|
|
Allow larger maxlen (fixes #100) (#101)
|
 Myle Ott
|
|
6 years ago |
|
|
Adjust weight decay by the current learning rate to make it work correctly during annealing
|
 Sergey Edunov
|
|
6 years ago |
|
|
Merge pull request #91 from facebookresearch/prepare_wmt
|
 Sergey Edunov
|
|
6 years ago |
|
|
spelling
|
Sergey Edunov
|
|
6 years ago |
|
|
Update README with new models
|
Sergey Edunov
|
|
6 years ago |
|
|
Merge pull request #95 from bastings/patch-1
|
Sergey Edunov
|
|
6 years ago |
|
|
Adding README and more parameters to En2De script
|
Sergey Edunov
|
|
6 years ago |
|
|
Ratio should be predlen/reflen not reflen/predlen
|
 Joost Bastings
|
|
6 years ago |
|
|
Merge branch 'master' of github.com:facebookresearch/fairseq-py into prepare_wmt
|
Sergey Edunov
|
|
6 years ago |
|
|
Switch to news-commentary-v12
|
Sergey Edunov
|
|
6 years ago |
|
|
Fixed Weight Decay Regularization in Adam
|
 Michael Auli
|
|
6 years ago |
|
|
Fix tests
|
Myle Ott
|
|
6 years ago |
|
|
Output correct perplexity when training with --sentence-avg
|
Myle Ott
|
|
6 years ago |
|
|
Fix max_positions calculation in train.py
|
Myle Ott
|
|
6 years ago |
|
|
Better warning message for inputs that are too long
|
Myle Ott
|
|
6 years ago |
|
|
ATen Fix
|
 Michael Auli
|
|
6 years ago |
|
|
Momentum correction
|
Michael Auli
|
|
6 years ago |
|
|
Report log likelihood for label smoothing
|
Sergey Edunov
|
|
6 years ago |
|
|
Share input/output embed
|
Sergey Edunov
|
|
6 years ago |
|
|
Better support for torch.no_grad (since volatile is deprecated)
|
Myle Ott
|
|
6 years ago |
|
|
Fix training
|
Myle Ott
|
|
6 years ago |
|
|
Save dictionary in model base classes
|
Myle Ott
|
|
6 years ago |
|
|
Fix gradient clipping when --clip-norm=0
|
Myle Ott
|
|
6 years ago |
|
|
Fix LearnedPositionalEmbedding
|
Myle Ott
|
|
6 years ago |
|
|
Move normalization of model output (e.g., via LSM) into model definition
|
Myle Ott
|
|
6 years ago |
|
|
Move positional embeddings into LearnedPositionalEmbedding module
|
Myle Ott
|
|
6 years ago |
|
|
Fix warning about deprecated `volatile` kwarg for Variables
|
Myle Ott
|
|
6 years ago |
|
|
Add option to SequenceGenerator to retain dropout
|
Myle Ott
|
|
6 years ago |
|
|
Add --max-sentences-valid to train.py
|
Myle Ott
|
|
6 years ago |
|
|
Streamline data formatting utils
|
Myle Ott
|
|
6 years ago |
Connect to our Discord channel
Models
Annotations