Prediction using keras

[ylikx]

Prediction using keras

[ylikx]

The error says that there is no call_py function specifically for ndarrays as first argument. You can however use an object and the use cast to create a ndarray

type(object) :: ypred_obj
!...
ierror = call_py(ypred_obj, loaded_model, "predict",  args)
ierror = cast(Ypred, ypred_obj)

[ylikx]

Hi, sorry for the late reply.
Maybe this helps: https://github.com/ylikx/forpy/wiki/Working-with-arrays#accessing-array-elements-with-ndarrayget_data

I think the most efficient way would be to keep working with the Fortran pointer which points to the internal representation of the numpy array. Unfortunately one has to then keep working with the transpose on the Fortran side.
Sometimes it might be possible to write the Python part in a way that it produces not the original matrix but the transpose (in C order). Then in Fortran one could work with the original matrix (this time in Fortran order).