Work

Enhancing Transformer Architecture: Techniques for Efficient Inference

Year: 2025

Type: article

Abstract: This paper explores recent advancements in optimizing transformer architectures for efficient inference. We investigate various techniques including pruning, quantization, knowledge distillation, and architectural modifications. Our experimental results demonstrate that combining these approaches can reduce inference time by up to 74% while maintaining over 95% of the original performance. We also introduce a novel attention mechanism that dynamically allocates computational resources based on input complexity. Our implementation shows promise for edge device deployment where computational resources are constrained.

Source: International Journal of Scientific Research in Computer Science Engineering and Information Technology

DOI: 10.32628/cseit25112757

Author: Kingsuk Chakrabarty

Language: English

Topic: Advanced Neural Network Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: diamond

Funder: National Science Foundation

Award: Collaborative Research: Robust Deep Learning in Real Physical Space: Generalization, Scalability, and Credibility