Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on May 7, 2019

• More specific error messages clarifying the "two outer dimensions" error: explains why the output of trajectory.from_transition() cannot be used with agent.train().

  Show description for 3a64c31
  

  nealwu

  authored and

  copybara-github

  committed
  3a64c31

  Copy full SHA for 3a64c31

Commits on May 3, 2019

• When scaling an action distribution, ensure that the distribution matches the action spec bounds.

  Show description for 21e8129
  

  bkungfoo

  authored and

  copybara-github

  committed
  21e8129

  Copy full SHA for 21e8129

• Internal Change

  Show description for 734021a
  

  kbanoop

  authored and

  copybara-github

  committed
  734021a

  Copy full SHA for 734021a

Commits on May 2, 2019

• Add value estimation baseline to reinforce agent

  Show description for 49a8409
  

  kuanghuei

  authored and

  copybara-github

  committed
  49a8409

  Copy full SHA for 49a8409

• Fix TF agent test failures after change that switched to using V2 name scope in Keras in V2.

  Show description for 6bd5b4c
  

  TF-Agents Team

  authored and

  copybara-github

  committed
  6bd5b4c

  Copy full SHA for 6bd5b4c

Commits on Apr 30, 2019

• Remove pinned tfp nightly version.

  Show description for 414cb0a
  

  Oscar Ramirez

  committed
  414cb0a

  Copy full SHA for 414cb0a

• Enables passing a callable for epsilon greedy and using, for example, tf.train.exponential_decay in eager mode. See also cl/235273088, which added the same functionality to the BoltzmannPolicy.

  Show description for 55c3d38
  

  TF-Agents Team

  authored and

  Oscar Ramirez

  committed
  55c3d38

  Copy full SHA for 55c3d38

Commits on Apr 29, 2019

• Add optional saving of train_step to policy_saver.

  Show description for fb8d330
  

  Oscar Ramirez

  authored and

  copybara-github

  committed
  fb8d330

  Copy full SHA for fb8d330

Commits on Apr 26, 2019

• [TF-Agents] Initial bits of support for CompositeTensor in TF-Agents.

  Show description for fa40049
  

  ebrevdo

  authored and

  copybara-github

  committed
  fa40049

  Copy full SHA for fa40049

• Remove trailing spaces

  Show description for 45d806f
  

  kuanghuei

  authored and

  copybara-github

  committed
  45d806f

  Copy full SHA for 45d806f

• Add/update init files

  Show description for fa5cae1
  

  kbanoop

  authored and

  copybara-github

  committed
  fa5cae1

  Copy full SHA for fa5cae1

• Add tutorial colab for REINFORCE

  Show description for d58743a
  

  TF-Agents Team

  authored and

  copybara-github

  committed
  d58743a

  Copy full SHA for d58743a

• tests_release: Remove tf-nightly version pin.

  Show description for b7469ea
  

  samfishman

  authored and

  copybara-github

  committed
  b7469ea

  Copy full SHA for b7469ea

• Update code to use tf.math.mod

  Show description for 75aaf3d
  

  Oscar Ramirez

  authored and

  copybara-github

  committed
  75aaf3d

  Copy full SHA for 75aaf3d

Commits on Apr 24, 2019

• QPolicy maps action returned within action_spec bounds provided by the user.

  Show description for 11c3f07
  

  vcarbune

  authored and

  copybara-github

  committed
  11c3f07

  Copy full SHA for 11c3f07

• Add ShiftedCategorical to simplify honoring user-provided action specs.

  Show description for 4d04a20
  

  vcarbune

  authored and

  copybara-github

  committed
  4d04a20

  Copy full SHA for 4d04a20

Commits on Apr 18, 2019

• Fix sac_agent debug summary bug w/ entropy.

  Show description for 8231773
  

  TF-Agents Team

  authored and

  copybara-github

  committed
  8231773

  Copy full SHA for 8231773

Commits on Apr 16, 2019

• Normal projection network initializes the stddev projection layer with the given std_initializer_value argument and rename to the std_bias_initializer_value.

