numba@jit compilation fail for function "_error_fun_triangulation"

[hejieshi]

Error message:

NumbaWarning:
Compilation is falling back to object mode WITHOUT looplifting enabled because Function "_error_fun_triangulation" failed type inference due to: No implementation of function Function(<function diff at 0x0000023562457EB0>) found for signature:

 >>> diff(array(float64, 3d, C), n=int64, axis=Literal[int](0))

There are 2 candidate implementations:
  - Of which 2 did not match due to:
  Overload in function 'np_diff_impl': File: numba\np\arraymath.py: Line 3531.
    With argument(s): '(array(float64, 3d, C), n=int64, axis=int64)':
   Rejected as the implementation raised a specific error:
     TypingError: got an unexpected keyword argument 'axis'
  raised from C:\Users\DELL\anaconda3\envs\DEEPLABCUT\lib\site-packages\numba\core\typing\templates.py:783

During: resolving callee type: Function(<function diff at 0x0000023562457EB0>)
During: typing of call at C:\Users\DELL\anaconda3\envs\DEEPLABCUT\lib\site-packages\aniposelib\cameras.py (1288)


File "C:\Users\DELL\anaconda3\envs\DEEPLABCUT\lib\site-packages\aniposelib\cameras.py", line 1288:
    def _error_fun_triangulation(self, params, p2ds,
        <source elided>
        # temporal constraint
        errors_smooth = np.diff(p3ds, n=n_deriv_smooth, axis=0).ravel() * scale_smooth
        ^

  @jit(forceobj=True, parallel=True)

 >>> diff(array(float64, 3d, C), int64, Literal[int](0))

Error code:
errors_smooth = np.diff(p3ds, n=n_deriv_smooth, axis=0).ravel() * scale_smooth
After checking, I found that numba's function np_diff_impl does not support the passed axis parameter, because np_diff_impl only supports operations on the last dimension. In order to accommodate np_diff_impl, I changed the code to:
errors_smooth = np.diff(p3ds.transpose(2, 1, 0), n=n_deriv_smooth).transpose(2, 1, 0).ravel() * scale_smooth
However, more errors have been reported. How should I use numba acceleration correctly?
My package version:
aniposelib 0.7.0
numba 0.61.0