Skip to content
Grigori Fursin edited this page Feb 5, 2017 · 8 revisions

This framework and documentation is continuously updated - please help us improve it via CK mailing list

We need to have easily customizable Caffe builds via JSON API to be able to plug it to our framework for collaborative benchmarking and optimization of workloads across diverse inputs and hardware provided by volunteers (see project page, live repo and papers 1, 2, 3 for more details).

  • Introduction
 * Collective knowledge customizable builds: [http://github.com/ctuning/ck GitHub sources], [https://github.com/ctuning/ck/wiki/Portable-workflows  documentation]
 * Collaborative optimization concepts: [http://arxiv.org/abs/1506.06256 CPC'15], [http://schedule.armtechcon.com/session/know-your-workloads-design-more-efficient-systems ARM TechCon'16 talk], [https://github.com/ctuning/ck/wiki/Demo-ARM-TechCon'16 ARM TechCon'16 demo], [http://tinyurl.com/zyupd5v DATE'16].
 * Other DNN engines with JSON API for customized builds via CK: [https://github.com/dividiti/ck-tensorflow CK-TensorFlow], [https://github.com/ctuning/ck-tiny-dnn CK-TinyDNN], [https://play.google.com/store/apps/details?id=openscience.crowdsource.video.experiments Android app for DNN crowd-benchmarking and crowd-tuning]

Clone this wiki locally