JuliaAI / MLJFlow.jl

Connecting MLJ and MLFlow
MIT License
9 stars 0 forks source link

Changing test URI to the defined ENV variable #30

Closed pebeto closed 12 months ago

pebeto commented 12 months ago

As defined in #20 Closes #20

codecov[bot] commented 12 months ago

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (5aebc7c) 97.77% compared to head (d7658cb) 97.77%.

Files Patch % Lines
src/types.jl 0.00% 1 Missing :warning:
Additional details and impacted files ```diff @@ Coverage Diff @@ ## dev #30 +/- ## ======================================= Coverage 97.77% 97.77% ======================================= Files 4 4 Lines 45 45 ======================================= Hits 44 44 Misses 1 1 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

ablaom commented 12 months ago

Thanks @pebeto for returning to this. This now works for me locally: I launch an mlflow service on my mac, set MLFLOW_URI="http://127.0.0.1:5000" and tests run successfully!

The error I get if I don't set the env variable is not too helpful. But this is hard to fix completely, because there could be other sources for the error. Perhaps we could improve the error message issued here. Something like:

 error("It seems that the MLFlow server is not running at specified "*   
           "location, $baseuri. For more information, see https://mlflow.org/docs/latest/quickstart.html")

which adds a reference to the failing baseuri.

Another problem I noticed is this warning:

│  WARNING: Method definition log_evaluation(Any, Any) in module MLJBase at /Users/anthony/.julia/packages/MLJBase/fEiP2/src/resampling.jl:866 overwritten in module MLJFlow at /Users/anthony/MLJ/MLJFlow/src/base.jl:1.
│    ** incremental compilation may be fatally broken for this module **

I believe this is fixed by replacing this line with:

function log_evaluation(logger::Logger, performance_evaluation)

Otherwise the overloading is not specialising to MLFlow.Logger.