Cython module _utils not initialized. Fallback to python.

[stashkov]

Hi Chris!

When I executed import mnnpy I got

~/PyCharm.app/Contents/helpers/pydev/_pydev_bundle/pydev_import_hook.py:19: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  module = self._system_import(name, *args, **kwargs)
Cython module _utils not initialized. Fallback to python.

Does it mean that I don't get any Cython speed benefits? Maybe you have an idea why is this hapening?

[chriscainx]

Yes, this means the cython fails. In this case, the program still work but will take much more time and memory, since it's now running in python. If your sample size is not too big (thousands of cells) it would be acceptable... What is the version you are installing? do you have cython package installed?

[suestring7]

Got a similar warning here. I am installing the latest version mnnpy-0.1.9.5 with the latest cython 0.29.33.

\mnnpy\mnnpy\utils.py:30: NumbaPerformanceWarning: np.dot() is faster on contiguous arrays, called on (array(float32, 1d, A), array(float32, 1d, A))
   dist[i, j] = np.dot(m[i], n[j])
 \mnnpy\mnnpy\utils.py:205: NumbaPerformanceWarning: np.dot() is faster on contiguous arrays, called on (array(float32, 1d, C), array(float32, 1d, A))
   scale = np.dot(working, grad)
 \mnnpy\mnnpy\utils.py:215: NumbaPerformanceWarning: np.dot() is faster on contiguous arrays, called on (array(float32, 1d, C), array(float32, 1d, A))
   curproj = np.dot(grad, curcell)
 Cython module _utils not initialized. Fallback to python.