issue with template scaling

[marius10p]

I am trying to make Phy use the whitened and registered binary file instead of the raw, and I am stuck on the template un-whitening step. Since the binary file is already white and high-pass, I replaced the whitening matrix with an identity matrix divided by the scaling constant, and I set hp_filtered to True in params.py. This makes the templates too small, so I am clearly missing some scaling factor from somewhere. Any ideas? Could you point me to where in the code the templates are "un-whitened"?

[rossant]

There have been long-standing scaling issues, which is partly why I've been working on refactoring this part of the code for a while, it's in phylib/io/model.py (there's an ongoing branch model-refactor where there's an additionnal loader.py module). I hope this will be ready within a few weeks.
Note that hp_filtered is a historical leftover and is not used currently.