Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 22, 2019

• Add fallback for SLURM config

  Show description for bccfa7d
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  bccfa7d

  Copy full SHA for bccfa7d

• Move Masked LM components to legacy/ -- new ones are coming

  Show description for 47fd985
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  47fd985

  Copy full SHA for 47fd985

• Misc improvements to torch hub interface

  Show description for 9c89e88
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  9c89e88

  Copy full SHA for 9c89e88

Commits on Jul 21, 2019

• Update GPT-2 BPE

  Show description for 62b5498
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  62b5498

  Copy full SHA for 62b5498

• Default to mmap and infer dataset implementations automatically

  Show description for 5f78106
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  5f78106

  Copy full SHA for 5f78106

• Fix topp sampling issues (facebookresearch#882)

  Show description for 1f96d28
  

  Liang Wang

  authored and

  facebook-github-bot

  committed
  1f96d28

  Copy full SHA for 1f96d28

• Rename data.transforms -> data.encoders

  Show description for f812e52
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  f812e52

  Copy full SHA for f812e52

Commits on Jul 19, 2019

• Rename _load_model_ensemble -> load_model_ensemble_and_task

  Show description for 69d0f7f
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  69d0f7f

  Copy full SHA for 69d0f7f

• Allow not specifying --warmup-init-lr

  Show description for 7efde22
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  7efde22

  Copy full SHA for 7efde22

• Create standalone label_smoothed_nll_loss

  Show description for ffe53d6
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  ffe53d6

  Copy full SHA for ffe53d6

• Store task in the criterion base class

  Show description for c811e0e
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  c811e0e

  Copy full SHA for c811e0e

• Improve interactive generation (support --tokenizer and --bpe)

  Show description for 8af5554
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  8af5554

  Copy full SHA for 8af5554

• Switch to torch.nn.functional.gelu when available

  Show description for be5821b
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  be5821b

  Copy full SHA for be5821b

• v0.7.1 -> v0.7.2 (facebookresearch#891)

  Show description for b002d00
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  b002d00

  Copy full SHA for b002d00

Commits on Jul 17, 2019

• Support Latent Variable Model in base training (facebookresearch#879)

  Show description for 1f5b414
  

  cndn

  authored and

  facebook-github-bot

  committed
  1f5b414

  Copy full SHA for 1f5b414

• Nucleus (top-P) sampling (facebookresearch#710)

  Show description for e46b924
  

  Xing Zhou

  authored and

  facebook-github-bot

  committed
  e46b924

  Copy full SHA for e46b924

• Small bug fix for generation when batch_size is small

  Show description for 473389a
  

  jiajunshen

  authored and

  facebook-github-bot

  committed
  473389a

  Copy full SHA for 473389a

• Add suppress_defaults functionality to options parser (facebookresearch#723)

  Show description for 61e328c
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  61e328c

  Copy full SHA for 61e328c

• Forcing static shapes in loss computation (LSCE) (facebookresearch#876)

  Show description for 8db7b1c
  

  taylanbil

  authored and

  facebook-github-bot

  committed
  8db7b1c

  Copy full SHA for 8db7b1c

Commits on Jul 14, 2019

• removing tensor resizing in future_mask (facebookresearch#877)

  Show description for c38b1f9
  

  taylanbil

  authored and

  facebook-github-bot

  committed
  c38b1f9

  Copy full SHA for c38b1f9

Commits on Jul 11, 2019

• adding __getstate__ to fairseq_optimizer (facebookresearch#872)

  Show description for 397ba26
  

  taylanbil

  authored and

  facebook-github-bot

  committed
  397ba26

  Copy full SHA for 397ba26

• Small fix for trainer

  Show description for 0a4911f
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  0a4911f

  Copy full SHA for 0a4911f

Commits on Jul 10, 2019

• add __len__ for progress_bar (facebookresearch#871)

  Show description for 2bf7e75
  

  taylanbil

  authored and

  facebook-github-bot

  committed
  2bf7e75

  Copy full SHA for 2bf7e75

Commits on Jul 9, 2019

• Fix multilingual evaluation bug (facebookresearch#592)

  Show description for e433500
  

  pipibjc

  authored and

  facebook-github-bot

  committed
  e433500

  Copy full SHA for e433500

Commits on Jul 8, 2019

• Integrate torch.nn and fairseq MultiheadAttention (facebookresearch#772)

  Show description for 6d2e083
  

  Guanheng Zhang

  authored and

  facebook-github-bot

  committed
  6d2e083

  Copy full SHA for 6d2e083

Commits on Jul 6, 2019

• Added a comment to inform coders that pack_padded_sequence requires that padding must be on the right (facebookresearch#860)

  Show description for 417ecb4
  

  vineetk1

  authored and

  facebook-github-bot

  committed
  417ecb4

  Copy full SHA for 417ecb4

• Add specific compile flags for macOS (facebookresearch#862)

  Show description for cc292af
  

  louismartin

  authored and

  facebook-github-bot

  committed
  cc292af

  Copy full SHA for cc292af

Commits on Jul 4, 2019

• support streaming iterator

  Show description for 5c241c8
  

  Spencer Poff

  authored and

  facebook-github-bot

  committed
  5c241c8

  Copy full SHA for 5c241c8

Commits on Jul 2, 2019

• add --max-tokens-valid option for validation

  Show description for bccfddb
  

  Xutai Ma

  authored and

  facebook-github-bot

  committed
  bccfddb

  Copy full SHA for bccfddb

Commits on Jul 1, 2019

• Fixes checkpointing bug introduced in 89e077c

  Show description for 39bbc9a
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  39bbc9a

  Copy full SHA for 39bbc9a

• Make segment_labels optional

  Show description for b6d420c
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  b6d420c

  Copy full SHA for b6d420c

• Better distributed init for SLURM

  Show description for 1757ef6
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  1757ef6

  Copy full SHA for 1757ef6

Commits on Jun 30, 2019

• Fix docs (fixes facebookresearch#843)

  Show description for b594937
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  b594937

  Copy full SHA for b594937

• Add linear activation

  Show description for ca5b1da
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  ca5b1da

  Copy full SHA for ca5b1da

• Fix --end-learning-rate in polynomial LR schedule

  Show description for 0bac688
  

  Myle Ott

  authored and

  facebook-github-bot

  committed
  0bac688

  Copy full SHA for 0bac688

Pagination

Previous

Next