cdec-dtrain/word-aligner, branch sa_mmap Mirror of https://github.com/pks/cdec-dtrain.git https://git.simianer.de/mirrored/cdec-dtrain/atom?h=sa_mmap 2012-12-15T07:53:56Z enable kenlm compression 2012-12-15T07:53:56Z Chris Dyer cdyer@allegro.clab.cs.cmu.edu 2012-12-15T07:53:56Z urn:sha1:29a47a94bfc09450802484e5cd3f835d39c9f66c optionally optimize diagonal tension 2012-11-25T05:28:07Z Chris Dyer cdyer@allegro.clab.cs.cmu.edu 2012-11-25T05:28:07Z urn:sha1:9b3306332a27f23a36e96a93b5ff97caee1b6e3c comment out asserts 2012-11-25T01:25:51Z Chris Dyer cdyer@allegro.clab.cs.cmu.edu 2012-11-25T01:25:51Z urn:sha1:efa888a623913e72c5dd0e527ba7da014d52597e victor's geometric series trick 2012-11-25T01:22:45Z Chris Dyer cdyer@allegro.clab.cs.cmu.edu 2012-11-25T01:22:45Z urn:sha1:147238b755eeeb4623ce74aad79d62c378b54ea5 fixes for 2011 optimizer 2012-11-20T18:56:08Z Chris Dyer cdyer@allegro.clab.cs.cmu.edu 2012-11-20T18:56:08Z urn:sha1:6802ac200ef614b4935d597ed4cfc3857c1f6c06 major restructure of the training code 2012-11-18T18:35:42Z Chris Dyer cdyer@allegro.clab.cs.cmu.edu 2012-11-18T18:35:42Z urn:sha1:1b8181bf0d6e9137e6b9ccdbe414aec37377a1a9 rename model1 to fast_aligner, since it does way more than model1 but is mostly just fast 2012-09-21T01:51:11Z Chris Dyer cdyer@cs.cmu.edu 2012-09-21T01:51:11Z urn:sha1:78518f1f417616633b300a361cd5e0c1bcb1ff24 more alignment stuff 2012-01-23T20:47:29Z Chris Dyer cdyer@cs.cmu.edu 2012-01-23T20:47:29Z urn:sha1:26d9ad04bd81508163d75c99726f970dd75f5127 'pseudo model 2' that strictly favors a diagonal, with tunable parameters for p(null) and how sharp/flat the alignment distribution is around the diagonal 2012-01-20T20:35:47Z Chris Dyer cdyer@cs.cmu.edu 2012-01-20T20:35:47Z urn:sha1:f0bdd4de6455855d705d9056deb2e90c999dc740 lbfgs + time-series minibatch optimization 2011-10-31T18:03:22Z Chris Dyer cdyer@cs.cmu.edu 2011-10-31T18:03:22Z urn:sha1:6de8f58cd13813bf33af4903bf386439683c0fd6