Numpy-UHD (npuhd), is a very minimalist approach to providing programmatic python/numpy radio access to the RF spectrum.  In many cases, especially when considering RadioML type tasks, its handy to have a very minimal way to simply grab or insert numpy vectors of sample data to or frame the radio front-end into your favorite numpy based tensor processor or machine learning backend.  In some of these cases GNU Radio can seem a bit heavy handed for such tasks, since you may not actually want to set up processing flowgraph that doesn’t anything more than capture samples.  This library attempts to fill this gap and provide a rapid link between the RF and ML worlds.

Additional detail and source code is avilable at