Iota2 won't start if no classifier is set
I am working on adding the possibility to use regression to the pretrained model builder. The first step being the generation of samples, I modified a configuration file to stop the chain to the sampling step. I also deleted the arg_train
section, as I wouldn't need it. But if this section isn't present in the config file, iota2 won't start, even if the section/parameter isn't used (e.g. using iota2 only until the sampling step) because no classifier is set:
iota2.configuration_files.sections.cfg_utils.ConfigError: iota2 configuration file error : No classifier detected in the configuration file, please set one.
The same issue occurs when the parameter classifier
isn't present in the case of a classification run.
This issue comes from the fact that each builder has different step names, meaning that this parameter may be used in steps with differing names (depending on the builder). Therefore, each builder should integrate its own specific method to handle this issue and deactivate parameters depending on planned steps.