3 ms·
* I haven't looked into it myself, so I can't really address your specific question about using fast/optimized libraries such as BLAS. This doesn't (seem to) d
by abathur 7y ago
* I haven't looked into it myself, so I can't really address your specific question about using fast/optimized libraries such as BLAS.
This doesn't (seem to) distinguish your system python (i.e., the ones that'll run in a default execution environment if you run "python" or "python3") and project-specific pythons.
For the latter, I guess it's common to use something like virtualenv or a container/VM to isolate this python instance from your system Python.
Personally, I use Nix (the package manager and corresponding packaging language) for declaring both my python system and project environments.
Nix still takes some work to ease into, but I think your question suggests enough understanding/need/drive to push through this. Costrouc (https://discourse.nixos.org/u/costrouc https://discourse.nixos.org/u/costrouc) may be a good person to ask; I can recall him making a number of posts about data science on Nix. I also noticed this one, which seems related to what you're looking into: https://discourse.nixos.org/t/benchmarking-lapack-blas-libraries-and-compilers-with-nix/1176/3 https://discourse.nixos.org/t/benchmarking-lapack-blas-libra...