As of now pygpufit has type checks and only allows for np.float32. The REALdatatype in Gpufit/definitions.h allows to switch between 32bit single prec. and 64bit double prec. floating point in Gpufit ...