NUMPY 1.20 above error: TypeError: expected dtype object, got 'numpy.dtype[float64]'

[pinghongyeh]

Hi Rutger,

I have encounterd errors while running NODDI_mod_watson.fit,

Here is the error

NODDI_fit_watson = NODDI_mod_watson.fit(gtab_dmipy, data, mask=mask) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/core/modeling_framework.py", line 1201, in fit self, self.scheme, x0_, Ns, N_sphere_samples) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/optimizers/brute2fine.py", line 76, in init model, x0_vector_, Ns, N_sphere_samples) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/optimizers/brute2fine.py", line 168, in precompute_signal_grid self.acquisition_scheme, self.parameter_grid) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/core/modeling_framework.py", line 1300, in simulate_signal E_2d[i] = self(acquisition_scheme, **parameters) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/core/modeling_framework.py", line 1370, in call acquisition_scheme_or_vertices, **parameters)) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/distributions/distribute_models.py", line 390, in call return self.sh_convolved_model(acquisition_scheme, **kwargs) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/distributions/distribute_models.py", line 482, in sh_convolved_model **distribution_parameters) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/distributions/distributions.py", line 149, in spherical_harmonics_representation watson_sf = self(hemisphere.vertices, mu=mu, odi=odi) File "/home/yehp/anaconda3/lib/python3.7/site-packages/dmipy/distributions/distributions.py", line 123, in call mu_cart = utils.unitsphere2cart_1d(mu) TypeError: expected dtype object, got 'numpy.dtype[float64]'

I have numpy=1.21.1 installed. The scripts ran fine using numpy< 1.18.0 It seems that API change in numpy has caused this issue, see https://github.com/numba/numba/issues/6041

I wonder if you plan to upgrade dmipy for the current version of numpy.

Thank you. Ping