search

SciML / diffeqpy

Solving differential equations in Python using DifferentialEquations.jl and the SciML Scientific Machine Learning organization
MIT License
508 stars 39 forks source link

diffeqpy GPU isn't automatically installing DiffEqGPU on Collab? #126

Closed ChrisRackauckas closed 8 months ago

ChrisRackauckas commented 8 months ago

https://colab.research.google.com/drive/1bnQMdNvg0AL-LyPcXBiH10jBij5QUmtY?usp=sharing

@LilithHafner do you happen to see what I may have done wrong? It's odd to me that it auto-installs DiffEq and MTK but not DiffEqGPU/CUDA when done the same way.

LilithHafner commented 8 months ago

This runs on diffeqpy.install() (and on initial load in the event that Julia is missing from the system)

https://github.com/SciML/diffeqpy/blob/42ce4d3fe7e6ffd620254b9cff9da4b039eecb9e/diffeqpy/install.jl#L3

and is where we install DiffEq and MTK.

Other packages can be added to that list (but I would not want to install and attempt to precompile Metal.jl or Cuda.jl on hardware that does not support those packages)

ChrisRackauckas commented 8 months ago

ahh I see. Yeah we should make separate diffeqpy.install_cuda(), diffeqpy.install_amdgpu(), etc.