PyCall → PythonCall #38
Conversation
Add handling for logical fallacies in MACE@v0.3.3
|
I've decided to move this to a separate package instead to avoid requiring Python in NQCModels.jl. Instead, this PR now makes the ASE and ASE friction interfaces usable with PythonCall.jl to fit in the structure of using PythonCall and CondaPkg.jl for packaging Python dependencies. |
| @@ -42,7 +42,12 @@ jobs: | |||
| run: julia -e 'using Pkg; Pkg.Registry.add(RegistrySpec(url="https://github.com/JuliaMolSim/MolSim"))' | |||
|
|
|||
| - uses: julia-actions/julia-buildpkg@v1 | |||
| env: | |||
| LD_LIBRARY_PATH: $HOME/libxc-5.1.5/lib | |||
There was a problem hiding this comment.
Why is there a libxc dependency here? this only seems to be in the CI. does anything in NQCModels or its tests require that?
There was a problem hiding this comment.
test/ase_pycall.jl:17-22 uses it to test the ASE calculator interface with GPAW. I couldn't find a working conda package for GPAW, or Id've changed the test mechanism to install it through CondaPkg.jl and make the tests work on any machine without having to set up ase and GPAW first.
|
|
||
| @testset "EMT" begin | ||
| emt = pyimport("ase.calculators.emt") | ||
| h2.calc = emt.EMT() |
There was a problem hiding this comment.
just a note: it used to be that calculators require setting with atoms.set_calculator(calc). EMT might be a rare exception where you can directly set it like this
Update:
I've decided to move this to a separate package instead to avoid requiring Python in NQCModels.jl.
Instead, this PR now makes the ASE and ASE friction interfaces usable with PythonCall.jl to fit in the structure of using PythonCall and CondaPkg.jl for packaging Python dependencies.
Avoid ase calculator and as many Julia -> Python -> Torch pipelines as reasonably possible without re-writing MACE for Julia.