[WIP] Integration for CP2K #383

nwinner · 2020-03-15T03:25:45Z

Summary

Integration of CP2K into the atomate framework. This WIP PR is here so that maintainers can have a sense of what I'm proposing to add.

New firetasks, fireworks, workflows, and database/drone classes for cp2k. Meant to be used in conjunction with the new pymatgen.io.cp2k module (also a work in progress).

TODO (if any)

More workflows/fireworks are needed. There have also been a few issues with gluing tasks with file exchange. The drone also needs to include the optional Dos, bader, and cube parsing that will be needed.

…ooks like. Still working on making it a more seemless integration with pymatgen, but it seems to be close to functional at least.

avoid using reservation fws.

…instead of reservation, then we can just specify the number of images there will be in each calculation and the env variable.

…d happen if starting a difficult-to-converge calculation from a previous WAVECAR) then it needs to handle bytes instead of strings for the file contents.

testing, I need to push it so I can test on cluster a bit.

…. This seems to be the first version that runs everything (create workflow, run, stich together, push to database). There will sitll be testing that needs to be done though.

…ross different Cp2k executions could be problematic. Probably the best way to deal with it is to assert a universal project name with each workflow. I don't like this, because its not as flexible, but it fixes the problem.

successfully. Might have to decide on universal file naming in order to avoid these problems in the future.

utf

Hi @nwinner Thank you for submitting this!

I know this is a work in progress but I made some comments on the PR so we can hopefully clean this up as you go along.

Would you also be able to run the cp2k module through the black python formatter? Some of your lines are rather long so this will enforce PEP8 formatting of your code.

You can install black from pip:

pip install black

And run it on the cp2k module using a max line length of 80 (presuming you are in the atomate root directory) via:

black atomate/cp2k -l 80

utf · 2020-03-17T21:32:00Z

atomate/cp2k/database.py

+logger = get_logger(__name__)
+
+
+class Cp2kCalcDb(CalcDb):


Is this different to VASPCalcDb at all? If not, we should probably rename VASPCalcDb something more general and then it can be used in the CP2K module also.

Its not different at all I think, a more general VASPCalcDb would be a good idea.

utf · 2020-03-17T21:34:45Z

atomate/cp2k/drones.py

+from __future__ import division, print_function, unicode_literals, absolute_import
+
+"""
+This Drone tries to produce a more sensible task dictionary than the default VaspToDbTaskDrone.


You can remove these comments as they aren't relevant.

utf · 2020-03-17T21:36:13Z

atomate/cp2k/drones.py

@@ -0,0 +1,463 @@
+# coding: utf-8
+
+from __future__ import division, print_function, unicode_literals, absolute_import


Atomate does not support python 2 any more, so you can remove all the __future__ imports from all of the cp2k module files.

utf · 2020-03-17T21:38:37Z

atomate/cp2k/drones.py

+
+    def post_process(self, dir_name, d):
+        """
+        Post-processing for various files other than the vasprun.xml and OUTCAR.


Can you update this docstring.

utf · 2020-03-17T21:46:26Z

atomate/cp2k/firetasks/parse_outputs.py

+                pass
+            elif defuse_unsuccessful == "fizzle":
+                raise RuntimeError(
+                    "VaspToDb indicates that job is not successful "


Update VaspToDb

utf · 2020-03-17T21:55:51Z

atomate/cp2k/fireworks/core.py

+            t.append(UpdateStructureFromPrevCalc(prev_calc_loc=prev_calc_loc))
+
+        # if prev calc directory is being REPEATED, copy files
+        if prev_calc_dir:


It's not clear to me how prev_calc_loc and prev_calc_dir are interacting here? The docstring for previous_calc_dir says:

prev_calc_dir (str): Path to a previous calculation to copy from.

but that doesn't happen in the code. Can you clarify how prev_calc_loc and prev_calc_dir should be used?

utf · 2020-03-17T22:29:41Z

atomate/cp2k/firetasks/write_inputs.py

+from __future__ import division, print_function, unicode_literals, absolute_import
+
+"""
+This module defines tasks for writing vasp input sets for various types of vasp calculations


Update please.

utf · 2020-03-17T22:38:10Z

atomate/cp2k/firetasks/write_inputs.py

+@explicit_serialize
+class WriteCp2kFromPrevious(FiretaskBase):
+
+    optional_params = ['cp2k_input_params', 'prev_calc_loc', 'original_input_filename',


cp2k_input_params and new_input_filename are unused here.

utf · 2020-03-17T22:43:53Z

atomate/cp2k/firetasks/glue_tasks.py

+class UpdateStructureFromPrevCalc(FiretaskBase):
+    """
+    Using the location of a previous calculation. The CP2K output parser will
+    get the final structure from a previous calculation and update this FW's


Is there a reason why you are updating the structure by updating the firework spec rather than either:

Just writing the structure to file

Doing the structure updating as part of WriteCp2kFromPrevious.

