Casper da Costa-Luis
Casper da Costa-Luis
currently part of #196 :)
deprioritising in favour of #2198
I'd prefer GHA cached art[ie]facts (e.g. https://github.com/dask/distributed/pull/6855) rather than committing lockfiles.
thanks @purepani; I can fix this after #2148 & #2145. Just a quick note for now - could you use `pybind11` in lieu of `nanobind`? In future we will need...
As per #2132 (and #1912 #1126 #572 etc.) we're looking to support arbitrary (non-numpy) arrays, so buffer protocol is safest. `nanobind` is a stripped-down (quicker to compile) fork of `pybind11`...
I'm uncertain; have you tried casting a custom (non-numpy/cupy etc) object (such as the [`pybind11::class_` example](https://pybind11.readthedocs.io/en/stable/advanced/pycpp/numpy.html#buffer-protocol)) to a `nanobind::array`? Last time I checked it didn't work.
Just tested in https://github.com/AMYPAD/NumCu/pull/9 and it looks like `nanobind` works but needs: - `python>=3.8` - no `-Wpedantic`