Commits

Branch selector

 develop

User selector

All users

Datepicker

All time

Commit History

Commits on May 24, 2016

• fixed is_legal caching bug when include_eyes changes

  

  wrongu

  committed
  b6bf819

  Copy full SHA for b6bf819

• stone_age feature in GameState to make turns_since feature faster

  

  wrongu

  committed
  520b5bd

  Copy full SHA for 520b5bd

Commits on May 21, 2016

• "legal" feature added to preprocessing.py

  

  wrongu

  committed
  b1f0771

  Copy full SHA for b1f0771

• Merge pull request #112 from wrongu/nn_refactor

  Show description for 876871f
  wrongu

  committed
  876871f

  Copy full SHA for 876871f

• cleaner introspection with a NeuralNetBase registration decorator

  

  wrongu

  committed
  7e81f19

  Copy full SHA for 7e81f19

• NeuralNetBase class in nn_util handling generic policy/value things

  

  wrongu

  committed
  f08d52d

  Copy full SHA for f08d52d

Commits on May 20, 2016

• A few caching optimizations to go.py

  Show description for baa2c81
  

  wrongu

  committed
  baa2c81

  Copy full SHA for baa2c81

Commits on May 19, 2016

• benchmark for RL policy trainer

  

  wrongu

  committed
  e94f924

  Copy full SHA for e94f924

• fixed typo in reinforcement_policy_trainer: oppenents-->opponents

  

  wrongu

  committed
  6e77710

  Copy full SHA for 6e77710

• reinforcement_policy_trainer takes policy network's board size into account

  

  wrongu

  committed
  069b489

  Copy full SHA for 069b489

Commits on May 14, 2016

• fixed loading Bias layer using Keras' model_from_json

  

  wrongu

  committed
  79f4935

  Copy full SHA for 79f4935

• adding Bias layer to policy networks

  

  wrongu

  committed
  010a96c

  Copy full SHA for 010a96c

• updated ResnetPolicy to use BatchNormalization before ReLU

  

  wrongu

  committed
  6c592e7

  Copy full SHA for 6c592e7

• miniscule typo

  

  wrongu

  committed
  b40532c

  Copy full SHA for b40532c

• policy.py able to support Dropout and BatchNormalization

  Show description for db405c0
  

  wrongu

  committed
  db405c0

  Copy full SHA for db405c0

• (re) fixed initial weights metadata in reinforcement_policy_trainer

  

  wrongu

  committed
  840075c

  Copy full SHA for 840075c

Commits on May 13, 2016

• fixed to inconsequential typos, reinforcement_policy_trainer.py

  

  wrongu

  committed
  a6d51ca

  Copy full SHA for a6d51ca

• Merge pull request #108 from Rochester-NRT/reinforce

  Show description for 1fb0251
  wrongu

  committed
  1fb0251

  Copy full SHA for 1fb0251

• test of reinforcement_policy_trainer

  

  wrongu

  committed
  691a627

  Copy full SHA for 691a627

• reinforcement_policy_trainer interface updates, more like supervised_..

  

  wrongu

  committed
  cab9fde

  Copy full SHA for cab9fde

• whitespace/formatting in reinforcement_policy_trainer

  

  wrongu

  committed
  cccdd9a

  Copy full SHA for cccdd9a

• Merge branch 'develop' into reinforce

  

  wrongu

  committed
  a3c6318

  Copy full SHA for a3c6318

• Work continued on RL policy trainer

  

  ttrine

  authored and

  wrongu

  committed
  84acfa5

  Copy full SHA for 84acfa5

• policy.py save/load saves class as well

  

  wrongu

  committed
  2cd7a07

  Copy full SHA for 2cd7a07

Commits on May 11, 2016

• Merge pull request #105 from Rochester-NRT/resnet

  Show description for fd30639
  wrongu

  committed
  fd30639

  Copy full SHA for fd30639

• ResnetPolicy added to policy.py

  

  wrongu

  committed
  41d46d7

  Copy full SHA for 41d46d7

Commits on May 10, 2016

• slight modification to sgf_iter_states to handle AB and AW-only files

  

  wrongu

  committed
  dee1532

  Copy full SHA for dee1532

• Merge branch 'feature-fix' into develop

  

  wrongu

  committed
  e2e2c2a

  Copy full SHA for e2e2c2a

• fixes feature bug when stones captured

  Show description for fc80a88
  

  wrongu

  committed
  fc80a88

  Copy full SHA for fc80a88

• preprocessing tests to expose bugs in features after capture

  

  wrongu

  committed
  2163f90

  Copy full SHA for 2163f90

• comment typo

  

  wrongu

  committed
  045464e

  Copy full SHA for 045464e

Commits on May 9, 2016

• Python3 changes in game converter files merged from branch 'python3_changes' in thouis/AlphaGo

  

  wrongu

  committed
  a727763

  Copy full SHA for a727763

Commits on May 6, 2016

• fixes bug in usage of one_hot_action, supervised training

  Show description for 2bfce2b
  

  wrongu

  committed
  2bfce2b

  Copy full SHA for 2bfce2b

• rotations/reflections set by command line in training

  

  wrongu

  committed
  963fca4

  Copy full SHA for 963fca4

Commits on Apr 30, 2016

• Merge pull request #103 from BeierZhu/patch-2

  Show description for 134adb1
  wrongu

  committed
  134adb1

  Copy full SHA for 134adb1

Pagination

Previous

Next