This repository is the the open-source implementation of the tool perf4sight proposed in the paper perf4sight: A toolflow to model CNN training performance on Edge GPUs published at the Workshop on "Embeeded and Real-World Computer Vision in Autonomous Driving" at ICCV 2021.

git submodule update --init --recursive

Please refer to https://github.com/ICIdsl/performance_modelling.git on how to use this tool.

If you reference this work in a publication, please cite it through the following citation:

@misc{rajagopal2021perf4sight,
      title={perf4sight: A toolflow to model CNN training performance on Edge GPUs}, 
      author={Aditya Rajagopal and Christos-Savvas Bouganis},
      year={2021},
      eprint={2108.05580},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}