Commits

Branch selector

 develop

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 13, 2016

• updated keras and theano; fixed resulting bugs and warnings

  Show description for 9da84ee
  

  wrongu

  committed
  9da84ee

  Copy full SHA for 9da84ee

• whitespace fix in benchmarks

  

  wrongu

  committed
  082a6c2

  Copy full SHA for 082a6c2

• updated keras and theano; fixed resulting bugs and warnings

  

  wrongu

  committed
  428fac0

  Copy full SHA for 428fac0

• variable rename

  Show description for e96c85a
  IAmARobot

  committed
  e96c85a

  Copy full SHA for e96c85a

• Added supervised policy training benchmark

  Show description for 8162931
  IAmARobot

  committed
  8162931

  Copy full SHA for 8162931

Commits on Apr 12, 2016

• fixed weight loading when resuming supervised training

  

  wrongu

  committed
  fa3a4a2

  Copy full SHA for fa3a4a2

Commits on Apr 11, 2016

• binary_crossentropy --> categorical_crossentropy

  Show description for 99b98e6
  

  wrongu

  committed
  99b98e6

  Copy full SHA for 99b98e6

• fixed test_supervised_policy_trainer weights file as per last commit

  

  wrongu

  committed
  2aba4cb

  Copy full SHA for 2aba4cb

• reset readme - accidental commit...

  

  wrongu

  committed
  f3e502f

  Copy full SHA for f3e502f

• supervised_policy_trainer saves 5-digit weights filenames

  

  wrongu

  committed
  1fd9a98

  Copy full SHA for 1fd9a98

Commits on Apr 10, 2016

• readme update

  

  wrongu

  committed
  9dd393d

  Copy full SHA for 9dd393d

Commits on Apr 8, 2016

• updated supervised training for flattened 1D output

  

  wrongu

  committed
  15d0877

  Copy full SHA for 15d0877

• CNNPolicy using Flattened board; cleaned up forward(); closes #81

  Show description for a1921cf
  

  wrongu

  committed
  a1921cf

  Copy full SHA for a1921cf

Commits on Apr 6, 2016

• save accuracy as well in supervised training. fixes #87

  

  wrongu

  committed
  e4d04ba

  Copy full SHA for e4d04ba

Commits on Apr 5, 2016

• Merge pull request #86 from thouis/opponent_capture_size_fix

  Show description for fc95c72
  wrongu

  committed
  fc95c72

  Copy full SHA for fc95c72

• check for opponent stones for capture size

  thouis

  committed
  4ba2d5a

  Copy full SHA for 4ba2d5a

• size parameter needed to be an int. Fixes #82

  

  wrongu

  committed
  6ffee5d

  Copy full SHA for 6ffee5d

Commits on Apr 3, 2016

• recursion test for eyes

  

  wrongu

  committed
  2bc6d9c

  Copy full SHA for 2bc6d9c

Commits on Apr 2, 2016

• Merge pull request #79 from Rochester-NRT/feature-fix

  Show description for 380bfda
  wrongu

  committed
  380bfda

  Copy full SHA for 380bfda

• whitespace fixes

  

  wrongu

  committed
  c04f5e0

  Copy full SHA for c04f5e0

• fixed preprocessing_benchmark according to new hdf5 game_converter

  Show description for 35361da
  

  wrongu

  committed
  35361da

  Copy full SHA for 35361da

Commits on Apr 1, 2016

• true vs false eye test in test_gamestate

  

  wrongu

  committed
  7884138

  Copy full SHA for 7884138

• GameState.is_eye --> is_eyeish, and the new is_eye does full check

  

  wrongu

  committed
  3eb86fa

  Copy full SHA for 3eb86fa

• fixed off-by-one in liberties_after feature

  Show description for b6b6b91
  

  wrongu

  committed
  b6b6b91

  Copy full SHA for b6b6b91

• updated comment in get_capture_size

  

  wrongu

  committed
  57ec07c

  Copy full SHA for 57ec07c

• turns_since feature fixed for captured stones

  

  wrongu

  committed
  76529f4

  Copy full SHA for 76529f4

• 'ones' feature fixed

  Show description for 141ff92
  

  wrongu

  committed
  141ff92

  Copy full SHA for 141ff92

Commits on Mar 31, 2016

• Merge pull request #77 from lilin00000/develop

  Show description for 43a0b63
  wrongu

  committed
  43a0b63

  Copy full SHA for 43a0b63

• modify AlphaGo/preprocessing/game_converter.py to avoid WindowsError: [Error 32] in Windows platform when testing

  

  Li Lin

  committed
  4ec561b

  Copy full SHA for 4ec561b

• add src/ to .gitignore

  

  Li Lin

  committed
  a974a71

  Copy full SHA for a974a71

Commits on Mar 30, 2016

• added --epoch-length argument to supervised training (default size of training set)

  

  wrongu

  committed
  65ac912

  Copy full SHA for 65ac912

• supervised trainer saves all logs to metadata and is safe if no validation data

  

  wrongu

  committed
  e52a1af

  Copy full SHA for e52a1af

Commits on Mar 28, 2016

• Update README.md

  wrongu

  committed
  a05b2af

  Copy full SHA for a05b2af

• Merge pull request #71 from Rochester-NRT/hdf5

  Show description for e11186e
  wrongu

  committed
  e11186e

  Copy full SHA for e11186e

• game_converter and supervised_policy_trainer tests updated

  Show description for 2a511a0
  

  wrongu

  committed
  2a511a0

  Copy full SHA for 2a511a0

Pagination

Previous

Next