just embrace the numba and add it in the test...

Author: haruishi43

.github/workflows/ci.yml

@@ -53,6 +53,7 @@ jobs:
       run: |
         pip install pytest
         pip install torch torchvision
+        pip install numba
         pytest tests
 
     # - name: Install from dist

tests/equi2pers/numpy_inv.py

@@ -5,11 +5,7 @@
 
 from timeit import timeit
 
-try:
-    from numba import jit
-except ImportError:
-    print("numba not available")
-    jit = None
+from numba import jit
 
 import numpy as np
 
@@ -59,7 +55,7 @@ def hdinv(A):
     return invA
 
 
-# @jit("float64[:,:](float64[:,:])", cache=True, nopython=True, nogil=True)
+@jit("float64[:,:](float64[:,:])", cache=True, nopython=True, nogil=True)
 def fast_inverse(A):
     inv = np.empty_like(A)
     a = A[0, 0]
@@ -89,7 +85,7 @@ def fast_inverse(A):
     return inv
 
 
-# @jit(cache=True, nopython=True, nogil=True)
+@jit(cache=True, nopython=True, nogil=True)
 def vecinv(A):
     invA = np.zeros_like(A)
     for i in range(A.shape[0]):

tests/grid_sample/numpy/nearest.py

@@ -4,11 +4,7 @@
 
 """
 
-try:
-    from numba import njit
-except ImportError:
-    print("numba not available")
-    njit = None
+from numba import njit
 
 import numpy as np
 
@@ -80,7 +76,7 @@ def faster_nearest(
     return out
 
 
-# @njit
+@njit
 def run(img, grid, out, b, h, w):
     for i in range(b):
         for y_out in range(h):