Cannot Precompile DifferentialEquations.jl #683
Comments
|
Getting the same error with tags v6.14.0 and v6.15.0 with Julia 1.5.2. Seemed to work fine with Julia 1.4 for me. UPDATE: Confirming that both versions precompile fine with Julia 1.4.2 but not with 1.5.2. |
|
I'm using Julia 1.4.0 right now, if that helps |
|
Looks like we may have had a versioning issue in the latest release. Can you share |
@DilumAluthge I am not quite sure how to use RegistryTools.jl to do this, so I did it by hand. Could this be checked? We missed an important lower bound: SciML/DifferentialEquations.jl#683 SciML/ModelingToolkit.jl#601
|
I can't recreate the error locally but I know that it's due to this missing lower bound JuliaRegistries/General#22615 and so we'll get the correct version bound amended into the registry. If you do |
|
It's still not working for me. I tried ]up and that didn't seem to solve it. Here's my ]st list. [7d9fca2a] Arpack v0.4.0 |
|
Hmm, can I get |
|
Here it is: [c3fe647b] AbstractAlgebra v0.10.0 |
|
If you do |
|
That seems to conflict with DataStructures and CUDA |
|
Oh yes, the later versions of DiffEqBase want DataStructures v0.18 which is what's keeping it back. So the ModelingToolkit bound will fix this by just not giving you the latest version, but the real offender here is whatever other package has an upper bound on the core DataStructures library... that just is going to cause issues. It's interesting you're on CUDA 0.1 though: is that by choice? CUDA.jl released its 2.0, which should work fine here. |
|
JuMP v0.21.4 also has such a bound and should be v0.21.5. This is all saying that there's something deeper in your system that is causing CUDA, JuMP, DiffEqBase, etc. to all be held back. |
|
Oh: [3a865a2d] CuArrays v2.2.2
[8faf48c0] NeuralNetDiffEq v1.6.0You might want to remove those, since those were renamed CUDA.jl and NeuralPDE.jl respectively, and those old versions may be causing your issue. |
|
Thank you very much! Is there a good way to identify what packages have upper bounds on other packages? |
@DilumAluthge I am not quite sure how to use RegistryTools.jl to do this, so I did it by hand. Could this be checked? We missed an important lower bound: SciML/DifferentialEquations.jl#683 SciML/ModelingToolkit.jl#601
|
🤷 Add and find out. At this point I just follow a few thousand packages and tend to know, but that's obviously not a scalable solution. Take that to the Discourse though: for now, I'm closing this since the precompilation issue is fixed by appropriate version bounding: |
Hello,
When I try to precompile DifferentialEquations.jl I get the following error:
"ERROR: LoadError: LoadError: UndefVarERror: AbstractADType not defined
ERROR: LoadError: Faield to precompile ModelingToolkit"
Do you have any suggestions for how to deal with this?
Thank you.
The text was updated successfully, but these errors were encountered: