graph building and file I/O with Xtals.jl #112
Conversation
If this ends up working well, I may just PR that function to Xtals.jl as it could definitely be useful for others, but going to play with it here for now. Signed off by Rachel Kurchin [email protected]
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
Codecov Report
@@ Coverage Diff @@
## main #112 +/- ##
==========================================
+ Coverage 86.97% 87.42% +0.44%
==========================================
Files 17 17
Lines 476 493 +17
==========================================
+ Hits 414 431 +17
Misses 62 62
Continue to review full report at Codecov.
|
Signed-off-by: Rachel Kurchin <[email protected]>
More details: * Made separate dispatch of build_graph function onto a Crystal object; this should be directly usable as the first layer in a model and differentiate with respect to c.atoms.coords.xf (or convert to Cartesian first) * ASE dependency is no more! The cost of this was loss of compatibility with some structure file formats. Some of these (most notably XYZ) will be recoverable with a pymatgen update. * I removed the normalize_weights option, because I can't conceive of a case where you wouldn't want to do this * I also removed the test for nonperiodicity because I realized this is not generally stored in the structure files but rather in the ASE Atoms object (which has certain defaults when reading in files), so it wasn't an actual information stream Signed off by Rachel Kurchin [email protected] 25 August 2021
Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
Co-authored-by: Anant Thazhemadam <[email protected]>
…k for cutoff radius size, change from loops to list comprehensions for neighbor_list Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Anant Thazhemadam <[email protected]>
… a better longterm solution though Signed-off-by: Rachel Kurchin <[email protected]>
Signed-off-by: Rachel Kurchin <[email protected]>
| @@ -58,7 +58,7 @@ using Xtals | |||
| ag2 = deserialize(abspath(@__DIR__, "..", "test_data", "strucs", "testgraph.jls")) | |||
| @test adjacency_matrix(ag.graph) == adjacency_matrix(ag2.graph) | |||
| @test elements(ag) == ag2.elements | |||
| @test Atoms.normalized_laplacian(ag) == ag2.laplacian | |||
| @test ChemistryFeaturization.Atoms.normalized_laplacian(ag) == ag2.laplacian | |||
There was a problem hiding this comment.
Could you please expand on what exactly the namespace conflict is? I don't recall seeing any such conflicts while testing with the OFD PR.
It might also help to actually make a note of it for future reference.
There was a problem hiding this comment.
it's because Xtals also defines Atoms so while this worked before:
using ChemistryFeaturization.Atoms
# then later...
Atoms.whatever()
that's now ambiguous as to which Atoms it's referring to, so now I just have
using ChemistryFeaturization
# then later...
ChemistryFeaturization.Atoms.stuff...
There was a problem hiding this comment.
Ah, I see. In that case, I can't think of any other long term solution that's better than what we have rn.
There was a problem hiding this comment.
yeah I mean short of renaming our Atoms module to something else there isn't really another option
| - `max_num_nbr::Integer=12`: maximum number of neighbors to include (even if more fall within cutoff radius) | ||
| - `dist_decay_func::Function=inverse_square`: function to determine falloff of graph edge weights with neighbor distance | ||
| """ | ||
| function build_graph( |
There was a problem hiding this comment.
Could we have a reference to this docstring also under Utilities/Graph Building?
There was a problem hiding this comment.
shouldn't the @autodocs do that?
There was a problem hiding this comment.
I can't check right now because my local setup for building the docs is somehow broken again :/ but I thought the @autodocs block should do it?
|
|
||
| Returns as is, js, dists to be compatible with ASE's output format for the analogous function. | ||
| """ | ||
| function neighbor_list(crys::Crystal; cutoff_radius::Real = 8.0) |
There was a problem hiding this comment.
Again, can we have a reference to this docstring in the docs? 😄
Signed-off-by: Rachel Kurchin <[email protected]>
Prerequisite for #10 but won't resolve it on its own. Will also help #111 go more smoothly.
Need to actually implement the neighbor list function, which will be the most intensive part of this. Not really started yet, but I have a clear picture of how it will go and it's sketched out in some comments...