kwdagger 0.2.5

The kwdagger module

Read the Docs	http://kwdagger.readthedocs.io/en/latest/
Gitlab (main)	https://gitlab.kitware.com/computer-vision/kwdagger
Github (mirror)	https://github.com/Kitware/kwdagger
Pypi	https://pypi.org/project/kwdagger

Overview

KWDagger is a lightweight framework for defining bash-centric DAGs and running large parameter sweeps. It builds on top of cmd_queue and scriptconfig to provide:

• Reusable kwdagger.pipeline.Pipeline and kwdagger.pipeline.ProcessNode abstractions for wiring inputs / outputs together.

• A scheduling CLI (kwdagger.schedule) that materializes pipeline definitions over a parameter grid and executes them via Slurm, tmux, or a serial backend.

• An aggregation CLI (kwdagger.aggregate) that loads job outputs, computes metrics, and optionally plots parameter/metric relationships.

• A self-contained demo pipeline in kwdagger.demo.demodata that is used in CI and serves as a reference implementation.

Repository layout

• kwdagger/pipeline.py – core pipeline and process node definitions, networkx graph construction, and configuration utilities.

• kwdagger/schedule.py – ScheduleEvaluationConfig CLI for expanding parameter grids into runnable jobs and dispatching them through cmd_queue backends.

• kwdagger/aggregate.py – AggregateEvluationConfig CLI for loading job outputs, computing parameter hash IDs, and generating text/plot reports.

• kwdagger/demo/demodata.py – end-to-end demo pipeline with prediction and evaluation stages plus CLI entry points for each node.

• docs/ – Sphinx sources, including an example user module under docs/source/manual/tutorials/twostage_pipeline.

• tests/ – unit and functional coverage for pipeline wiring, scheduler behavior, aggregation, and import sanity checks.

Quickstart

Run the demo pipeline locally to see the CLI workflow end-to-end:

TMP_DPATH=$(mktemp -d --suffix "-kwdagger-demo")
cd "$TMP_DPATH"
echo "demo" > input.txt

EVAL_DPATH=$PWD/pipeline_output
python -m kwdagger.schedule \
    --params="
        pipeline: 'kwdagger.demo.demodata.my_demo_pipeline()'
        matrix:
            stage1_predict.src_fpath:
                - input.txt
            stage1_predict.param1:
                - 123
            stage1_evaluate.workers: 2
    " \
    --root_dpath="${EVAL_DPATH}" \
    --backend=serial --skip_existing=1 --run=1

python -m kwdagger.aggregate \
    --pipeline='kwdagger.demo.demodata.my_demo_pipeline()' \
    --target "
        - $EVAL_DPATH
    " \
    --output_dpath="$EVAL_DPATH/full_aggregate" \
    --eval_nodes="
        - stage1_evaluate
    " \
    --stdout_report="
        top_k: 10
        concise: 1
    "

The scheduler will generate per-node job directories with invoke.sh and job_config.json metadata. The aggregator then consolidates results, computes parameter hash IDs, and prints a concise report.

A novel graph based symlink structure allows for navigation of dependencies within a node. The .succ folder holds symlinks to successors (i.e. results that depend on the current results), and .pred holds symlinks to folders of results that the current folder depends on.

For more in-depth information see tutorials:

• Tutorial #1: twostage_pipeline

• Tutorial #2: ollama_benchmark

Command line entry points

• python -m kwdagger.schedule or kwdagger schedule – build and run a pipeline over a parameter matrix (see kwdagger.schedule.ScheduleEvaluationConfig).

• python -m kwdagger.aggregate or kwdagger aggregate – load completed runs and generate tabular and plotted summaries (kwdagger.aggregate.AggregateEvluationConfig).

• python -m kwdagger – modal CLI that exposes the schedule and aggregate commands via kwdagger.__main__.KWDaggerModal.