Want to train a model without any evaluate or test dataset

[nerner94]

Hello,

I want to train a model on my knowledge graph via pipeline class without providing a test set and want to later use predict_target to predict potential tails for a head-relationship pair.

The package throws an error when I define None as argument to testing in pipeline. I also tried providing a fake file unrelated to training file, again I obtain an error. How do I go around this problem without testing file affecting the training procedure?

And may I suggest that, you do not force your users to specify a testing file if they do not need that procedure within pipeline.

Thanks.

[mberr]

Hi @nerner94 ,

if you want a more customized training that what the pipeline function offers as a default, please take a look at what is described under "Beyond the Pipeline" in this tutorial.

[nerner94]

thanks, will try the approach there, will let you know if something comes up :)