Commits · 5ed6be65d1ea16e0202c09ffbedd3cdb73916083 · Franck Dary / macaon · GitLab

Snippets Groups Projects

Apr 30, 2020
- Changed name of optimizer checkpoints to avoid confusion with models checkpoints · 47cbed2e
  Franck Dary authored 4 years ago
  
  47cbed2e
- Making sure optimizer have access to embeddings parameters · b4c4cd4c
  Franck Dary authored 4 years ago
  
  b4c4cd4c
- Corrected bug where embeddings were not loaded when training resumed · 93b2c58c
  Franck Dary authored 4 years ago
  
  93b2c58c
Apr 29, 2020
- Corrected a bug where dict was modified if training was resumed · 29907cb5
  Franck Dary authored 4 years ago
  
  29907cb5
- Corrected a bug where embeddings module was not sent to cuda · 431fab99
  Franck Dary authored 4 years ago
  
  431fab99
- Rare values are now treated as unknown values. Embeddings sizes now exactly match dict size · d217cf58
  Franck Dary authored 4 years ago
  
  d217cf58
- Fixed error message · 7982090e
  Franck Dary authored 4 years ago
  
  7982090e
- Implemented all modules · ca72f53e
  Franck Dary authored 4 years ago
  
  ca72f53e
- First draft of modular neural network · b4228d7b
  Franck Dary authored 4 years ago
  
  b4228d7b
Apr 24, 2020
- Updated syntax for pytorch 1.5 · bfd53a17
  Franck Dary authored 4 years ago
  
  bfd53a17
Apr 23, 2020
- Fixed bugs when there was no HEAD col in mcd · b43d29d9
  Franck Dary authored 4 years ago
  
  b43d29d9
Apr 20, 2020
- Each classifier can have a different output layer in neural networks · a31419f9
  Franck Dary authored 4 years ago
  
  a31419f9
Apr 17, 2020
- split unknown only when extracting train dataset · 28045459
  Franck Dary authored 4 years ago
  
  28045459
- Added option to chose if embeddings dropout is 2d or not · e4e9d906
  Franck Dary authored 4 years ago
  
  e4e9d906
- Added total input dropout to LSTMNetwork · 70896a64
  Franck Dary authored 4 years ago
  
  70896a64
Apr 16, 2020
- Optimizer is now a member of Classifier, and its parameters can be given in rm file · e350f836
  Franck Dary authored 4 years ago
  
  e350f836
- embeddings dropout is now a LSTM classifier parameter · 979104ff
  Franck Dary authored 4 years ago
  
  979104ff
Apr 14, 2020
- do not close dict during example extraction · 0daae795
  Franck Dary authored 4 years ago
  
  0daae795
Apr 11, 2020
- Fixed a bug with focused indexes where comments where not ignored during training · d5a1c96e
  Franck Dary authored 4 years ago
  
  d5a1c96e
Apr 10, 2020
- Working tokeparser_seq · 757de891
  Franck Dary authored 4 years ago
  
  757de891
- Working tokeparser_incr · 209dd02c
  Franck Dary authored 4 years ago
  
  209dd02c
- Integrated DepthLayerTreeEmbedding · 58a388d4
  Franck Dary authored 4 years ago
  
  58a388d4
- Updated LSTMNetwork and attach action to work with tree embedding networks · 4bfe6147
  Franck Dary authored 4 years ago
  
  4bfe6147
Apr 09, 2020
- Fixed bug in MLP init · ff3894a3
  Franck Dary authored 4 years ago
  
  ff3894a3
- Added MLP · d1e40df1
  Franck Dary authored 4 years ago
  
  d1e40df1
- print number of parameters of the model · 86ded225
  Franck Dary authored 4 years ago
  
  86ded225
Apr 08, 2020
- setting neural network to correct device · 991e94f3
  Franck Dary authored 4 years ago
  
  991e94f3
- Fixed some problems in dependency parsing · 15d915ca
  Franck Dary authored 4 years ago
  
  15d915ca
- changed the way classifiers are initialized · cac30a69
  Franck Dary authored 4 years ago
  
  cac30a69
- code refactoring · 8adb3462
  Franck Dary authored 4 years ago
  
  8adb3462
Apr 05, 2020
- Made splitwords transition set optional in definition of reading machine · b9a3230c
  Franck Dary authored 4 years ago
  
  b9a3230c
- Sequential · 050839ed
  Franck Dary authored 4 years ago
  
  050839ed
Apr 03, 2020
- Corrected a bug where splitword had 0 cost even when it didn't match the size of the gold multiword · 03221c3f
  Franck Dary authored 4 years ago
  
  03221c3f
Apr 02, 2020
- added transition split · d438d3c3
  Franck Dary authored 4 years ago
  
  d438d3c3
Apr 01, 2020
- Checkpoints are created after each training epoch and it is possible to resume... · e45d45e6
  Franck Dary authored 4 years ago
  
  Checkpoints are created after each training epoch and it is possible to resume a training by training again on the same directory
  e45d45e6
- Deleted OneWordNetwork · 1cf6cf2f
  Franck Dary authored 4 years ago
  
  1cf6cf2f
- Fixed bug in Transition splitword where we used a non utf8string · 07748169
  Franck Dary authored 4 years ago
  
  07748169
- Ignore empty node when reading tsv · b4d97a44
  Franck Dary authored 4 years ago
  
  b4d97a44
Mar 31, 2020
- Throwing exception when rm file not found · 7e7d8d52
  Franck Dary authored 4 years ago
  
  7e7d8d52
- .ts file path is now relative to expdir · 2ac42b6d
  Franck Dary authored 4 years ago
  
  2ac42b6d