Commits

Branch selector

 master

User selector

All users

All time

Commit history

Commits on Apr 30, 2018

• Merge pull request #759 from deepsense-ai/fix_data_generation

  Show description for b42e7ba
  lukaszkaiser

  authored
  b42e7ba

  Copy full SHA for b42e7ba

• Fixing data generation from simulated env.

  

  Błażej O

  committed
  8296aeb

  Copy full SHA for 8296aeb

Commits on Apr 26, 2018

• Merge pull request #751 from rsepassi/push

  Show description for ed9e3bd
  lukaszkaiser

  authored
  ed9e3bd

  Copy full SHA for ed9e3bd

• Add travis ignore for r_transformer_test

  Show description for 1027772
  

  Ryan Sepassi

  committed
  1027772

  Copy full SHA for 1027772

• Autoencoder improvements and related corrections; ordered and discrete residual MNIST samples look reasonable.

  Show description for d901097
  

  Lukasz Kaiser

  authored and

  Ryan Sepassi

  committed
  d901097

  Copy full SHA for d901097

• Add h5py to deps

  Show description for 25b101d
  

  Ryan Sepassi

  committed
  25b101d

  Copy full SHA for 25b101d

• Corrections to model-based RL code.

  Show description for cb31f9f
  

  Lukasz Kaiser

  authored and

  Ryan Sepassi

  committed
  cb31f9f

  Copy full SHA for cb31f9f

• Sort the losses_dict before passing it to the sum op, in order to make the tf graph deterministic.

  Show description for 456ea8b
  

  T2T Team

  authored and

  Ryan Sepassi

  committed
  456ea8b

  Copy full SHA for 456ea8b

• Internal merges

  Show description for 8461a03
  

  Ryan Sepassi

  committed
  8461a03

  Copy full SHA for 8461a03

• StringIO Py2/3 fix

  Show description for 5a55341
  

  Ryan Sepassi

  committed
  5a55341

  Copy full SHA for 5a55341

• don't crash if hparams.max_length is unspecified

  Show description for aa6ad69
  

  T2T Team

  authored and

  Ryan Sepassi

  committed
  aa6ad69

  Copy full SHA for aa6ad69

• Internal merges

  Show description for 6f17c7a
  

  Ryan Sepassi

  committed
  6f17c7a

  Copy full SHA for 6f17c7a

• Fix pending bug :(

  Show description for 4b705cb
  

  Niki Parmar

  authored and

  Ryan Sepassi

  committed
  4b705cb

  Copy full SHA for 4b705cb

• Release Wikipedia Summarization dataset

  Show description for 65c797f
  

  Ryan Sepassi

  committed
  65c797f

  Copy full SHA for 65c797f

• Implement span-positioned SquAD text problem.

  Show description for 4130e71
  

  T2T Team

  authored and

  Ryan Sepassi

  committed
  4130e71

  Copy full SHA for 4130e71

• Version bump

  Show description for 79114db
  

  Ryan Sepassi

  committed
  79114db

  Copy full SHA for 79114db

• Add support for PREDICT on TPU, tested with resnet50

  Show description for 432317a
  

  Ryan Sepassi

  committed
  432317a

  Copy full SHA for 432317a

• don't crash if hparams.max_length is unspecified

  Show description for 0ef2e3f
  

  nshazeer

  authored and

  Ryan Sepassi

  committed
  0ef2e3f

  Copy full SHA for 0ef2e3f

• faster implementation of relative attention for masked-self-attention case.

  Show description for 2983108
  

  nshazeer

  authored and

  Ryan Sepassi

  committed
  2983108

  Copy full SHA for 2983108

• Update ImportError logging in data_generators.all_problems

  Show description for 861235f
  

  Ryan Sepassi

  committed
  861235f

  Copy full SHA for 861235f

• Use TF 1.7 in Cloud TPU path, use gcloud beta, add --cloud_skip_confirmation

  Show description for 9301ef8
  

  Ryan Sepassi

  committed
  9301ef8

  Copy full SHA for 9301ef8

• Recurrent Transformers.

  Show description for 485beeb
  

  T2T Team

  authored and

  Ryan Sepassi

  committed
  485beeb

  Copy full SHA for 485beeb

• Create data_dir before vocab save

  Show description for b269683
  

  Ryan Sepassi

  committed
  b269683

  Copy full SHA for b269683

• Only make shuffle parallel if >1 file

  Show description for fec36fb
  

  Ryan Sepassi

  committed
  fec36fb

  Copy full SHA for fec36fb

• Shuffle in parallel

  Show description for f259410
  

  Ryan Sepassi

  committed
  f259410

  Copy full SHA for f259410

• Make base Transformer behave well on 2D data for baselines.

  Show description for f4dd38e
  

  Lukasz Kaiser

  authored and

  Ryan Sepassi

  committed
  f4dd38e

  Copy full SHA for f4dd38e

• Add note about output_dir (#747)

  

  Ankush Agarwal

  authored and

  rsepassi

  committed
  757b529

  Copy full SHA for 757b529

Commits on Apr 25, 2018

• Add ROC AUC (#737)

  

  jjtan

  authored and

  rsepassi

  committed
  42a7883

  Copy full SHA for 42a7883

• en-mk: update training and development dataset for en-mk (#738)

  

  stefan-it

  authored and

  rsepassi

  committed
  310b02b

  Copy full SHA for 310b02b

• Add English-Estonian (WMT2018) as new problem (#739)

  Show description for cd7e5e5
  

  stefan-it

  authored and

  rsepassi

  committed
  cd7e5e5

  Copy full SHA for cd7e5e5

• readme: fix link to unit tests for translation problems (#742)

  

  stefan-it

  authored and

  rsepassi

  committed
  331d17a

  Copy full SHA for 331d17a

• Minor pylint fixes (#746)

  Show description for 328d808
  

  stefan-it

  authored and

  rsepassi

  committed
  328d808

  Copy full SHA for 328d808

Commits on Apr 20, 2018

• Merge pull request #736 from rsepassi/push

  Show description for 99750c4
  lukaszkaiser

  authored
  99750c4

  Copy full SHA for 99750c4

• Rm all uses of xrange and fix decode_length for slow_greedy_infer

  Show description for 8cf5fa4
  

  Ryan Sepassi

  committed
  8cf5fa4

  Copy full SHA for 8cf5fa4

• Update text_encoder unicode utilities

  Show description for 347084c
  

  Ryan Sepassi

  committed
  347084c

  Copy full SHA for 347084c