search

epfml / sent2vec

General purpose unsupervised sentence representations
Other
1.19k stars 256 forks source link

nnSent segfault? #7

Closed Gldkslfmsd closed 7 years ago

Gldkslfmsd commented 7 years ago

I'm still getting this segfault. My machine seems to have lots of free memory and I have no idea what's happening. Any help?

$ ./fasttext nnSent torontobooks_unigrams.bin 1 < sentence
Pre-computing sentence vectors... done.
Segmentation fault (core dumped)
martinjaggi commented 7 years ago

did you try ./fasttext nnSent torontobooks_unigrams.bin some_sentences.txt ?

Gldkslfmsd commented 7 years ago

I've tried it right now. It doesn't segfault anymore but I can't see any meaningful output...

I followed the usage message, it seems to be obsolete:

$ ./fasttext nnSent
usage: fasttext nn <model> <k>

  <model>      model filename
  <k>          (optional; 10 by default) predict top k labels
martinjaggi commented 7 years ago

if you want sentences (nnSent) not words (nn), then a list of candidate sentences is mandatory after the model argument

Gldkslfmsd commented 7 years ago

Then the help message is wrong. Can you update it, please? And where is the output printed?

Gldkslfmsd commented 7 years ago

Is the "list of candidate sentences" file supposed to have some special format or is it just one tokenized sentence per file?

martinjaggi commented 7 years ago

tokenized sentence per line

On Fri, Sep 1, 2017 at 2:23 PM Dominik Macháček notifications@github.com wrote:

Is the "list of candidate sentences" file supposed to have some special format or is it just one tokenized sentence per file?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/epfml/sent2vec/issues/7#issuecomment-326566801, or mute the thread https://github.com/notifications/unsubscribe-auth/AEaGR9sOaugBmNkvc4TTf5pbH1BiW17qks5sd_dcgaJpZM4PJHo6 .