Fixing MLFlow Tracking Directory

[aritraghsh09]

In the last two weeks at some point, we had erroneously changed where the MLFlow results were being stored.

We want the MLFlow results to be stored,not inside individual results directories, but at the root. This was the behaviour before. Not sure when it got changed.

[Copilot]

Pull request overview

Fixes Hyrax’s MLflow file-store location so experiment/run artifacts are tracked under the configured results root directory (the historical behavior) rather than inside each per-run results subdirectory.

Changes:

• Update MLflow tracking URI in the train verb to use <config.general.results_dir>/mlflow instead of <results_dir>/mlflow.

[Copilot]

Building the MLflow tracking URI via string concatenation ("file://" + str(...)) can produce invalid URIs on some platforms (notably Windows paths). Prefer constructing the URI with Path(...).as_uri() (or equivalent) so the file:// URL is correctly formed across OSes.

Suggested change

	mlflow.set_tracking_uri("file://" + str(results_root_dir / "mlflow"))
	mlflow_tracking_dir = (results_root_dir / "mlflow").resolve()
	mlflow.set_tracking_uri(mlflow_tracking_dir.as_uri())

[Copilot]

This change is a regression fix in where MLflow stores its backend (root results_dir vs per-run results_dir). There are existing end-to-end/unit tests for training (tests/hyrax/test_train.py), but none assert the MLflow tracking directory behavior; consider adding a test that mocks mlflow.set_tracking_uri and verifies it is called with <config.general.results_dir>/mlflow (not <results_dir>/mlflow).

[aritraghsh09]

@copilot open a new pull request to apply changes based on this feedback

[Copilot]

@aritraghsh09 I've opened a new pull request, #694, to work on those changes. Once the pull request is ready, I'll request review from you.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 63.26%. Comparing base (68c8e0c) to head (2ff808b).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #693      +/-   ##
==========================================
+ Coverage   63.25%   63.26%   +0.01%     
==========================================
  Files          59       59              
  Lines        5769     5771       +2     
==========================================
+ Hits         3649     3651       +2     
  Misses       2120     2120              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[drewoldag]

This seems fine, especially if this unblocks you.

[github-actions]

Before [68c8e0c]	After [80ca930]	Ratio	Benchmark (Parameter)
1.93±0.1s	2.05±0.05s	1.06	benchmarks.time_download_help
1.94±0.08s	2.03±0.01s	1.05	benchmarks.time_help
1.88±0.08s	1.97±0.04s	1.05	benchmarks.time_train_help
1.89±0.07s	1.97±0.04s	1.04	benchmarks.time_database_connection_help
1.91±0.08s	1.98±0.02s	1.04	benchmarks.time_infer_help
203±5ms	209±3ms	1.03	benchmarks.time_import
1.93±0.07s	1.97±0.03s	1.02	benchmarks.time_lookup_help
1.94±0.06s	1.97±0.03s	1.02	benchmarks.time_prepare_help
1.93±0.08s	1.97±0.04s	1.02	benchmarks.time_rebuild_manifest_help
1.90±0.09s	1.94±0.02s	1.02	benchmarks.time_umap_help

Click here to view all benchmarks.