Commits · 057e1d9d919282e24b65fbf540dd3f46d139943c · Franck Dary / RL-Parsing

Browse files

Sep 18, 2021
- Do not read conllu file for pretrained embeddings · 057e1d9d
  Franck Dary authored 3 years ago
  
  057e1d9d
- Feature canBack is now only used when back is available · 98e2ffb0
  Franck Dary authored 3 years ago
  
  98e2ffb0
- Added tagparserlabel ts · 1e33b269
  Franck Dary authored 3 years ago
  
  1e33b269
- Print reward in debug mode · 57382909
  Franck Dary authored 3 years ago
  
  57382909
Sep 14, 2021
- Added mechanism to restart training for RL · d6407379
  Franck Dary authored 3 years ago
  
  d6407379
Sep 08, 2021
- Added lexicon · 5afbc8a4
  Franck Dary authored 3 years ago
  
  5afbc8a4
Aug 11, 2021
- Added pretrained WE for base network · 09f12081
  Franck Dary authored 3 years ago
  
  09f12081
Jul 20, 2021
- tout · 9d39440a
  Franck Dary authored 3 years ago
  
  9d39440a
Jul 11, 2021
- Added parameter probaStateBack · 792754e4
  Maxime Petit authored 4 years ago
  
  792754e4
Jul 09, 2021
- Proba oracle and random per state · 3360b487
  Franck Dary authored 4 years ago
  
  3360b487
Jul 06, 2021
- Corrected bug where system wanted to back at each end of sentence · e3a110f0
  Franck Dary authored 4 years ago
  
  e3a110f0
Jul 05, 2021
- Added tagger transition set · 6c8dcca4
  Franck Dary authored 4 years ago
  
  6c8dcca4
Jun 28, 2021
- Added argument to chose neural network type. Added separated network, where... · ade901ad
  Franck Dary authored 4 years ago
  ade901ad
- States working for RL · f336968b
  Franck Dary authored 4 years ago
  
  f336968b
Jun 25, 2021
- States working for oracle learning · 67435b57
  Franck Dary authored 4 years ago
  
  67435b57
Jun 24, 2021
- Started to work on states · bbdc365a
  Franck Dary authored 4 years ago
  
  bbdc365a
Jun 17, 2021
- New parameters : Nb of unaplayable actions picked before breaking the analysis · ead830cc
  Maxime Petit authored 4 years ago
  
  ead830cc
Jun 11, 2021
- Scores are now printed in deterministic order · 85a275f0
  Franck Dary authored 4 years ago
  
  85a275f0
Jun 10, 2021
- Added suffix feature to neural network · cdfa91d8
  Franck Dary authored 4 years ago
  
  cdfa91d8
Jun 08, 2021
- Added TAG transition and tagparser transition set · fc820be0
  Franck Dary authored 4 years ago
  
  fc820be0
May 26, 2021
- Added arc swift transition system · 7cc3e7f1
  Franck Dary authored 4 years ago
  
  7cc3e7f1
May 12, 2021
- Added hyperparameters as program arguments · 2038e1bc
  Franck Dary authored 4 years ago
  
  2038e1bc
- Added multiple reward functions and added program argument to chose them · e8b9c9f0
  Franck Dary authored 4 years ago
  
  e8b9c9f0
May 06, 2021
- Added incremental mode (-i) : parser cannot see right context if it has not... · 62b38a8c
  Franck Dary authored 4 years ago
  62b38a8c
May 04, 2021
- Corrected path of train log file · fdfc07d3
  Franck Dary authored 4 years ago
  
  fdfc07d3
- Fixed missing argument · 157585cf
  Franck Dary authored 4 years ago
  
  157585cf
- Improved debug print · 22de30ac
  Franck Dary authored 4 years ago
  
  22de30ac
- modified reward for back · 394637aa
  Maxime Petit authored 4 years ago
  
  394637aa
Apr 28, 2021
- Added history of transitions in feature function · b699477b
  Franck Dary authored 4 years ago
  
  b699477b
Apr 27, 2021
- Added Back transitions, problem : system can cycle · bf43cbf5
  Franck Dary authored 4 years ago
  
  bf43cbf5
- Added history in config · 39e488e9
  Franck Dary authored 4 years ago
  
  39e488e9
- ProbaOracle and probaRandom get smaller with number of epochs · fb6f6ffa
  Franck Dary authored 4 years ago
  
  fb6f6ffa
Apr 26, 2021
- Added form in feature function, and added threshold for number of occurences in dicts · 7729907f
  Franck Dary authored 4 years ago
  
  7729907f
- Refactored feature extraction, and added extractFeature function to NeuralNetwork · 3d80804a
  Franck Dary authored 4 years ago
  
  3d80804a
Apr 24, 2021
- Added bootstrap mode for oracle training · b1c976ae
  Franck Dary authored 4 years ago
  
  b1c976ae
- Reduced probaRandom and probaORacle · e0926705
  Franck Dary authored 4 years ago
  
  e0926705
- In RL, consider non appliable action · 9547a7de
  Franck Dary authored 4 years ago
  
  9547a7de
- Added CUDA support and changed Neural network weight init to avoid Q value explosion in RL · 7ff12b50
  Franck Dary authored 4 years ago
  
  7ff12b50
- Improved Replay Memory · 18082daf
  Franck Dary authored 4 years ago
  
  18082daf
- Code refactoring and cosmetic changes to print · bc508cdb
  Franck Dary authored 4 years ago
  
  bc508cdb