Closed LukasBeiske closed 6 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
75a38ee
) 92.44% compared to head (19791c9
) 92.47%. Report is 21 commits behind head on main.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Good start, but you should
a) Check overwrite conditions in
__init__
: if output file exists and not overwrite, raise an error, if output file exists and overwrite=true truncate it (tables.open_file(path, mode='w').b) Move the writing into the cv loop using
write_table(..., append=True)
to further reduce memory usagec) Add writing of metadata to the output file
a) The error if file exists and not overwrite is already raised by the train tools via Tool.check_output
.
a)/b) Using write_table(..., append=True, mode='w')
does not work, does it? Isn't this basically the same as write_table(..., append=True, overwrite=True)
?
c) Is there any other metadata requiered besides some column descriptions?
a) The error if file exists and not overwrite is already raised by the train tools via Tool.check_output.
Also for the cv output file and not just the model?
a)/b) Using write_table(..., append=True, mode='w') does not work, does it? Isn't this basically the same as write_table(..., append=True, overwrite=True)?
You need to truncate the file once in __init__
and then only use append=True
a) The error if file exists and not overwrite is already raised by the train tools via Tool.check_output.
Also for the cv output file and not just the model?
Yes, both output paths are passed and CrossValidator.ouput_path
gets ignored if it is None
Fixes #2480