SkipViT: Speeding Up Vision Transformers with a Token-Level Skip Connection

no code implementations • 27 Jan 2024 • Foozhan Ataiefard, Walid Ahmed, Habib Hajimolahoseini, Saina Asani, Farnoosh Javadi, Mohammad Hassanpour, Omar Mohamed Awad, Austin Wen, Kangling Liu, Yang Liu

Our method does not add any parameters to the ViT model and aims to find the best trade-off between training throughput and achieving a 0% loss in the Top-1 accuracy of the final model.

SwiftLearn: A Data-Efficient Training Method of Deep Learning Models using Importance Sampling

no code implementations • 25 Nov 2023 • Habib Hajimolahoseini, Omar Mohamed Awad, Walid Ahmed, Austin Wen, Saina Asani, Mohammad Hassanpour, Farnoosh Javadi, Mehdi Ahmadi, Foozhan Ataiefard, Kangling Liu, Yang Liu

In this paper, we present SwiftLearn, a data-efficient approach to accelerate training of deep learning models using a subset of data samples selected during the warm-up stages of training.

GQKVA: Efficient Pre-training of Transformers by Grouping Queries, Keys, and Values

no code implementations • 6 Nov 2023 • Farnoosh Javadi, Walid Ahmed, Habib Hajimolahoseini, Foozhan Ataiefard, Mohammad Hassanpour, Saina Asani, Austin Wen, Omar Mohamed Awad, Kangling Liu, Yang Liu

We tested our method on ViT, which achieved an approximate 0. 3% increase in accuracy while reducing the model size by about 4% in the task of image classification.

Image Classification

Speeding up Resnet Architecture with Layers Targeted Low Rank Decomposition

no code implementations • 21 Sep 2023 • Walid Ahmed, Habib Hajimolahoseini, Austin Wen, Yang Liu

Compression of a neural network can help in speeding up both the training and the inference of the network.