This PR introduces single-node fault tolerance for Tune.
## Previous behavior:
- Actors will be restarted without checking if resources are available. This can lead to problems if we lose resources.
## New behavior:
- RUNNING trials will be resumed on another node on a best effort basis (meaning they will run if resources available).
- If the cluster is saturated, RUNNING trials on that failed node will become PENDING and queued.
- During recovery, TrialSchedulers and SearchAlgorithms should receive notification of this (via `trial_runner.stop_trial`) so that they don’t wait/block for a trial that isn’t running.
Remaining questions:
- Should `last_result` be consistent during restore?
Yes; but not for earlier trials (trials that are yet to be checkpointed).
- Waiting for some PRs to merge first (#3239)
Closes#2851.
* Use F.softmax instead of a pointless network layer
Stateless functions should not be network layers.
* Use correct pytorch functions
* Rename argument name to out_size
Matches in_size and makes more sense.
* Fix shapes of tensors
Advantages and rewards both should be scalars, and therefore a list of them
should be 1D.
* Fmt
* replace deprecated function
* rm unnecessary Variable wrapper
* rm all use of torch Variables
Torch does this for us now.
* Ensure that values are flat list
* Fix shape error in conv nets
* fmt
* Fix shape errors
Reshaping the action before stepping in the env fixes a few errors.
* Add TODO
* Use correct filter size
Works when `self.config['model']['channel_major'] = True`.
* Add missing channel major
* Revert reshape of action
This should be handled by the agent or at least in a cleaner way that doesn't
break existing envs.
* Squeeze action
* Squeeze actions along first dimension
This should deal with some cases such as cartpole where actions are scalars
while leaving alone cases where actions are arrays (some robotics tasks).
* try adding pytorch tests
* typo
* fixup docker messages
* Fix A3C for some envs
Pendulum doesn't work since it's an edge case (expects singleton arrays, which
`.squeeze()` collapses to scalars).
* fmt
* nit flake
* small lint
* trying to fix jenkins tests
* comment out more tests
* remove pytorch stuff
* use non-monotonic clock (monotonic not supported on python 2.7)
* whitespace
* Test example applications in Jenkins.
* Fix default upload_dir argument for Algorithm class.
* Fix evolution strategies.
* Comment out policy gradient example which doesn't seem to work.
* Set --env-name for evolution strategies.
* attempt to build on travis using docker
* run tests in foreground
* add examples to travis tests
* test from current checkout
* attempt to fix docker version issues
* try build with xenial
* attempt docker upgrade
* avoid hang on configuration files
* matrix osx and linux w/ docker
* restore non-test docker builds
* fix typo
* tuning and cleanup
* add missing file
* comment cleanup
* Ray with Docker
* cleanup based on comments
* rename docker user to ray-user
* add examples docker image
* working toward reliable Docker devel image
* adjust ray-user uid for Linux builds on AWS
* update documentation
* reduced dependencies for examples
* updated Docker documentation
* experimental notice on developing with Docker
