microsoft/Windows-Machine-Learning

Question : how to get the intermediate data of a evaluation?

oldoldman opened this issue · 1 comments

It seems I can just bind to the input/output features returned by LearningModel.InputFeatures and LearningModel.OutputFeatures. Take inception_v1 as an example , I can bind to data_0 and prob_1 only, an attempt to bind to some intermediate node (for example, conv1/7x7_s2_1 ) will fail. So is there some way to bind the intermediate node? Many thanks!!!

hi @oldoldman , please see the instructions for how to do this with WinMLDashboard. While you can't bind to the intermediate nodes, you can dump the outputs to disk.