I think option 2 makes the most sense. If you only want to copy the structure and nothing else from the previous calculation, you could add options to that effect in WriteCp2kFromPrevious.

Hey Alex, number 2 probably does make more sense. Number 1 isn't the case because cp2k doesn't use the same formatting as VASP where a separate file holds the structure. Generally cp2k has everything in a single input file. This can actually be modified, but its not a cp2k handles stuff by default so i didn't want to complicate that until a later time.

Ok great. Let's go with option 2!

utf · 2020-03-17T22:44:19Z

atomate/cp2k/fireworks/core.py

+        fw_name = "{}-{}".format(structure.composition.reduced_formula if structure else "unknown", name)
+
+        cp2k_input_set = cp2k_input_set or StaticSet(structure, **cp2k_input_set_params)
+        cp2k_input_set.project_name = name  # Currently, this assertion is needed to keep all FWs in a workflow


What is the problem you are trying to solve by using project_name?

The CP2K output files are named according to the "PROJECT_NAME" variable inside the input file. Notably, the wavefunction and DOS files have the project name prepended in their filename. So, in order to deal with this we would need to either (a) check for the project name every time you try to find a file (like a restart file) to copy over, (b) use a universal project name so that the files still have the same format but with a standard prefix, (c) override the file name in the cp2k input file so that it adheres to some agreed upon overall file name. I chose to go with (b) as the easiest one, but I'm open to changing it.

Note we're going to have the same question for my (perpetual backburner project) of CASTEP support, so interested in whatever you settle on. I might pick the seed name to be the firework id, unless that has issues I'm not anticipating.

Ok, thanks for clarifying that. With option b) are you using the same project name for EVERY calculation or is it specific to the calculation type? Maybe what I am describing is option c?

Are there any downsides to using a fixed project name, and just doing the calculations in different directories? That way there is never any ambiguity over which files to load. it seems like this would remove a lot of the complexity and make the parsing and file copying more similar to how it is handled in the VASP module?

Currently option (b) uses the same name for the workflow. So you could name the files according to "Static" for static WFs as an example. Just to give you a sense of what the output files are like. If you did that you would get a directory like this (for Si):

launcher-X:

Static-ALPHA_k1-1.pdos # elemental dos for atom kind 1, Spin up channel

Static-BETA_k1-1.pdos # elemental dos for atom kind 1, Spin down channel

Static-ALPHA_list1-1.pdos # Site projected dos onto site list 1, spin up

Static-ALPHA_list2-1.pdos # Site projected dos onto site list 2, spin up

Static-BETA_list1-1.pdos # Site projected dos onto site list 1, spin down

Static-BETA_list2-1.pdos # Site projected dos onto site list 2, spin down

Static-RESTART.wfn # Restart wavefunction file

Static-RESTART.wfn.bak-1 # Backup wavefunction file (dumped intermittently)

Static-v_hartree-1_0.cube. # Hartree (electrostatic potential)

Static-ELECTRON_DENSITY-1_0.cube # Electron density

Static-SPIN_DENSITY-1_0.cube # spin density

So you can see that it changes the naming, but its pretty repeatable. Something I've been doing in the pymatgen output parsing is globbing to find files containing, say, "pdos", to identify the pdos files and globbing for "RESTART" to get the wavefunction file name. We could think about doing this instead of trying to record what the file name is going to be.

Drones needs to use num_sites of structure object.

of of Danny Broberg's implementation for VASP.

utils: return entire scaling matrix

implemented properly.

nwinner added 30 commits April 22, 2019 15:49

Fixed import.

c856382

A lot of clean-up. Had not been addressed in a while by the devs it l…

8ff38d1

…ooks like. Still working on making it a more seemless integration with pymatgen, but it seems to be close to functional at least.

