Reading a trained model!

[kordjamshidi]

trafficstars

[kordjamshidi]

It seems we need a different constructor for the classifier inside Learnable. When loading model the lex and lc should not be removed but used.

[kordjamshidi]

Ah! I see the same problem here :https://github.com/IllinoisCogComp/saul/blob/master/saul-examples/src/main/scala/edu/illinois/cs/cogcomp/saulexamples/nlp/EmailSpam/spamClassifiers.scala#L28 this usecache flag should not be here inside the definition of the classifier, it should be passed to the constructor. How we can do it? Do @sameersingh or anyone else have any idea? I need to do the same thing for loading a pre trained classifier.

[sameersingh]

You can create a function that takes the parameter and returns this object with the appropriate usecache flag set.

[kordjamshidi]

Thank you, for now, the save() and load() seems to be fine. But the implementation is still an issue because the model has not been saved properly in general. https://github.com/IllinoisCogComp/saul/blob/master/saul-core/src/main/scala/edu/illinois/cs/cogcomp/saul/classifier/Learnable.scala#L108 . I keep this issue open.

[danyaljj]

@kordjamshidi is it still a valid issue?

[kordjamshidi]

I think load and save can be refined still, yes. When we need to specify the path, if a path is not there what happens, etc. It is both syntax variation and documentation issue.