Commits · bb3be24de31140ed9c6e36696b0017c8a0a034d0 · OpenDAS / Fairseq

01 Nov, 2017 1 commit
- Update README with note about Docker (#49) · bb3be24d
  Myle Ott authored Nov 01, 2017
  
  bb3be24d
24 Oct, 2017 1 commit
- Fix for building under clang: specify C++ build and use C++ linkage (#42) · 30953d8b
  James Reed authored Oct 24, 2017
  
  30953d8b
19 Oct, 2017 21 commits
- Merge pull request #33 from facebookresearch/oss-merge-internal · c83efd21
  Myle Ott authored Oct 19, 2017
```
Changes:
Add support for NCCL v2
Add support for additional optimizers
SequenceGenerator returns attention matrix
Misc bugfixes (e.g., fixes #32) and cleanup
```
  c83efd21
- Set seed after each epoch to improve consistency when resuming · 104cead1
  Myle Ott authored Oct 19, 2017
  
  104cead1
- Prevent math overflow when loss is too high · 8b4c45a2
  Louis Martin authored Oct 19, 2017
  
  8b4c45a2
- Fix handling of continuation tokens that precede <unk> in generate.py · ff2e8cf2
  Myle Ott authored Oct 17, 2017
  
  ff2e8cf2
- Fix language inference in generate.py · 8943fc78
  Myle Ott authored Oct 17, 2017
  
  8943fc78
- Simplify deps of build_model to only depend on dict (instead of dataset) · 84b82dc6
  Myle Ott authored Oct 17, 2017
  
  84b82dc6
- Add support for additional optimizers · d646a4a8
  Myle Ott authored Oct 17, 2017
  
  d646a4a8
- Refactor code in Tokenizer · cab76554
  Louis Martin authored Oct 12, 2017
  
  cab76554
- Refactor model saving/loading to be more reusable · eea50f38
  Myle Ott authored Oct 12, 2017
  
  eea50f38
- More flexible gradient normalization · 3f970086
  Myle Ott authored Oct 06, 2017
  
  3f970086
- Fix smoothed (sentence-level) BLEU calculation · 88a8bd42
  Myle Ott authored Oct 06, 2017
  
  88a8bd42
- Ignore generated files for temporal convolution tbc · d92ce54c
  Louis Martin authored Oct 06, 2017
  
  d92ce54c
- Add attention matrix to output of SequenceGenerator · 84754894
  Louis Martin authored Oct 03, 2017
  
  84754894
- Add support for NCCL v2 · 376c265f
  Myle Ott authored Oct 02, 2017
  
  376c265f
- Better logging from criterions · 8bafae2e
  Myle Ott authored Sep 28, 2017
  
  8bafae2e
- Add optimizer history to checkpoints (and rearrange criterions slightly) · e432459b
  Myle Ott authored Sep 27, 2017
  
  e432459b
- Allow --max-len-a to be a float · 48631f7a
  Myle Ott authored Sep 27, 2017
  
  48631f7a
- Don't save/restore convolutional layers in incremental inference · 813352e1
  Myle Ott authored Sep 27, 2017
  
  813352e1
- Fix flake8 warnings · cb0d7b2a
  Louis Martin authored Sep 25, 2017
  
  cb0d7b2a
- Support configurable BPE symbol · 7333d04d
  Myle Ott authored Sep 25, 2017
  
  7333d04d
- Move helper functions from generate.py to fairseq/dictionary.py · 59d599a2
  Myle Ott authored Sep 25, 2017
  
  59d599a2
17 Oct, 2017 1 commit
- Update En2Fr model · af86c1ac
  Sergey Edunov authored Oct 17, 2017
  
  af86c1ac
15 Oct, 2017 1 commit
- Fix --no-progress-bar option in generate.py (#115) · 2b482f63
  Myle Ott authored Oct 14, 2017
  
  2b482f63
13 Oct, 2017 2 commits
- Update README.md · 9c5fa5c5
  Myle Ott authored Oct 13, 2017
```
Don't suggest Miniconda (see #24)
```
  9c5fa5c5
- Update README.md · 0ccd81d4
  Myle Ott authored Oct 12, 2017
```
Update PyTorch install instructions
```
  0ccd81d4
11 Oct, 2017 5 commits
- Ignore invalid sentences in test and valid · 3f9b9838
  Sergey Edunov authored Oct 11, 2017
  
  3f9b9838
- Don't generate during training, add --quiet to generate.py · 8f058ea0
  Sergey Edunov authored Oct 11, 2017
  
  8f058ea0
- BPE transformation for IWSLT · a8260d52
  Sergey Edunov authored Oct 11, 2017
  
  a8260d52
- Fix call ordering to ATen addmm and sum (#22) · ae0c05d9
  Sam Gross authored Oct 11, 2017
  
  ae0c05d9
- Update progress_bar to be more robust to changes in tqdm (#21) · 7aba6084
  Myle Ott authored Oct 11, 2017
  
  7aba6084
02 Oct, 2017 1 commit
- Refactor PaddingCollater · 2ad58885
  Myle Ott authored Sep 27, 2017
  
  2ad58885
28 Sep, 2017 1 commit
- Fix handling of partially-empty initial batch (#11) · 4593ebfa
  Myle Ott authored Sep 28, 2017
  
  4593ebfa
26 Sep, 2017 1 commit
- Fix generation when vocabulary is small relative to beam size (fixes #7) · 03c4a716
  Myle Ott authored Sep 26, 2017
  
  03c4a716
24 Sep, 2017 1 commit
- Issue #2, Checking size attribute of dst when dst is None · 2d3161da
  Sergey Edunov authored Sep 24, 2017
  
  2d3161da
19 Sep, 2017 1 commit
- Better training support when GPUs are in "exclusive mode" · a6155337
  Myle Ott authored Sep 19, 2017
  
  a6155337
18 Sep, 2017 2 commits
- Small fix · a8bc4d0a
  Sergey Edunov authored Sep 18, 2017
  
  a8bc4d0a
- More fixes · c6de2190
  Sergey Edunov authored Sep 18, 2017
  
  c6de2190
15 Sep, 2017 1 commit
- Architecture settings and readme updates · a15acdb0
  Sergey Edunov authored Sep 15, 2017
  
  a15acdb0