Sparse Training (DST) It avoids dense gradients during backpropagation by sampling a subset of important vectors.
  arXiv  Detail & Related papers  (2024-11-19T03:48:48Z)
• Compute Better Spent: Replacing Dense Layers with Structured Matrices [77.61728033234233]
  We identify more efficient alternatives to dense matrices, as exemplified by the success of convolutional networks in the image domain. We show that different structures often require drastically different initialization scales and learning rates, which are crucial to performance. We propose a novel matrix family containing Monarch matrices, the Block-Train, which we show performs better than dense for the same compute on multiple tasks.
  arXiv  Detail & Related papers  (2024-06-10T13:25:43Z)
• VeLoRA: Memory Efficient Training using Rank-1 Sub-Token Projections [35.133698935322634]
  Large language models (LLMs) have recently emerged as powerful tools for tackling many language-processing tasks. We identify and characterise the important components needed for effective model convergence using gradient descent. This result leads us to a cheap and memory-efficient algorithm for both fine-tuning and pre-training LLMs.
  arXiv  Detail & Related papers  (2024-05-28T09:23:14Z)
• Dynamic Sparsity Is Channel-Level Sparsity Learner [91.31071026340746]
  Dynamic sparse training (DST) is a leading sparse training approach. Channel-aware dynamic sparse (Chase) seamlessly translates the promise of unstructured dynamic sparsity to channel-level sparsity. Our approach translates unstructured sparsity to channel-wise sparsity.
  arXiv  Detail & Related papers  (2023-05-30T23:33:45Z)
• Dynamic Sparse Training with Structured Sparsity [11.778353786208765]
  Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training. We propose a sparse-to-sparse DST method, Structured RigL (SRigL), to learn a variant of fine-grained structured N:M sparsity. We demonstrate a real-world acceleration of 3.4x/2.5x on CPU for online inference and 1.7x/13.0x on GPU for inference with a batch size of 256.
  arXiv  Detail & Related papers  (2023-05-03T17:48:55Z)
• Distributed Adversarial Training to Robustify Deep Neural Networks at Scale [100.19539096465101]
  Current deep neural networks (DNNs) are vulnerable to adversarial attacks, where adversarial perturbations to the inputs can change or manipulate classification. To defend against such attacks, an effective approach, known as adversarial training (AT), has been shown to mitigate robust training. We propose a large-batch adversarial training framework implemented over multiple machines.
  arXiv  Detail & Related papers  (2022-06-13T15:39:43Z)
• Improving Semantic Segmentation via Self-Training [75.07114899941095]
  We show that we can obtain state-of-the-art results using a semi-supervised approach, specifically a self-training paradigm. We first train a teacher model on labeled data, and then generate pseudo labels on a large set of unlabeled data. Our robust training framework can digest human-annotated and pseudo labels jointly and achieve top performances on Cityscapes, CamVid and KITTI datasets.
  arXiv  Detail & Related papers  (2020-04-30T17:09:17Z)
• Learning Low-rank Deep Neural Networks via Singular Vector Orthogonality Regularization and Singular Value Sparsification [53.50708351813565]
  We propose SVD training, the first method to explicitly achieve low-rank DNNs during training without applying SVD on every step. We empirically show that SVD training can significantly reduce the rank of DNN layers and achieve higher reduction on computation load under the same accuracy.
  arXiv  Detail & Related papers  (2020-04-20T02:40:43Z)

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.