Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Sep 24, 2018

• Merge pull request facebookresearch#287 from pytorch/oss-master

  Show description for 25524f1
  edunov

  authored
  25524f1

• Update readme with WMT'18 model (facebookresearch#433)

  

  edunov

  authored and

  Sergey Edunov

  committed
  86b5cfe

Commits on Sep 18, 2018

• Merge pull request facebookresearch#279 from pytorch/oss-master

  Show description for 5d15085
  edunov

  authored
  5d15085

• Readme fix

  

  Sergey Edunov

  committed
  74b3f1e

• Fix docs

  

  Sergey Edunov

  committed
  fe2d158

• Fix readme

  

  Sergey Edunov

  committed
  5d944b0

Commits on Sep 7, 2018

• modified stories readme to include sample preprocessing code to split stories to 1k tokens

  huihuifan

  committed
  5d00e8e

Commits on Sep 4, 2018

• Update documentation

  myleott

  committed
  4a47b88

Commits on Sep 3, 2018

• Add documentation

  myleott

  committed
  6381cc9

• Misc changes to simplify upcoming tutorial

  myleott

  committed
  0e101e9

• Test max_positions

  myleott

  committed
  d473620

• fix cosine lr sched for t_mult=1 with warmup

  

  alexeib

  authored and

  myleott

  committed
  dfd7771

• Further generalize EpochBatchIterator and move iterators into new file

  myleott

  committed
  0a7f9e6

• Fix comment

  myleott

  committed
  75f6ba0

• fix max_positions comparison

  

  alexeib

  authored and

  myleott

  committed
  b3cd43b

• Clean up FairseqTask so that it's easier to extend/add new tasks

  myleott

  committed
  2e507d3

• Add --upsample-primary

  myleott

  committed
  6296de8

• Add adaptive softmax changes for lstm model

  

  Li Zhao

  authored and

  myleott

  committed
  5852d3a

• dont send dummy batch when reloading from checkpoint

  Show description for 343819f
  

  alexeib

  authored and

  myleott

  committed
  343819f

• Fix FP16 version comparison

  myleott

  committed
  b9956a6

• Merge internal changes

  myleott

  committed
  753935e

• word stats in eval_lm

  

  alexeib

  authored and

  myleott

  committed
  c7c567a

• Old checkpoints can't be loaded because of a new meter

  

  Sergey Edunov

  authored and

  myleott

  committed
  c9b800d

• Add training wall time meter

  myleott

  committed
  9c10278

• disable final layer norm for transformer decoder as it makes things worse

  

  alexeib

  authored and

  myleott

  committed
  f84e1ed

• Fix adaptive softmax cutoff comment

  

  louismartin

  authored and

  myleott

  committed
  81ba4c4

• Remove --normalization-constant from fconv

  myleott

  committed
  47f095f

• Diverse Beam Search

  myleott

  committed
  8c0ca1a

• add warmup support back to cosine lr sched (important for mt)

  

  alexeib

  authored and

  myleott

  committed
  ba9f32c

• Warn when using FP16 on pre-Volta GPUs

  myleott

  committed
  8d6665f

• Move read_binarized.py to scripts/

  myleott

  committed
  e742219

• script to read binarized data

  

  alexeib

  authored and

  myleott

  committed
  cd6590b

• Increase max buffer size in all_gather_list

  myleott

  committed
  5b533aa

• fix tests

  

  alexeib

  authored and

  myleott

  committed
  f1d81db

• Reset gnorm after each epoch

  

  Sergey Edunov

  authored and

  myleott

  committed
  97a6b13

Pagination

PreviousNext