Added lammps_input_set so that atom_style can be retained as a variable

87ea810

Playing with the NEB workflow. Might try to figure out a way to

29d7e94

avoid using reservation fws.

Currently thinking it would be best to use env_chk with vasp_neb_cmd …

6c845e6

…instead of reservation, then we can just specify the number of images there will be in each calculation and the env variable.

fix

02776ff

fix

68e5465

Minor correction to glue_tasks. If reading in the WAVECAR (which coul…

2600dca

…d happen if starting a difficult-to-converge calculation from a previous WAVECAR) then it needs to handle bytes instead of strings for the file contents.

Lammps should support env_chk

f294cf3

Lammps should support env_chk

2665de6

Exploring a new firetask.

c2a17fa

Debug.

6c3872d

Lammps should support env_chk

5108d19

Trying something different. Adding a seperate file for my user task.

21b9fb2

Added a brief function to transmute after relaxing.

17a466a

Working on this new task.

9b1cd17

Working on Lammps run_calc

914b9cd

Debug

7005625

Lammps Run

1d2e29e

Working on it.

e3d670f

debug

748e0c9

Lammps should support env_chk

9bc9563

Lammps should support env_chk

e46e77a

Debug

31830ba

debug

39d04a7

debug

33e6987

debug

e298e49

debug

d71f00c

Debug.

f76d774

First design of cp2k module for atomate.

b087ea2

nwinner added 8 commits February 23, 2020 14:09

First commit for the cp2k module in atomate. Pretty rough, but to start

6262805

testing, I need to push it so I can test on cluster a bit.

Continuing testing and refinement. Added database and drone functions…

d22d72d

…. This seems to be the first version that runs everything (create workflow, run, stich together, push to database). There will sitll be testing that needs to be done though.

Wavefunction files are bytes files. File copying needs to handle that.

9168197

File copying is now to the point where things will at least run

ab1bd74

successfully. Might have to decide on universal file naming in order to avoid these problems in the future.

Merge with master

7051dd7

Merge https://github.com/hackingmaterials/atomate into cp2k

10f846e

user_tasks.py not needed, now in MPMorph

229a187

utf reviewed Mar 17, 2020

View reviewed changes

nwinner added 17 commits March 18, 2020 13:19

Database was missing kwargs, led to errors.

808cb60

Drones needs to use num_sites of structure object.

Cleanup.

58f2152

Beginning to integrate the ability to do defect workflows by building

361cad5

of of Danny Broberg's implementation for VASP.

Debugging last commit on the cluster.

c0e7acd

Minor refinements

226eabe

Minor refinements

0eaded2

Typo.

b8867c5

Testing from very simple cell_opt wfs on the cluster.

2dd859a

Database manipulation for CalcDb task

b801820

Missing import statement.

0bac0d9

Drones: fft can only apply to array, not array of arrays

f0cc95a

utils: return entire scaling matrix

Incorrect hartree file parsing. Works now, but might change again later.

6edb5db

Bug.

c67165e

Corrections to the drone assimilation.

6d9c4e1

Written incorrectly. It was working because the underlying sets were not

2b3b579

implemented properly.

Fixing write inputs.py.

60c15a9

Quick commit. Saving state.

4201338

Base automatically changed from master to main March 10, 2021 18:03

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] Integration for CP2K #383

[WIP] Integration for CP2K #383

nwinner commented Mar 15, 2020

utf left a comment

utf Mar 17, 2020

nwinner Mar 18, 2020

utf Mar 17, 2020

utf Mar 17, 2020

utf Mar 17, 2020

utf Mar 17, 2020

utf Mar 17, 2020

utf Mar 17, 2020

utf Mar 17, 2020

utf Mar 17, 2020

nwinner Mar 18, 2020

utf Mar 18, 2020

utf Mar 17, 2020

nwinner Mar 18, 2020

mkhorton Mar 18, 2020

utf Mar 18, 2020

nwinner Mar 18, 2020

		@@ -0,0 +1,463 @@
		# coding: utf-8

		from __future__ import division, print_function, unicode_literals, absolute_import

		logger = get_logger(__name__)


		class Cp2kCalcDb(CalcDb):

[WIP] Integration for CP2K #383

Are you sure you want to change the base?

[WIP] Integration for CP2K #383

Conversation

nwinner commented Mar 15, 2020

Summary

TODO (if any)

utf left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment