Nick

Some more clues as to the source of this issue: The first `seq_len -1` samples are always zero (which should perhaps be the title of the issue): ```python #!/usr/bin/env python3...

@relf : I've attempted a fix, but this one is beyond my powers and everything I've tried thus far makes the problem worse . . .

> Since we now support non-integer values, what is the best approach to generate a new test dataset? Does anyone know of a reliable source for non-integer hypergeometric PMF values?...

@timholy : Just to give additional context: This ask arose from an optimization routine where the matrix construction and decomposition is an implementation detail; the objective function F(x) can have...

I have analyzed this is a bit more depth. If I have a virtualenv `deleteme`, then I can fix this issue by copying the `pybind11/share/cmake` into `lib`: ```bash (deleteme) ➜...

@henryiii : Yes, both of those suggestions work, but the first is not available for users on HPC systems (which is why I'm putting `pybind11` in a virtualenv in the...

> The pybind11[global] option does exactly that. aaand I'm a moron. Sorry for the noise. Validated that the syntax `Requires-Dist: pybind11[global] (>= 2.7.1)` works in wheels as well. While we're...

Note that the generalized eigenvalue problem $Ax = \lambda Bx$ _is_ solved in this package, at least for $B \succ 0$. Example code: ```julia using GenericLinearAlgebra: eigen, cholesky, Diagonal, Symmetric...

I have attempted to implement this. The code is still rather inelegant: ```python #!/usr/bin/env python3 import argparse from pathlib import Path import sys from packaging.utils import parse_wheel_filename import platform def...

So the recommended diff is? > the bounds given must bracket a ~~single~~ root. That seems ok. FWIW, we do have a [cubic solver](https://github.com/boostorg/math/blob/f7e6597a6044bd2a2ba2900479c550bee0fda337/doc/roots/cubic_roots.qbk#L4).