Commits · v0.5.0 · Simon Will / fairseq · GitLab

Jun 15, 2018
- 0.4.0 -> 0.5.0 · 388c520b
  Myle Ott authored Jun 15, 2018
  View commits for tag v0.5.0 v0.5.0
  388c520b
- Update README.md · 5383b5db
  Myle Ott authored Jun 15, 2018
  5383b5db
- add links to pretrained language models · 2845d033
  Alexei Baevski authored Jun 14, 2018
  
  2845d033
- add default architecture for gbw fconv lm · fbc42f2d
  alexeib authored Jun 14, 2018
  
  fbc42f2d
- Change --path to be colon-separated instead of comma-separated · 16caed31
  Myle Ott authored Jun 14, 2018
  
  16caed31
- Faster generation when using a single model (rather than ensemble) · ef179415
  Myle Ott authored Jun 14, 2018
  
  ef179415
- Fix bidirectional lstm · bfcc6ec7
  Myle Ott authored Jun 12, 2018
  
  bfcc6ec7
- Fix tests · 55dc4842
  Myle Ott authored Jun 12, 2018
  
  55dc4842
- Updates for latest PyTorch · e89329d6
  Myle Ott authored Jun 12, 2018
  
  e89329d6
- Add FairseqTask · ff68a9ef
  Myle Ott authored Jun 12, 2018
  ff68a9ef
- torch.arange default return type is changed in the latest pytorch version... · 2de93532
  Sergey Edunov authored Jun 11, 2018
  2de93532
- Fix length penalty when combined with --no-early-stop · fc87eea2
  Myle Ott authored Jun 11, 2018
  fc87eea2
- initialize normalization constant for fconv_lm · 1fb22af4
  alexeib authored Jun 07, 2018
  
  1fb22af4
- build optimizer only once, otherwise it leaks cuda memory · f4108909
  Alexei Baevski authored Jun 05, 2018
  
  f4108909
- Update README.md · 92050ef2
  Myle Ott authored Jun 04, 2018
  
  92050ef2
- Add more integration tests (LM, stories, transformer, lstm) · 16a72b4d
  Myle Ott authored Jun 04, 2018
  
  16a72b4d
- Suppress stdout in test_train · 736fbee2
  Myle Ott authored Jun 04, 2018
  
  736fbee2
- Small fixes · 13aa36cf
  Myle Ott authored May 31, 2018
  
  13aa36cf
- create examples dir and add conv lm + stories readme · c778a31e
  Alexei Baevski authored May 31, 2018
  
  c778a31e
- Merge validate and val_loss functions (simplify train.py) · a919570b
  Myle Ott authored May 30, 2018
  
  a919570b
- Use symlinks for redundant checkpoints · 6643d525
  Myle Ott authored May 30, 2018
  
  6643d525
- Unify various sharding into ShardedIterator · 24d7de44
  Myle Ott authored May 30, 2018
  
  24d7de44
- Migrate all binaries to use options.parse_args_and_arch · 76b5ecab
  Myle Ott authored May 30, 2018
  
  76b5ecab
- Nits · cf1c64a5
  Myle Ott authored May 30, 2018
  
  cf1c64a5
- fix model loading in eval_lm · 6eda8e47
  alexeib authored May 30, 2018
  
  6eda8e47
- save best val loss in checkpoint · 295ccee9
  Alexei Baevski authored May 30, 2018
  295ccee9
- minor parameter fixes for stories model · 316744d6
  Angela Fan authored May 30, 2018
  
  316744d6
- modified writing prompts model parameters to make readme cleaner · 7a36da42
  Angela Fan authored May 30, 2018
  
  7a36da42
- added multiscale gated self attention layer with multiple heads, and pretrained fusion models · b59815bc
  Angela Fan authored May 09, 2018
  
  b59815bc
- fix default params · 50931d69
  alexeib authored May 29, 2018
  
  50931d69
- use adaptive softmax only with adaptive loss · 7f5f2461
  alexeib authored May 29, 2018
  
  7f5f2461
- record end_of_epoch in checkpoint · 7d560402
  alexeib authored May 28, 2018
  
  7d560402
- fix restoring from middle of epoch; fix defaulting transformer dropout params · 978c125a
  alexeib authored May 27, 2018
  
  978c125a
- Generalize eval_str_list · 386847ee
  Myle Ott authored May 25, 2018
  
  386847ee
- add big en_fr transformer architecture · 09379ad8
  alexeib authored May 25, 2018
  
  09379ad8
- default normalization constant for older models · e774fda7
  alexeib authored May 25, 2018
  
  e774fda7
- Conv lm implementation · 4c2ef2de
  alexeib authored May 25, 2018
  4c2ef2de
- Merge OSS + internal changes · 4e1ec2d8
  myleott authored May 23, 2018
  
  4e1ec2d8
- Fix old model checkpoints · d4816034
  Myle Ott authored May 21, 2018
  
  d4816034
- Flake8 · 81e99d8d
  Myle Ott authored May 09, 2018
  
  81e99d8d