Dear HLS4ML,
Myself and my colleagues (Zhewen Yu and Dr. Christos-Savvas Bouganis) have been developing a tool which abstracts the CNN to FPGA mapping problem and implements optimisation methods that optimise the mapping of a particular CNN model to a target FPGA platform. We have integrated HLS4ML into this tool and demonstrated it's ability to discover high performance designs across a range of platforms. We would like to share this work with you, as we hope you and the users of your tool will find it useful. It would also be great to discuss how we could work together to improve both tools, as during our evaluation of HLS4ML we discovered some limitations.
The SAMO tool can be found on GitHub: https://github.com/AlexMontgomerie/samo
There is also a pre-print which details the tool: https://arxiv.org/abs/2112.00170?context=cs
Best,Alexander Montgomerie-Corcoran
This email from jdu...@physics.ucsd.edu originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list to disable email stamping for this address. |