Conversation
ElliottKasoar
left a comment
ElliottKasoar
left a comment
There was a problem hiding this comment.
Thanks for this! I need to test is properly, but from a first pass of the code it looks great!
I know they're relatively small, but are you happy for me to upload the data start files to the S3 bucket, just to slim things down?
ml_peg/calcs/physicality/oxidation_states/calc_oxidation_states.py
Outdated
Show resolved
Hide resolved
ml_peg/analysis/physicality/oxidation_states/analyse_oxidation_states.py
Outdated
Show resolved
Hide resolved
ml_peg/calcs/physicality/oxidation_states/calc_oxidation_states.py
Outdated
Show resolved
Hide resolved
|
I think the plot titles/axes aren't quite working. Since the plots are unique to each model, it would also be nice to include the model name somewhere in the title, if possible. 
|
PKourtis
force-pushed
the
oxidation_states_clean
branch
from
9bba7fa to
fb343dc
Compare
…ation function - edited to remove unecessary data files
Co-authored-by: Elliott Kasoar <45317199+ElliottKasoar@users.noreply.github.com>
…tor and edited the analysis script accordingly
… skipped if the MD trajs are not present
…e form of outputs / model / {salt}-traj.extxyz
ElliottKasoar
force-pushed
the
oxidation_states_clean
branch
from
642ad10 to
b8e9dc4
Compare
ElliottKasoar
left a comment
ElliottKasoar
left a comment
There was a problem hiding this comment.
Thanks again for this, @PKourtis!
I've rebased and tidied up a little, plus less a couple of suggestions, mainly
| for salt in IRON_SALTS: | ||
| struct_path = data_path / f"{salt}_start.xyz" | ||
| struct = read(struct_path, "0") | ||
| struct.calc = calc |
There was a problem hiding this comment.
| struct.calc = calc | |
| struct.calc = calc | |
| struct.info["charge"] = int(struct.info["charge"]) | |
| struct.info["spin"] = int(struct.info["spin"]) |
Unfortunately when these are read in, ASE seems to set them to numpy.int64, which causes Orb-omol to raise errors
| Name of model use and model. | ||
| """ | ||
| model_name, model = mlip | ||
| model.default_dtype = "float32" |
There was a problem hiding this comment.
This doesn't work for Orb models, as they have float32-high and float32-highest.
For now, I'd probably just check against the name of the model, but in future I will think about a way around this (e.g. defining high vs low precision per model)
| npt.run() | ||
|
|
||
|
|
||
| @pytest.mark.parametrize("mlip", MODELS.items()) |
There was a problem hiding this comment.
I think you can just do "model_name", MODELS if you just want the names
3 participants
Pre-review checklist for PR author
PR author must check the checkboxes below when creating the PR.
Summary
Benchmarks how well models capture the different oxidation states of Iron.
Main tests compares the O-Fe RDFs of Fe+2 and Fe+3 by running 20ps NVT MD for Fe 2Cl and Fe 3Cl ions in water.
Linked issue
Resolves #264
Progress
Testing
Tested on mp0-b3, currently running omol-0 MD.
New decorators/callbacks
Added new option to the plot_scatter decorator to remove markers from the plots.