So, if there’s a demo or how can I use the FPL to optimize my own deep learning op or network?
For example, our group tried to use PPCG to optimize the performance of complex operators (using the default configuration, such as tile size, etc.), and found that the effect was not good. Report Link(Sorry for the Chinese report here. If necessary, I will translate it into English and reissue it again.).
Therefore, I very much hope that FPL can have better results in dealing with such operators, and I am urgent to have such an experiment.
By the way, I find that the project has not been completed so far. Do I have the opportunity to participate in the development of unfinished projects?