clearml icon indicating copy to clipboard operation
clearml copied to clipboard

Store models offline

Open shaziaakbar opened this issue 2 years ago • 3 comments

Hi everyone, The "offline mode" feature is great when using HPC environments with no internet connection. Would it be possible to expand this to allows models to be saved (and later uploaded to ClearML)?


Good news guys, offline mode is now available 🎊 With the new release of trains 0.16 you can set Task.set_offline to True. Later you can import your session (either the zip file or folder) with Task.import_offline_session.

The offline session includes all console output, metric logs & artifacts. (It does not include stored models, let us know if you would like to see that included as well :)

Originally posted by @bmartinn in https://github.com/allegroai/clearml/issues/117#issuecomment-672302058

shaziaakbar avatar Apr 22 '22 13:04 shaziaakbar

Hi @shaziaakbar,

Just to make sure I understand correctly, you'd like to store locally the models created in offline mode, and then when importing the offline session to clearml, we will also import the models themselves?

erezalg avatar Apr 24 '22 07:04 erezalg

Hi @shaziaakbar,

Just to make sure I understand correctly, you'd like to store locally the models created in offline mode, and then when importing the offline session to clearml, we will also import the models themselves?

Yes exactly :)

shaziaakbar avatar Apr 24 '22 12:04 shaziaakbar

Hello @shaziaakbar

Please apology for not having answered faster. Parameters, plots, artefacts and so on are pulled when uploading an offline session ; but you are right, the model is not.

We are currently working on the issue and will soon release a version that will allow to also pull the model.

DavidNativ avatar May 19 '22 12:05 DavidNativ

Hi @shaziaakbar , this should be in the upcoming release that is pretty close :)

NathanClearML avatar Mar 19 '23 16:03 NathanClearML

Hi @shaziaakbar , this should be in the upcoming release that is pretty close :)

Great! Looking forward to this addition

shaziaakbar avatar Mar 30 '23 22:03 shaziaakbar

Hey @shaziaakbar ! Just letting you know that this issue has been resolved in the recently released v1.10.0. Let us know if there are any issues :)

pollfly avatar Apr 03 '23 07:04 pollfly

@shaziaakbar closing this, please reopen if required.

jkhenning avatar Aug 06 '23 07:08 jkhenning