Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

TensorGlue: a framework for FPGA-based deep learning design

Giakoumakis Pavlos

Full record


URI: http://purl.tuc.gr/dl/dias/69B2ABAE-EA90-4347-BFDC-5352F4F01B5E
Year 2019
Type of Item Master Thesis
License
Details
Bibliographic Citation Pavlos Giakoumakis, "TensorGlue: a framework for FPGA-based deep learning design", Master Thesis, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2019 https://doi.org/10.26233/heallink.tuc.83920
Appears in Collections

Summary

In a deep learning framework, the designer provides a description of the neural network architecture, in the form of a computational graph (data-flow graph). The tool is able process this graph and either run it efficiently on fixed-hardware, or generate automatically additional graphs to train the neural network. Nevertheless, this kind of formalization using computational graphs is very close to the hardware design process. The graph can be processed in many ways to not only run the described architecture on fixed-hardware, but to generate hardware designs as well.In this work, we designed and implemented a novel framework that resembles deep learning frameworks but generates hardware designs in the form of synthesizable C++.

Available Files

Services

Statistics