You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Having a Pathlib Path as a variable in a dataclass and then calling Task.connect on it will not report the variable under the configuration section at all.
Perhaps it is worth thinking about which types we can still include as long as they are reasonable to cast to a string. I expect the difficulty will come when changing this value and rerunning the experiment?
To reproduce
Example:
@dataclassclassSingleModeClassifierConfig:
commit: str=COMMIT# Commit of the particle extraction to useroot_folder: Path=ROOT_FOLDER
If in this case one does task.connect(SingleModeClassifierConfig()) the root_folder parameter will not be found under the configuration section. Probably ClearML does not find a way to turn it into a string?
Describe the bug
Having a Pathlib Path as a variable in a dataclass and then calling Task.connect on it will not report the variable under the configuration section at all.
Perhaps it is worth thinking about which types we can still include as long as they are reasonable to cast to a string. I expect the difficulty will come when changing this value and rerunning the experiment?
To reproduce
Example:
If in this case one does
task.connect(SingleModeClassifierConfig())
the root_folder parameter will not be found under the configuration section. Probably ClearML does not find a way to turn it into a string?Expected behaviour
The variable to be logged as a string
Environment
The text was updated successfully, but these errors were encountered: