Closed LukasNickel closed 3 years ago
Currently this just raises a KeyError
. Which is also pretty clear I'd say, but can surely be improved to be something like an AICTToolsConfigError
.
To elaborate:
It is not a problem for the scripts, the methods just dont behave the same way. I can do from aict_tools.io import read_data
to use that method in external scripts or when looking at FACT data when I do not necessarily know what columns are present.
Just handy sometimes.
But for CTA files I need to use read_telescope_data
, which requires to set up an AICTConfig
anyway (which I forgot), so thats probably not a use case right now
You mean features in the config? That's expected. Training a model without features doesn't make sense. Just choosing all available also doesn't really make sense. People need to make a choice here.
So probably we can just improve the error message?
"features" is a required configuration field"
?