High throughput wc #147
High throughput wc #147
Conversation
federicazanca
force-pushed
the
high-throughput-wc
branch
from
60c0a08 to
41fb3d2
Compare
|
UPDATES The high throughput |
| wg = WorkGraph("hts_workflow") | ||
|
|
||
| wg.add_task( | ||
| run_opt_calc, name="opt_task", folder=folder_path, janus_opt_inputs=inputs |
There was a problem hiding this comment.
Could this be generalised to take an arbitrary task using one of the function's parameters, so it wouldn't have to be hardcoded?
| folder_path = Path("/home/federica/aiida-mlip/tests/workflows/structures/") | ||
| inputs = { | ||
| "model": ModelData.from_local( | ||
| "/home/federica/aiida-mlip/tests/data/input_files/mace/mace_mp_small.model", | ||
| architecture="mace_mp", |
There was a problem hiding this comment.
Paths should be generalised
| ase = "^3.23.0" | ||
| voluptuous = "^0.14" | ||
| janus-core = "^v0.6.0b0" | ||
| janus-core = "^v0.6.3b0" |
There was a problem hiding this comment.
Might need to pin or check compatibility with newer versions
| # AT THE MOMENT WE ONLY CHECK THE PROCESS IS CREATED AT LEAST, | ||
| # WHEN WE FIX THE SUBMISSION THIS NEEDS TO BE CHANGED | ||
|
|
||
| assert wg.state == "CREATED" | ||
| # wg_node = load_node(wg.pk) | ||
| # assert isinstance(wg_node.outputs.opt_structures.h2o, StructureData) |
| "code": janus_code, | ||
| } | ||
| wg = HTSWorkGraph(folder_path=structure_folder2, inputs=inputs) | ||
| wg.wait(15) |
There was a problem hiding this comment.
Are there nicer ways of doing this?
|
|
||
| wg.to_html() | ||
|
|
||
| wg.max_number_jobs = 10 |
There was a problem hiding this comment.
Two small suggestions:
- set
max_number_jobsas an input parameter. - Not ot submit the wg inside the function, but return the
wg, and let the user runwg.submit(). In this case, user can also modify the inputs of the tasks or other settings before submission.
|
|
||
| wg.group_outputs = [{"name": "opt_structures", "from": "opt_task.final_structures"}] | ||
|
|
||
| wg.to_html() |
There was a problem hiding this comment.
will this work in the notebook? if not, maybe delete this line.
| wg.to_html() |
| [tool.poetry.plugins."aiida.workflows"] | ||
| "mlip.hts_wg" = "aiida_mlip.workflows.hts_workgraph:HTSWorkGraph" | ||
|
|
There was a problem hiding this comment.
We are working on a new entry point for the WorkGraph. Will let you know the syntax when the API is fixed.
Still very messy, just making the PR so we can actually see the code when discussing.
Ignore the comments in the model file, will remove and use them to fix other issues.
I made a
WorkChainfor submitting high-throughput screening, but I am facing the problems described in #146I also added the submission script for submitting the actual
Workchain(in examples/workflows)And a script to run high-throughput screening without writing a
WorkChain. We might end up only keeping this? Depending on discussion on how to implement some things