Work

Efficient mapping of neural networks on FPGA with generated VHDL code

Year: 2023
Type: article
Abstract: FPGAs are an ideal platform to accelerate AI processing at the edge. However, simple toolchains are lacking to deploy AI networks directly on the FPGA. We present a framework that allows the design of custom AI accelerators for FPGAs by generating application application-specific HDL code. Unlike other solutions, the framework proposes a dataflow-style hardware mapping at the RTL level that is specific to the neural network and makes optimal use of available resources. We focused on providing an easy-to-use interface and extensible core that allows developers to add custom operators, optimizers, objectives, metrics, and callbacks.
Authors: Delafontaine, Thierry, Rosenthal, Matthias
Language: English
Open Access status: green
OpenAlex ID: W7046580700
FWCI:
Cites:
Cited by:
Related to: 10