numba nopython mode 不支持numpy.int類型

一則小記，numba庫並不支持NumPy的int類型，需要指定整型的位寬，使用int32和int64都沒有問題。測試代碼如下

本機系統Ubuntu 18.04, Python 3.6.8, numba 0.46.0, NumPy 1.16.4

import numba
import numpy as np
import sys

@numba.jit(nopython=True)
def test_zeros(H, W):
    # array = np.zeros((H, W, 3), dtype=np.int) # Will cause jit error.
    array = np.zeros((H, W, 3), dtype=np.int64) # OK.
    array = np.zeros((H, W, 3), dtype=np.int32) # OK.

def main():
    test_zeros(100, 200)
    return 0

if __name__ == "__main__":
    sys.exit(main())

上述代碼報出的錯誤如下

numba.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Invalid use of Function(<built-in function zeros>) with argument(s) of type(s): ((int64, int64, Literal[int](3)), dtype=Function(<class 'int'>))
 * parameterized
In definition 0:
    All templates rejected with literals.
In definition 1:
    All templates rejected without literals.
This error is usually caused by passing an argument of a type that is unsupported by the named function.
[1] During: resolving callee type: Function(<built-in function zeros>)
[2] During: typing of call at TestNumba.py (8)


File "TestNumba.py", line 8:
def test_zeros(H, W):
    array = np.zeros((H, W, 3), dtype=np.int) # Will cause jit error.