d3m.exceptions.InvalidArgumentValueError: Cannot add step 4.
sagarshankar opened this issue · 7 comments
I am attempting to implement an isolation forests pipeline with cross validation, but when I run the pipeline I get error saying that my step 4 where I implement k-folds cross validation cannot be added to the pipeline. I have attached my script for generating this pipeline and I am wondering if there is anything that is currently out of order possibly?
To add a new primitive, you may want to 1) register the primitive into the entry.ini file, 2) re-install the package with "pip install -e ." to reload the registration. For debugging, you may want to check the "input" of the produce function to make sure the input and the output are consistently as "d3m_dataframe".
I get this error in return for some reason as well. Also, I checked the "entry.ini" file and it seems that the K-Fold primitive I am using is in there. I also got this other error as well. Which I am trying to figure out why that would be the case. Any feedback would be great.
raise exceptions.InvalidArgumentValueError("Cannot add step {step_index}.".format(step_index=step.index)) from error
Based on the provided snapshots, I don't think there is a problem. Would you mind to provide the full error message?
Yeah this is the full error. It mentions that the K-Fold_Split seems to require some additional arguments and that seems to be the main cause, but I am not sure why that is.
So I also noticed in the "KFoldSplit.py" file, that there is no function called "produce" which is used in many of the other primitives. There is a function called "def _get_splits", but when I say step_4.add_output('_get_splits), I get the following error as well.
In our backend D3M system, only certain APIs of a primitive (e.g., "produce", "fit", "fit_produce") will be approachable by the pipeline constructor. Therefore, you might want to make sure the produce function is the one performing the functionality.
So what exactly is the acceptable API for the K Fold primitive? I have tried "produce", "fit" and "fit_produce" but none of them sem to be working so far