Closed kordjamshidi closed 8 years ago
I am not sure what has changed in the recent versions of the Lbjava with the ILPInference, when using version 24
I get this error:
Error:(6, 112) overloaded method constructor ILPInference with alternatives:
(x$1: Any,x$2: edu.illinois.cs.cogcomp.infer.ilp.ILPSolver)edu.illinois.cs.cogcomp.lbjava.infer.ILPInference <and>
(x$1: edu.illinois.cs.cogcomp.infer.ilp.ILPSolver,x$2: Int)edu.illinois.cs.cogcomp.lbjava.infer.ILPInference
cannot be applied to (T, edu.illinois.cs.cogcomp.lbjava.infer.ILPSolver)
abstract class JointTemplate[T](head: T, solver: ILPSolver, norm: Normalizer = new IdentityNormalizer) extends ILPInference(head, solver) {
^
is it compile error?
yes.
it is a package name change. Remove the import (on top of the file) and add the correct one.
It seems you have two copies of some of the classes in lbjava.infer and core.infer, right?
I think you just need to fix the package import ...
I have it in many files, and some of them still get things from lbjava.infer
Too many package changes, I fixed the ones with I/O
for Readers for pos tagger... the ILPInference issue seems to be ok now!
where is this one? PennTreebankPOSReader
You don't really need to ask us; you can just search inside our super-project: https://github.com/IllinoisCogComp/illinois-cogcomp-nlp/blob/master/corpusreaders/src/main/java/edu/illinois/cs/cogcomp/nlp/corpusreaders/PennTreebankPOSReader.java
You don't need to answer if too annoying!
I don't see this file in current Saul's cogcomp nlp!
Did this happen after changing the lbjava version? You need to see what is the latest "illinois-core-utils" and "corpus-readers" in your classpath.
if the issue is not resolved I can look at it tonight
No, thanks. It has been resolved.
@kordjamshidi is this functional/ready?
Actually, not yet, there was something wrong with the feature indexes that I could not resolve, so it does not generate the correct results for ER.
it has been a while that I've left it, but I will get back to it this weekend.
Could you elaborate more? Is it a very big issue? Does it make sense to merge the current version and then add the fixes later?
I don't think it is a big issue, it is related to concatenation of the feature vectors of various classifiers and making them one global vector. I do it but somewhere, something goes wrong with this module. It should be a small fix but needs a careful debug. The confusion comes from the difference between the way sparse vectors are stored in Lbjava and SL, I think. I think the code is well structured and is on the right track, I can clean it if you think we better to merge it.
okei then let's aim at getting this reviewed and checked then, over the next week or so.
Ok, sure. Good idea.
Since we changed the semaphore account, it doesn't seem to be get triggered for this PR. Will close it and open another one.
This is a new PR from #247