aiida-workgraph 0.2.2

Design flexible node-based workflow for AiiDA calculation.

AiiDA-WorkGraph

Provides the third workflow component: WorkGraph, to design flexible node-based workflows using AiiDA.

In AiiDA, there are two workflow components: workfunction and WorkChain. Workfunction is easy to implement but it does not support automatic checkpointing, which is important for long-running calculations. Workchain supports automatic checkpointing but it is difficult to implement and also not as flexible as the workfunction. AiiDA-WorkGraph provides the third component: WorkGraph. It is easy to implement and supports automatic checkpointing. It is also flexible and can be used to design complex workflows.

Here is a detailed comparison between the WorkGraph with two AiiDA built-in workflow components.

Aspect	WorkFunction	WorkChain	WorkGraph
Use Case	Short-running jobs	Long-running jobs	Long-running jobs
Checkpointing	No	Yes	Yes
Execution order	Sequential	Hybrid Sequential-Parallel	Directed Acyclic Graph
Non-blocking	No	Yes	Yes
Implementation	Easy	Difficult	Easy
Dynamic	No	No	Yes
Ready to Use	Yes	Need PYTHONPATH	Yes
Subprocesses Handling	No	Launches & waits	Launches & waits
Flow Control	All	if, while	if, while, match
Termination	Hard exit	ExitCode	ExitCode
Data Passing	Direct passing	Context	Link & Context
Output Recording	Limited support	Out & validates	Out
Port Exposing	Limited support	Manual & automatic	Manual

Installation

    pip install aiida-workgraph

To install the latest version from source, first clone the repository and then install using pip:

git clone https://github.com/superstar54/aiida-workgraph
cd aiida-workgraph
pip install -e .

In order to use the widget, you also need to run:

cd aiida_workgraph/widget/
npm install
npm run build

Documentation

Check the docs and learn about the features.

Examples

Suppose we want to calculate (x + y) * z in two steps. First, add x and y, then multiply the result with z.

from aiida.engine import calcfunction

# define add calcfunction
@calcfunction
def add(x, y):
    return x + y

# define multiply calcfunction
@calcfunction
def multiply(x, y):
    return x*y

# Create a workgraph to link the nodes.
wg = WorkGraph("test_add_multiply")
wg.nodes.new(add, name="add1")
wg.nodes.new(multiply, name="multiply1")
wg.links.new(wg.nodes["add1"].outputs[0], wg.nodes["multiply1"].inputs["x"])

Prepare inputs and submit the workflow:

from aiida_workgraph import WorkGraph
from aiida import load_profile

load_profile()

wg.submit(inputs = {"add1": {"x": 2, "y": 3}, "multiply1": {"y": 4}}, wait=True)
print("Result of multiply1 is", wg.nodes["multiply1"].outputs[0].value)

Start the web app, open a terminal and run:

workgraph web start

Then visit the page http://127.0.0.1:8000/workgraph, you should find a first_workflow Worktree, click the pk and view the WorkGraph.

One can also generate the node graph from the process:

verdi node generate pk

Development

Pre-commit and Tests

To contribute to this repository, please enable pre-commit so the code in commits are conform to the standards.

pip install -e .[tests, pre-commit]
pre-commit install

Web app

See the README.md

Build and publish

Build package:

pip install build
python -m build

Upload to PyPI:

pip install twine
twine upload dist/*

License

MIT