Closed MARMOTatZJU closed 4 years ago
More details:
I have tried versions including 0.39.0 / 0. 42.0 0.45.0. None of them fixed this problem
expected_overlaps = np.zeros((max_len), np.float32)
@jit(nopython=True)
def calculate_expected_overlap(fragments, fweights):
max_len = fragments.shape[1]
expected_overlaps = np.zeros((max_len), np.float32)
expected_overlaps[0] = 1
# TODO Speed Up
for i in range(1, max_len):
mask = np.logical_not(np.isnan(fragments[:, i]))
if np.any(mask):
fragment = fragments[mask, 1:i + 1]
seq_mean = np.sum(fragment, 1) / fragment.shape[1]
expected_overlaps[i] = np.sum(seq_mean * fweights[mask]) / np.sum(
fweights[mask])
return expected_overlaps
A simple solution: comment the jit line
# @jit(nopython=True)
This solve the problem
The bug has been fixed since 0.47.0.
I encounter the following problem and hereby I report it to Numba Dev. team.
Awaiting for your reply and solution. Thanks in advance!