  Show description for 9aaa7de
  

  TF-Agents Team

  authored and

  copybara-github

  committed
  9aaa7de

  Copy full SHA for 9aaa7de

• Store and load experience using TFRecord files.

  Show description for fbc111a
  

  Hugo Ponte

  authored and

  copybara-github

  committed
  fbc111a

  Copy full SHA for fbc111a

Commits on Apr 15, 2019

• Update tf-nightly version in test.

  Show description for 442e05f
  

  samfishman

  authored and

  copybara-github

  committed
  442e05f

  Copy full SHA for 442e05f

• Add support for emitting log-probs in TfPolicy via a constructor argument.

  Show description for 748bede
  

  vcarbune

  authored and

  copybara-github

  committed
  748bede

  Copy full SHA for 748bede

Commits on Apr 13, 2019

• Enables using SAC with stateful networks and discrete action spaces. Also makes train_sequence_length configurable to enable training on entire episodes.

  Show description for 7725a7e
  

  TF-Agents Team

  authored and

  copybara-github

  committed
  7725a7e

  Copy full SHA for 7725a7e

Commits on Apr 12, 2019

• Move time_step, policy_step and trajectory to trajectories/

  Show description for 4abdace
  

  kbanoop

  authored and

  copybara-github

  committed
  4abdace

  Copy full SHA for 4abdace

Commits on Apr 11, 2019

• Switch to wrapt for trackable dict data structures instead of subclassing collections.Mapping

  Show description for 184e87f
  

  TF-Agents Team

  authored and

  copybara-github

  committed
  184e87f

  Copy full SHA for 184e87f

• Move metric_utils to eval/

  Show description for f2276c6
  

  kbanoop

  authored and

  copybara-github

  committed
  f2276c6

  Copy full SHA for f2276c6

Commits on Apr 10, 2019

• TD3: Add actor_update_period to control actor optimization frequency.

  Show description for 328b576
  

  samfishman

  authored and

  copybara-github

  committed
  328b576

  Copy full SHA for 328b576

• td3: Use Gaussian noise instead of OU noise for exploration.

  Show description for 92a71e3
  

  samfishman

  authored and

  copybara-github

  committed
  92a71e3

  Copy full SHA for 92a71e3

• Move q_network from agents/dqn/ to networks/

  Show description for 92b116b
  

  kbanoop

  authored and

  copybara-github

  committed
  92b116b

  Copy full SHA for 92b116b

• Merge pull request tensorflow#72 from npfp:bugfix-tensorflow#69

  Show description for 53f732b
  copybara-github

  committed
  53f732b

  Copy full SHA for 53f732b

Commits on Apr 9, 2019

• Add a unit test for default batch* properties.

  npfp

  authored
  dd0da02

  Copy full SHA for dd0da02

• Default the batch* properties to super

  Show description for 545d2a7
  npfp

  authored
  545d2a7

  Copy full SHA for 545d2a7

Commits on Apr 8, 2019

• Move numpy_store from replay_buffers/ to utils/

  Show description for 0d4ef35
  

  kbanoop

  authored and

  copybara-github

  committed
  0d4ef35

  Copy full SHA for 0d4ef35

• Fix a bug where batch* properties were not forwarded to the wrapped env.

  npfp

  committed
  c58ae8f

  Copy full SHA for c58ae8f

Commits on Apr 5, 2019

• Rename variable from tf_agent to agent in train_eval_atari.py to prevent confusion with tf_agent.py.

  Show description for 21ae8a6
  

  nealwu

  authored and

  copybara-github

  committed
  21ae8a6

  Copy full SHA for 21ae8a6

• [TF-Agents] Remove some problematic control dependencies.

  Show description for ad652a1
  

  ebrevdo

  authored and

  copybara-github

  committed
  ad652a1

  Copy full SHA for ad652a1

Pagination

Previous

Next