1. 14 May, 2018 3 commits
  2. 08 May, 2018 1 commit
  3. 07 May, 2018 2 commits
  4. 06 May, 2018 3 commits
  5. 04 May, 2018 1 commit
  6. 02 May, 2018 4 commits
  7. 01 May, 2018 1 commit
  8. 29 Apr, 2018 1 commit
  9. 25 Apr, 2018 5 commits
  10. 24 Apr, 2018 1 commit
  11. 20 Apr, 2018 3 commits
  12. 19 Apr, 2018 2 commits
  13. 18 Apr, 2018 6 commits
  14. 17 Apr, 2018 1 commit
  15. 16 Apr, 2018 1 commit
  16. 13 Apr, 2018 3 commits
  17. 11 Apr, 2018 1 commit
  18. 10 Apr, 2018 1 commit
    • Roman Sokolkov's avatar
      Disable additional shuffling while training · ffa40d63
      Roman Sokolkov authored
      Before this change dataset was shuffled once and
      then on every step random sample was returned.
      Problem is that this way we do not guarantee that
      all samples from dataset will be used.
      ffa40d63