Please wait ...
0% Complete
Home
/
14th International Conference on Computer and Knowledge Engineering
Improve the utility of tensor cores by compacting sparse matrix technique
Authors :
Mohammad.S Abazari
1
Mahsa Zahedi
2
Abdorreza Savadi
3
1- Ferdowsi university of mashhad
2- Ferdowsi university of mashhad
3- Ferdowsi university of mashhad
Keywords :
Tensor Cores،Neural Networks،Convolution Operations،Graphics Processing Unit
Abstract :
Neural networks have demanding computational requirements, particularly in matrix multiplication operations. To address this challenge, we propose a model that combines network pruning and matrix compression techniques. Our approach leverages NVIDIA's tensor cores, which excel at efficient matrix operations. We compress the network weights based on the tensor core structure and perform convolutions using the compressed weight matrix on the tensor cores. Our model incorporates neural network pruning, mixed-precision training, and compression of network weight tensors using the im2col algorithm and CSR format. We also utilize tensor kernels with a block size of 16x16 for multiplication. We evaluate the performance of various models, including pruned, AMP-optimized, combined pruning and AMP techniques, and our proposed model. Our evaluation reveals a significant improvement in performance compared to a simple baseline model. Through an extensive analysis of related works, we establish foundational concepts, present our proposed model, and share the obtained results.
Papers List
List of archived papers
An Automated Visual Defect Segmentation for Flat Steel Surface Using Deep Neural Networks
Dorna Nourbakhsh Sabet - Mohammad Reza Zarifi - Javad Khoramdel - Yasamin Borhani - Esmaeil Najafi
Optimization Resource Allocation in NOMA-based Fog Computing with a Hybrid Algorithm
Zohreh Torki - S.Mojtaba Matinkhah
Cluster Sampling: A Cluster-Driven Sampling Strategy for Deep Metric Learning
Hamideh Rafiee - Ahmad Ali Abin - Seyed Soroush Majd
A Self-Configurable Model for Cloud Resource Allocation
Ali Bazghandi
Cross-project Defect Prediction with An Enhanced Transfer Boosting Algorithm
Nazgol Nikravesh - Mohammad Reza Keyvanpour
Weakly Supervised Learning in a Group of Learners with Communication
Ali Ganjbakhsh - Ahad Harati
Blind image quality assessment based on Multi-resolution Local Structures
Seyed Majid Khorashadizadeh - Mehdi Sadeghi Bakhi - Fatemeh Seifishahpar - AliMohammad Latif
Novel Insights in Deep Learning for Predicting Climate Phenomena
Mohammad Naisipour - Saghar Ganji - Iraj Saeedpanah - Behnam Mehrakizadeh - Ahmad Reza Labibzadeh
Computational Microscopy Based on Fourier Ptychography using Embedded Architecture
Rezvan Mir - Abedin Vahedian
Efficient Vision Transformer for Accurate Traffic Sign Detection
Javad Mirzapour Kaleybar - Hooman Khaloo - Avaz Naghipour
more
Samin Hamayesh - Version 46.0.2