swiseman / nn_coref Goto Github PK

Hiya!

I'm trying to compile the modifiedBCS stuff and getting

[error] ./src/main/java/edu/harvard/nlp/moarcoref/MiniDriver.java:15: cannot find symbol
[error]   symbol:   variable WRITE_FEATS_SEP
[error]   location: class edu.harvard.nlp.moarcoref.MiniDriver.Mode
[error]   public static Mode mode = Mode.WRITE_FEATS_SEP;
[error] (compile:compileIncremental) javac returned nonzero exit code
[error] Total time: 32 s, completed Feb 8, 2017 1:22:34 PM

Should the Mode enum in that class have another value?

Hi!

Thanks for the code!

I'm trying to run the command to extract the features using the pre-compiled moarcoref-assembly-1.jar, and I get the message below about invalid options:

java -jar -Xmx30g modifiedBCS/moarcoref-assembly-1.jar ++modifiedBCS/base.conf -execDir /home/loaiciga/coref/features/ -numberGenderData /home/loaiciga/coref/conll-2012/gender.data -animacyPath /home/loaiciga/coref/conll-2012/animate.unigrams.txt -inanimacyPath /home/loaiciga/coref/conll-2012/inanimate.unigrams.txt -trainPath /home/loaiciga/coref/conll-2012/flat_train_2012 -devPath /home/loaiciga/coref/conll-2012/flat_dev_2012 -testPath /home/loaiciga/coref/conll-2012/flat_test_2012 -mode SMALLER -conjType NONE -pairwiseFeats FINAL+MOARANAPH+MOARP

Invalid enum: 'SMALLER'; valid choices: WRITE_FEATS_SEP

And using the WRITE_FEATS_SEP option, doesn’t produce the OPCs.txt files.

Any suggestions?

Thanks again,
Sharid

Hi Sam,

I've encountered an issue I believe related to hdf5 when running the vanilla_mr.lua. I have installed all necessary packages (nn, hdf5), and have put all resources in their right location (hdf5 files and other required data). When running th vanilla_mr.lua, the following error appears:

$ th vanilla_mr.lua -PT
/Users/Thanakorn/torch/install/bin/lua: /Users/Thanakorn/torch/install/share/lua/5.2/hdf5/group.lua:312: HDF5Group:read() - no such child 'feats' for [HDF5Group 33554432 /]
stack traceback:
[C]: in function 'error'
/Users/Thanakorn/torch/install/share/lua/5.2/hdf5/group.lua:312: in function </Users/Thanakorn/torch/install/share/lua/5.2/hdf5/group.lua:302>
(...tail calls...)
./sparse_doc_data.lua:10: in function 'loadFromH5'
vanilla_mr.lua:307: in function 'main'
vanilla_mr.lua:349: in main chunk
[C]: in function 'dofile'
...korn/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
[C]: in ?

Any suggestions would be appreciated! Thanks!

Hi Sam,

Let me first thank you. Your implementation helped me understand a lot of stuff with coreference, Torch and neural networks.

I am trying to use your version of the Berkeley feature extractor under modifiedBCS but I am having trouble using it for gold mentions. I used it on automatic mentions already without problems.

Here is what I tried:

java -jar -Xmx30g modifiedBCS/target/scala-2.11/moarcoref-assembly-1.jar ++modifiedBCS/base.conf -execDir execdir -numberGenderData modifiedBCS/gender.data -animacyPath modifiedBCS/animate.unigrams.txt -inanimacyPath modifiedBCS/inanimate.unigrams.txt -trainPath ./data/flat_gold_conll_train_2012/ -devPath ./data/flat_gold_conll_dev_2012/ -testPath ./data/flat_test_2012/ -mode SMALLER -conjType NONE -pairwiseFeats FINAL+MOARANAPH+MOARPW -trainOnGold
main() {
  Execution directory: execdir
  Using conjType = NONE
  Reading Bergsma and Lin file from modifiedBCS/gender.data
  Done!
  ERROR: java.lang.IllegalArgumentException: requirement failed:
scala.Predef$.require(Predef.scala:207)
edu.harvard.nlp.moarcoref.FeatureExtractor$.writeSeparatedFeatsAndOraclePredClustering(FeatureExtractor.scala:24)
edu.harvard.nlp.moarcoref.FeatureExtractor.writeSeparatedFeatsAndOraclePredClustering(FeatureExtractor.scala)
edu.harvard.nlp.moarcoref.MiniDriver.run(MiniDriver.java:70)
edu.berkeley.nlp.futile.fig.exec.Execution.runWithObjArray(Execution.java:479)
edu.berkeley.nlp.futile.fig.exec.Execution.run(Execution.java:432)
edu.harvard.nlp.moarcoref.MiniDriver.main(MiniDriver.java:64)
  Execution directory: execdir
1 errors, 0 warnings
} [3.9s]

The following line seems to be my problem (FeaturExtractor.scala):

def writeSeparatedFeatsAndOraclePredClustering(smaller:Boolean) {
    ...
    require(!MiniDriver.trainOnGold);
    ...

Why are you disabling gold mentions? I tried using the options -trainOnGold, -useGoldMentions, -docSuffix gold_conll to no avail.

I run the python script, but get the error:
luarocks/loader.lua:154: Invalid data in manifest file for module libcr (invalid data for cr scm-1)
stack traceback:
[C]: in function 'error'
.../torch/install/share/lua/5.2/trepl/init.lua:384: in function 'require'
init.lua:3: in main chunk
[C]: in function 'dofile'
full_g1_model.lua:2: in main chunk
[C]: in function 'dofile'
.../torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
[C]: in ?
I cd to the "nn" folder and compile: "luarocks make cr-scm-1.rockspec" and then re-run, but the problem still remains..
Anybody know the issue?