* Cleaner and more correct treatment of agent states in rollout.py
* support lstm_use_prev_action_reward in rollout.py
* Linter.
* appease flake8
* Use _DUMMY_AGENT_ID instead of 0.
* All agents have a policy_agent_mapping.
Reset the mapping cache at the start of each episode.
* Update rollout.py
* Fix rollout.py for single-agent envs.
* Use agent_id, not policy_id.
* Make full use of node
implement local node
fix bugs mentioned in comments
* Add more tests
* Use more specific exception handling
* fix, lint
* fix for py2.x
* Added support for running different docker containers on clusters
* Remove node specific container names
* Keep old options and expand with node specific configuration
* Optimized imports
* Changed docker fields for autoscaler
* Auto reformat
* Updated comments
* Updated condition
* Run linter
* Updated example
* Changed condition for docker images, updated examples
* Removed duplicate line
* Fixed setup_commands
* Update autoscaler.py
* fix_better_image
## What do these changes do?
This PR prepares for future version 0.2.0 of `nevergrad`, in which each suggestion is a `Candidate` instance having fields `args` and `kwargs` instead of being a `np.ndarray`. The proposed changes are compatible with all versions of `nevergrad` (manually tested with `nevergrad_example.py` on both `master` and current version `v0.1.6`).
See `nevergrad`'s [CHANGELOG](https://github.com/facebookresearch/nevergrad/blob/master/CHANGELOG.md) for more information on the change.
## Related issue number
None
## Linter
- [x] I've run `scripts/format.sh` to lint the changes in this PR.
## What do these changes do?
Add documentation for the `--output` flag for ls / lsx in the Tune CLI.
## Related issue number
Closes#4511
## Linter
- [x] I've run `scripts/format.sh` to lint the changes in this PR.
* Travis CI: Do not hard-code Trusty, it EOLs this month
Do not hard-code __Trusty__ because it reaches its end-of-life this month.
https://wiki.ubuntu.com/Releases
* Update .travis.yml
* Moved clip_action into policy_graph; Clip actions in compute_single_action
* Update policy_graph.py
* Changed formatting
* Updated codebase for convencience
* Adding support for resource variables
Currently resource variable go undetected by the `TensorFlowVariables` since they do not use the same ops for reading values. This change should fix this until a more robust solution is implemented.
* fix varhandle
