Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 5, 2020

• Bump Tensorboard to be compatible with updated TF

  davidcotton

  committed
  0f6e4c8

• Merge pull request #2 from davidcotton/dependabot/pip/tensorflow-gpu-1.15.2

  Show description for 3cafe80
  davidcotton

  authored
  3cafe80

Commits on Jan 28, 2020

• Bump tensorflow-gpu from 1.14.0 to 1.15.2

  Show description for 2a3b8db
  dependabot[bot]

  authored
  2a3b8db

Commits on Jan 26, 2020

• Add restore to mcts_trainer.py

  davidcotton

  committed
  bc723fe

• Use episodes_this_iter as stop criteria to ensure min eval episodes completed

  davidcotton

  committed
  b2aa2f1

Commits on Jan 24, 2020

• Fix up MCTS trainer & evaluator to work with longer policy IDs

  davidcotton

  committed
  b006397

Commits on Jan 20, 2020

• Add resume flag to mcts_trainer.py, remove name==main guff

  davidcotton

  committed
  fda92b2

Commits on Jan 19, 2020

• Use softmax action selection for DQN/APEX

  davidcotton

  committed
  54105c4

Commits on Jan 15, 2020

• Fix for APEX

  davidcotton

  committed
  64eb56d

• Move plateau scripts into base dir to fix dependencies issue

  davidcotton

  committed
  8bccbcf

• Minor improvements for learning plateau experiments

  davidcotton

  committed
  54df25e

Commits on Jan 14, 2020

• Added dir for learning plateau experiments scripts

  davidcotton

  committed
  2ca2e0a

• Minor docs update for C4 env

  davidcotton

  committed
  3791ccc

• Moved standard PPO config into utils

  davidcotton

  committed
  e152ee0

Commits on Dec 13, 2019

• Add support for DQN & APEX in main.py

  davidcotton

  committed
  0580d7f

Commits on Dec 12, 2019

• Add config for DQN in mcts_trainer

  davidcotton

  committed
  6d8f882

Commits on Dec 4, 2019

• Simplify C4Env constructor

  davidcotton

  committed
  b692f9f

• C4 Game to return list of ints instead of numpy array

  davidcotton

  committed
  cad643a

• Improve mcts_trainer trial naming

  davidcotton

  committed
  6c6980d

Commits on Dec 3, 2019

• Add network type to trial name in mcts_trainer

  davidcotton

  committed
  64bfa88

Commits on Dec 2, 2019

• Rename models to MaskedActionsMLP/CNN

  davidcotton

  committed
  aacd917

• Remove Flattened & Square C4 envs

  davidcotton

  committed
  c58f354

• Update mcts_trainer stop criteria to 0.6 (80% win rate)

  davidcotton

  committed
  a1b9bb6

• Fix config loading in policies

  davidcotton

  committed
  e061fd6

• Remove logger

  davidcotton

  committed
  97aa97d

• Remove logger

  davidcotton

  committed
  26c41ae

• Commit assumption learner changes

  davidcotton

  committed
  7a55ec0

Commits on Nov 10, 2019

• Refactor out AssumptionMixin, AssumptionPPO policy template & build new AssumptionDQN policy template

  davidcotton

  committed
  b181f74

Commits on Nov 8, 2019

• Remove some commented code & comment out debug prints in ABL

  davidcotton

  committed
  400807a

• Remove default LR & gamma from ABL

  davidcotton

  committed
  1e6c812

• Refactor get_worker_config() to make it easier to support other algorithms than PPO

  davidcotton

  committed
  7c3994d

• Improve the reporting for C4 env.step() on invalid action

  davidcotton

  committed
  bdcabc2

• Fix up bugs with assumption learner

  davidcotton

  committed
  5dcac5a

Commits on Nov 7, 2019

• Added assumption-based learner

  davidcotton

  committed
  889d716

Commits on Oct 30, 2019

• Policy config readability improvements

  davidcotton

  committed
  e7ffb29

Pagination

PreviousNext