I am aware that #10 already addresses this issue. However, the way that PR goes about fixing the problem in #9 (in my opinion) is not ideal.
load_model expects a str or "h5py.File or h5py.Group object from which to load the model". @rrwick handles the model files internally with the pathlib.Path class which is an obvious design choice as these objects are indeed paths and therefore the added functionality this class brings within python is something Ryan uses a lot in the code.
I don't think changing the function find_model to return a str instead of a Path to satisfy the single use of keras.load_model is warranted.
I am aware that #10 already addresses this issue. However, the way that PR goes about fixing the problem in #9 (in my opinion) is not ideal.
load_model
expects astr
or "h5py.File or h5py.Group object from which to load the model". @rrwick handles the model files internally with thepathlib.Path
class which is an obvious design choice as these objects are indeed paths and therefore the added functionality this class brings within python is something Ryan uses a lot in the code.I don't think changing the function
find_model
to return astr
instead of aPath
to satisfy the single use ofkeras.load_model
is warranted.