Skip to content

GitLab

Switch branch/tag
  1. 04 May, 2019 2 commits
  3. 03 May, 2019 2 commits
  5. 02 May, 2019 5 commits
    • Peng-Jen Chen's avatar
      Make learned positional embedding optional · 39264559
      Peng-Jen Chen authored 
      Summary:
- Add learned positional embedding binary flag to masked LM model.
- Add base arch config for masked LM model which sets all the binary parameters to False. Otherwise some of the binary flag parameters will always be override by config in `xlm_architecture` (e.g. encoder_learned_pos)

Reviewed By: liezl200

Differential Revision: D15054487

fbshipit-source-id: d78827f352b9160a89c9dc4f45b9fce15a2f234d
      39264559
    • Myle Ott's avatar
      Move distributed_init into DistributedFairseqModel (#687) · 34726d56
      Myle Ott authored 
      Summary:
This should make rendezvous happen as lazily as possible.
Pull Request resolved: https://github.com/pytorch/fairseq/pull/687

Differential Revision: D15151145

Pulled By: myleott

fbshipit-source-id: d70816a85414c5d509a6b12e2b339b4736db2c88
      34726d56
    • Myle Ott's avatar
      Validate on all sets based on --save-interval-updates · fb18be00
      Myle Ott authored 
      Summary: Pull Request resolved: https://github.com/pytorch/fairseq/pull/693

Differential Revision: D15174831

fbshipit-source-id: 98688b1269ead5694e5116659ff64507d3c0d1c0
      fb18be00
    • Myle Ott's avatar
      Fix inconsistent gradient check · 4a30a5f6
      Myle Ott authored 
      Summary: Pull Request resolved: https://github.com/pytorch/fairseq/pull/692

Differential Revision: D15174954

fbshipit-source-id: 1a7bff9aeed3e2cc658577be9d79e8c9f72314c2
      4a30a5f6
    • Kritika Singh's avatar
      Make CTC work with more encoder-only models · ffc9c8cc
      Kritika Singh authored 
      Summary:
Changes include:
1. Added get_normalized_probabilities to the encoder-only base class FairseqEncoderModel
2. Made CTCCriterion work for both batch_first (LSTMSubsampleEncoderModel) and batch_second (LSTMEncoderOnly) encoder types
3. Added tests for different encoder and CTC combinations.

TODO:
CTC still doesn't work for VGGLSTMEncoderModel so I have disabled that. Will debug and send out fix in another diff.

Reviewed By: jay-mahadeokar

Differential Revision: D15158818

fbshipit-source-id: acb484bad705c937d676d2c3dcde3e3562d68ed9
      ffc9c8cc
  7. 01 May, 2019 5 commits
  9. 30 Apr, 2019 6 commits
  11. 29 Apr, 2019 2 commits
  13. 27 Apr, 2019 2 commits
  15. 26 Apr, 2019 1 commit
  17. 25 Apr, 2019 6 commits
  19. 24 Apr, 2019 1 commit
  21. 22 Apr, 2019 2 commits
    • Max Ryabinin's avatar
      Fix generation with --no-early-stop (#627) · fa52d202
      Max Ryabinin authored 
      Summary:
Because the size of `unfinalized_scores` is equal to current `bsz` and not initial batch size, we need to index it by `unfin_idx` instead of `sent` in `is_finished`.
Fixes #588.
Pull Request resolved: https://github.com/pytorch/fairseq/pull/627

Differential Revision: D15034641

Pulled By: myleott

fbshipit-source-id: 2638e68e877ae01256cac7d8e69b5b7fec8f7017
      fa52d202
    • Yongqiang Wang's avatar
      reduce memory footprint for average_checkpoints (#647) · d63477e1
      Yongqiang Wang authored 
      Summary:
Pull Request resolved: https://github.com/pytorch/fairseq/pull/647

the current implementation of average_checkpoints requires loading all
the model parameters into memory and then do the averaging. To average large
models (e.g., transformer) over a large number of checkpoints (e.g., >50),
it may require over 100GB memory.

Loading all the parameters is not necessary, as we know the number of models in advance.

Reviewed By: skritika

Differential Revision: D15027513

fbshipit-source-id: 0afe37c9a031a9ab0f1e78844a37be49ec5f76f1
      d63477e1
  23. 17 Apr, 2019 3 commits
  25. 16 Apr, 2019 1 commit
  27. 15 Apr, 2019 2 commits