Looking at #79 , I figured that Observation.render could break if the model frame was initialized with Frame.from_observations, because frame and observations may have different WCSs but the same pixel scale. The ConvolutionRenderer would therefore be used, but frames boxes would not necessarily be the same sizes. This PR fixes it.
I also removed all the numpy dependencies in frame.py and used jax.numpy instead.
Looking at #79 , I figured that
Observation.render
could break if the model frame was initialized withFrame.from_observations
, because frame and observations may have different WCSs but the same pixel scale. TheConvolutionRenderer
would therefore be used, but frames boxes would not necessarily be the same sizes. This PR fixes it.I also removed all the
numpy
dependencies inframe.py
and usedjax.numpy
instead.