Open
Description
Description
pytensor/pytensor/link/numba/dispatch/scalar.py
Lines 307 to 314 in 594f46b
This was used to try and improve caching of numba functions, but when investigating it, it doesn't seem necessary. Numba is happy to cache the usual way we dispatch functions, with limitations lying elsewhere. Removing the "hack", should make the code a bit tidier.
@numba_funcify.register(Erfc)
def numba_funcify_Erfc(op, **kwargs):
@numba_basic.numba_njit
def erfc(x)
return math.erfc(x)
This is used in more places