无法访问 mlflow ui

Question

By following quickstart and tutorial at https://www.mlflow.org/docs/latest/quickstart.html , and https://www.mlflow.org/docs/latest/tutorial.html , the execution of train.py works fine.

Elasticnet model (alpha=0.500000, l1_ratio=0.500000):
  RMSE: 0.8222428497595403
  MAE: 0.6278761410160693
  R2: 0.12678721972772622

But when launching the ui mlflow ui , and accessing to the web page localhost:5000, the browser complains

Not Found

The requested URL was not found on the server. If you entered the URL manually please check your spelling and try again.

What went wrong and how to fix this?

Thanks

Answer 1

There's many different issues that may or may not be happening. You're best bet is to plain-old troubleshoot. I saw you're comment on GitHub, so let's try a new angle; it's more than likely a building/debugging file path or directory setting which is incorrect.

I noticed a file/directory setting named set_tracking_uri .

def get_tracking_uri():
    """
    Returns the current tracking URI. Note that this may not correspond to the tracking URI of
    the currently active run, since the tracking URI may be updated via `set_tracking_uri`.
    :return: the tracking URI
    """

mlflow/tracking/__init__.py - lines 52-56

You could also check out the package.json for the server file path.

 "proxy": "http://localhost:5000",
  "homepage": "/static-files",
  "jest": {
    "collectCoverageFrom": [
      "src/**/*.{js,jsx}",
      "!**/*.test.{js,jsx}",
      "!**/__tests__/*.{js,jsx}"
    ]

mlflow/mlflow/server/js/package.json - lines 37-44

Make sure your files are in the correct directory compared to your settings and I'm sure it will resolve the issue.

Answer 2

Inside docker container, adding the --host 0.0.0.0 solved my issue. Using mlflow 1.17.0

mlflow ui --host 0.0.0.0

https://github.com/mlflow/mlflow/issues/109#issuecomment-404146077