Code

Imitation Learning#973

Code

Open

akashvelu wants to merge 70 commits intoi210_devflow-project/flow:i210_devfrom akash-daggerflow-project/flow:akash-daggerCopy head branch name to clipboard

+2,496-256

Commits

Commits on Mar 26, 2020

• dagger initial implementation

  akashvelu

  committed
  50b017a

• Changed environment to be single agent RL

  akashvelu

  committed
  1fc027d

• cleaned up code

  akashvelu

  committed
  d01aeb5

Commits on Mar 29, 2020

• check for None or Nan action before adding to buffer

  akashvelu

  committed
  920dd73

• Fixed dimension bug

  akashvelu

  committed
  630a100

Commits on Apr 13, 2020

• Added testing for imitation learning with dagger, more code cleanup

  akashvelu

  committed
  722f439

• Code cleanup, added testing/eval for imitation learning

  akashvelu

  committed
  e9d7634

• Moved file to within controller class

  akashvelu

  committed
  83a7887

• Renamed directory, code cleanup, evaluation script

  akashvelu

  committed
  c668336

• Delete dagger.iml

  akashvelu

  authored
  f54eebc

Commits on Apr 14, 2020

• initial multiagent imitation learning implementation

  akashvelu

  committed
  eb7b3a2

• Merge branch 'akash-dagger' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  4f1b2ad

• Merge branch 'i210_dev' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  c36f010

Commits on Apr 23, 2020

• Added multiagent capabilities for imitation learning

  akashvelu

  committed
  cb4cae8

Commits on May 1, 2020

• Multiagent changes, added stochastic policies

  akashvelu

  committed
  23e2ba3

Commits on May 4, 2020

• Bug fixes to stochastic policies and singlagent with multiple RL vehicles

  akashvelu

  committed
  f924d9c

• Delete Untitled.ipynb

  akashvelu

  authored
  0b08c33

• Removed files

  akashvelu

  committed
  21ee5ce

• Merge branch 'akash-dagger' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  75716d3

• minor cleanup

  akashvelu

  committed
  cffc33d

Commits on May 6, 2020

• Merge branch 'master' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  6c0c590

Commits on May 21, 2020

• Pulled updated changes

  akashvelu

  committed
  61f9a3a

Commits on May 23, 2020

• Bug fixes for stochastic policies

  akashvelu

  committed
  b4f844f

• Merge branch 'i210_dev' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  37f2c2e

Commits on May 26, 2020

• Ported to Keras, initial implementation of loading to RLLib

  akashvelu

  committed
  db0442b

• Merge branch 'i210_dev' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  39ad373

• Bug fixes for starting training from imitation model

  akashvelu

  committed
  ed065b3

• Minor cleanup

  akashvelu

  committed
  cc0aa32

• Minor cleanup

  akashvelu

  committed
  3a2e135

• Removed usage of rllib.utils.freamwork

  akashvelu

  committed
  288a1cf

Commits on May 27, 2020

• Changed location of h5 file

  akashvelu

  committed
  c1db60a

• Merge branch 'i210_dev' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  13a797c

• Function for ppo architecture

  akashvelu

  committed
  4b853b5

• Load weights for training in train.py

  akashvelu

  committed
  5c0923d

Commits on May 28, 2020

• Code structure changes

  akashvelu

  committed
  c785944

• Combine imitation and PPO training into one step

  akashvelu

  committed
  fef3a83

• Code cleanup

  akashvelu

  committed
  fd29e0f

• test with cluster

  akashvelu

  committed
  2b6cc08

• Merge i210_dev

  akashvelu

  committed
  8ef0179

Commits on May 29, 2020

• Bug fix

  akashvelu

  committed
  3dfafe1

• Minor cleanup

  akashvelu

  committed
  ba67961

Commits on Jun 2, 2020

• Load weights into rllib

  akashvelu

  committed
  bff9e47

• new merge

  akashvelu

  committed
  db29793

Commits on Jun 3, 2020

• Value function learning

  akashvelu

  committed
  d10f8e5

• Tensorboard plotting for loss

  akashvelu

  committed
  7fa3e3a

• Bug fixes

  akashvelu

  committed
  91fab74

Commits on Jun 4, 2020

• Code cleanup

  akashvelu

  committed
  d38839f

Commits on Jun 11, 2020

• Test file changes

  akashvelu

  committed
  81e8d6a

Commits on Jun 14, 2020

• Code cleanup

  akashvelu

  committed
  240cc05

Commits on Jun 17, 2020

• Flag reconfig

  akashvelu

  committed
  10bf24f

• flag cleanup

  akashvelu

  committed
  aa72b2e

• Reorganize method arguments

  akashvelu

  committed
  e209de2

• Argument reorganizing

  akashvelu

  committed
  5ce3c4d

Commits on Jun 18, 2020

• Cleanup and rearrange args

  akashvelu

  committed
  1aae8f8

• Custom PPO to log value function predictions:

  akashvelu

  committed
  f7451d0

• Merged i210 dev

  akashvelu

  committed
  9519c1f

• Cleanup to train.py

  akashvelu

  committed
  2d96460

Commits on Jun 19, 2020

• add imitation custom models

  akashvelu

  committed
  18c0d9e

• code cleanup

  akashvelu

  committed
  024cb93

• Merge branch 'i210_dev' of https://github.com/flow-project/flow into akash-dagger

  akashvelu

  committed
  16a9ced

• custom ppo for vf plotting edits

  akashvelu

  committed
  885ab6f

Commits on Jun 20, 2020

• i210 imitation model file

  akashvelu

  committed
  85fdd63

Commits on Jun 25, 2020

• Code cleanup

  akashvelu

  committed
  9dd65c8

• test files synced to i210_dev

  akashvelu

  committed
  739c2ca

Commits on Jun 26, 2020

• Cleanup code

  akashvelu

  committed
  ddce32e

• Handle case with vehicle in no-control edge

  akashvelu

  committed
  4e6302e

Commits on Jul 10, 2020

• Add learning rate as a parameter, override import_from_h5 method using setattr

  akashvelu

  committed
  29eb5a0

Commits on Jul 13, 2020

• Move imitation to algorithms folder

  akashvelu

  committed
  6c68800

• Merge i210 into branch

  akashvelu

  committed
  d73612f

• Revert model architecture and # rollouts to previous defaults

  akashvelu

  committed
  6aca7c5