Tridiagonal Solve: Fix dtype inference

Scipy helper doesn't have special handling for ipiv int32 variable, and always assumes it must be cast to a float64

Tridiagonal Solve: Fix dtype inference
eba75f6f · ricardoV94 · Ricardo Vieira · f4fb4833 · eba75f6f
--- a/pytensor/tensor/_linalg/solve/tridiagonal.py
+++ b/pytensor/tensor/_linalg/solve/tridiagonal.py
@@ -146,9 +146,8 @@ class SolveLUFactorTridiagonal(Op):
                n = nb
        dummy_arrays = [
-            np.zeros((), dtype=inp.type.dtype) for inp in (dl, d, du, du2, ipiv)
+            np.zeros((), dtype=inp.type.dtype) for inp in (dl, d, du, du2, b)
        ]
-        # Seems to always be float64?
        out_dtype = get_lapack_funcs("gttrs", dummy_arrays).dtype
        if self.b_ndim == 1:
            output_shape = (n,)