[BUG] Each step is no longer displayed in the metric plots

[ibaris]

Willingness to contribute

No. I cannot contribute a bug fix at this time.

MLflow version

2.11.3

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows 10
• Python version: 3.9
• yarn version, if running the dev UI:

Describe the problem

The metric plots does not show every step anymore:

• Version 2.11.3

• Version 2.9.3

Steps to reproduce the bug

I displayed the same training with two different version of mlflow.

Code to generate data required to reproduce the bug

No response

Is the console panel in DevTools showing errors relevant to the bug?

No response

Does the network panel in DevTools contain failed requests relevant to the bug?

No response

[daniellok-db]

Metric sampling was added in MLflow 2.11 to improve UI performance. At the moment the number of sampled points isn't configurable, but we could look into adding support for this.

[daniellok-db]

@ibaris in the meantime, if you install the dev version of mlflow (pip install git+https://github.com/mlflow/mlflow.git), does it help? there was a commit a couple weeks ago that fixed some sampling logic (it doesn't remove sampling, but may cause more points to show up).

[ibaris]

Thank you for the reply!

@ibaris in the meantime, if you install the dev version of mlflow (pip install git+https://github.com/mlflow/mlflow.git), does it help? there was a commit a couple weeks ago that fixed some sampling logic (it doesn't remove sampling, but may cause more points to show up).

Unfortunatelly, I run into the problem:

Unable to display MLflow UI - landing page (index.html) not found.

You are very likely running the MLflow server using a source installation of the Python MLflow
package.

If you are a developer making MLflow source code changes and intentionally running a source
installation of MLflow, you can view the UI by running the Javascript dev server:
https://github.com/mlflow/mlflow/blob/master/CONTRIBUTING.md#running-the-javascript-dev-server

Otherwise, uninstall MLflow via 'pip uninstall mlflow', reinstall an official MLflow release
from PyPI via 'pip install mlflow', and rerun the MLflow server.

It would be really helpful, if one could configure the number of samples. The Problem is, that the validation plots consists sometimes of only two data points which makes mlflows not usable anymore (this is only my opinion).

[daniellok-db]

Ahh, darn 😕. The next release should be sometime in the next couple of weeks, and the sampling fix should be included there. I hear you about wanting the samples to be configurable though—I think others would probably like this too. We can look into supporting it!

[ibaris]

Thank you very much for the support. I am looking forward to the new release. You guys are awesome 👍🏻

[github-actions[bot]]

@mlflow/mlflow-team Please assign a maintainer and start triaging this issue.