openml / automlbenchmark

OpenML AutoML Benchmarking Framework
https://openml.github.io/automlbenchmark
MIT License
399 stars 132 forks source link

Simplify output paths #510

Open Innixma opened 1 year ago

Innixma commented 1 year ago

Currently, when running AMLB on AWS mode (also happens in local mode), the output paths are longer than necessary:

This is the path saved to S3 for the predictions.csv output:

s3://bucket/autogluon.test.test.aws.20230220T221019/aws.test.test.cholesterol.0.autogluon/output/predictions/cholesterol/0/predictions.csv

Ideally, this could be simplified to:

s3://bucket/autogluon.test.test.aws.20230220T221019/aws.test.test.cholesterol.0.autogluon/output/predictions.csv

This is already the case for results.csv, which is found in:

s3://bucket/autogluon.test.test.aws.20230220T221019/aws.test.test.cholesterol.0.autogluon/output/results.csv

(Note: This applies to all frameworks, not just AutoGluon)

It would be really nice to simplify these paths so that coding logic that fetches these files is easier, since currently the path structure under output depends on the name of the task, which makes shell script parsing very complicated compared to if the output folder structure wasn't altered by the name of the task.