I am using Numba to speed up a series of functions as shown below. if I set the step_size
variable in function PosMomentSingle
to a float (e.g. step_size = 0.5
), instead of an integer (e.g step_size = 1.0
), I get the following error:
Cannot unify array(float32, 1d, C) and array(float64, 1d, C) for 'axle_coords.2', defined at <ipython-input-182-37c789ca2187> (12)
File "<ipython-input-182-37c789ca2187>", line 12:
def nbSimpleSpanMoment(L, axles, spacings, step_size):
<source elided>
while np.min(axle_coords) < L:
I found it quite hard to understand what the problem is, but my guess is there is an issue with the function after @jit
(nbSimpleSpanMoment
), with some kind of a datatype mismatch. I tried setting all variables to float32, then to float64 (e.g. L = np.float32(L)
) but whatever I try creates a new set of errors. Since the error message is quite cryptic, I am unable to debug the issue. Can someone with numba experience explain what I am doing wrong here?
I placed my code below to recreate the problem.
Thank you for the help!
import numba as nb
import numpy as np
@nb.vectorize(nopython=True)
def nbvectMoment(L,x):
if x<L/2.0:
return 0.5*x
else:
return 0.5*(L-x)
@nb.jit(nopython=True)
def nbSimpleSpanMoment(L, axles, spacings, step_size):
travel = L + np.sum(spacings)
maxmoment = 0
axle_coords = -np.cumsum(spacings)
moment_inf = np.empty_like(axles)
while np.min(axle_coords) < L:
axle_coords = axle_coords + step_size
y = nbvectMoment(L,axle_coords)
for k in range(y.shape[0]):
if axle_coords[k] >=0 and axle_coords[k] <= L:
moment_inf[k] = y[k]
else:
moment_inf[k] = 0.0
moment = np.sum(moment_inf * axles)
if maxmoment < moment:
maxmoment = moment
return np.around(maxmoment,1)
def PosMomentSingle(current_axles, current_spacings):
data_list = []
for L in range (1,201):
L=float(L)
if L <= 40:
step_size = 0.5
else:
step_size = 0.5
axles = np.array(current_axles, dtype='f')
spacings = np.array(current_spacings, dtype='f')
axles_inv = axles[::-1]
spacings_inv = spacings[::-1]
spacings = np.insert(spacings,0,0)
spacings_inv = np.insert(spacings_inv,0,0)
left_to_right = nbSimpleSpanMoment(L, axles, spacings, step_size)
right_to_left = nbSimpleSpanMoment(L, axles_inv, spacings_inv, step_size)
data_list.append(max(left_to_right, right_to_left))
return data_list
load_effects = []
for v in range(14,31):
load_effects.append(PosMomentSingle([8, 32, 32], [14, v]))
load_effects = np.array(load_effects)
from Numba data type error: Cannot unify array
No comments:
Post a Comment