Mar 22, 2023 · We propose the first silicon photonic hardware neural network accelerator called TRON for transformer-based models such as BERT, and Vision Transformers.
In this paper, we introduce TRON, the first silicon-photonic- based transformer accelerator that can accelerate inference of a broad family of transformer ...
The proposed TRON architecture used noncoherent optical computing, together with the cross-layer design principles first proposed in Crosslight, power ...
Our analysis demonstrates that TRON exhibits at least 14x better throughput and 8x better energy efficiency, in comparison to state-of-the-art transformer ...
Deep Learning Hardware Accelerators with Silicon Photonics
www.engr.colostate.edu › siph-accelerators
The research objective of this project is to design new hardware accelerators for machine learning workloads that leverage the remarkable communication and ...
2020. Tron: Transformer neural network acceleration with non-coherent silicon photonics. S Afifi, F Sunny, M Nikdast, S Pasricha. Proceedings of the Great ...
We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Transformer neural networks are rapidly being integrated into state-of-the-art solutions for natural language processing (NLP) and computer vision.
This paper outlines how optical communication and computation can be leveraged in 2.5D platforms to realize energy-efficient and high throughput 2.5D ML ...
Jun 4, 2023 · Pasricha, “TRON: Transformer Neural Network Acceleration with Non-Coherent Silicon Photonics”, ACM GLSVLSI, 2023. (https://lnkd.in/gGE4qWwP) ...