Something like this, SuccessfulKoala55 ?
Open a bash session on the docker ( docker exec -it <docker id> /bin/bash
) Open a mongo shell ( mongo
) Switch to backend db ( use backend
) Get relevant project IDs ( db.project.find({"name": "ClearML Examples"})
and db.project.find({"name": "ClearML - Nvidia Framework Examples/Clara"})
) Remove relevant tasks ( db.task.remove({"project": "<project_id>"})
) Remove project IDs ( db.project.remove({"name": ...})
)
Hi UnevenDolphin73 , the quickest way would be to locate the relevant project in mongodb under the backend.project
collection, delete the project document and any task documents in the backend.task
collection who'se project
field refers to the project in question.
This would of course not delete the example experiment's metrics and files, but these are relatively small and should not pose any problem.
Hi SuccessfulKoala55 !
Could you elaborate on how best to delete these from the database?
Perfect, I'll just reinstall then, don't have any important models to lose yet. Thanks!
Hi TroubledJellyfish71 ,
Well, you can disable these when you install the server (i.e. make sure they will not be installed at all) by changing the following line in the docker-compose file to "false"
:CLEARML__apiserver__pre_populate__enabled: "true"
On an already-installed server, you can only remove them manually by deleting them from the database 😕