[buildkite] Add CI builds on GPUs #113
Conversation
|
test-self-hosted can be commented out |
.buildkite/pipeline.yml
Outdated
| using Pkg | ||
| Pkg.add("CUDA") | ||
| Pkg.add("KernelAbstractions") | ||
| Pkg.add("NLPModels") |
There was a problem hiding this comment.
It depends on how you want to run the tests.
If you just want to use runtests.jl, we can modify it.
But with the current pipeline.yml, we can also run specific and longer tests that target specific GPUs without environment variables.
I did something similar in Krylov.jl, using buildkite to run GPU tests separately from CPU tests, which is recommended to avoid occupying the GPUs for too long, allowing other users and repositories to access them.
.buildkite/pipeline.yml
Outdated
| Pkg.add("NLPModelsJuMP") | ||
| Pkg.add("Percival") | ||
| Pkg.add("PowerModels") | ||
| Pkg.add("MadNLP") |
There was a problem hiding this comment.
not sure if this is ideal. Maybe we can create a separate Project.toml file for testing?
There was a problem hiding this comment.
The best solution is to create one Project.toml for each dependency. We can create a CI build for each extension if we do that.
There was a problem hiding this comment.
Can't we do, say, Project.toml on /test and have Pkg.add("CUDA.jl") and Pkg.add("KernelAbstractions.jl") before project environment instantiation?
There was a problem hiding this comment.
So that we don't need to keep duplicate copies of Project.toml for testing
There was a problem hiding this comment.
We can do that, it's what I do for KrylovPreconditioners.jl:
https://github.com/JuliaSmoothOptimizers/KrylovPreconditioners.jl/blob/main/.buildkite/pipeline.yml
There was a problem hiding this comment.
I think this approach is good. Can we do it in this way?
|
Thanks @amontoison for setting this up! |
@sshin23