* add marvil policy graph
* fix typo
* add offline optimizer and enable running marwil
* fix loss function
* add maintaining the moving average of advantage norm
* use sync replay optimizer for unifying
* remove offline optimizer and use sync replay optimizer
* format by yapf
* add imitation learning objective
* fix according to eric's review
* format by yapf
* revise
* add test data
* marwil
* Refactor code about ray.ObjectID.
* remove from_random and use nil_id instead of constructor
* remove id() in hash
* Lint and fix
* Change driver id to ObjectID
* Replace binary_to_hex(ObjectID.id()) to ObjectID.hex()
* Implement Node class and move most of services.py into it.
* Wait for nodes as they are added to the cluster.
* Fix Redis authentication bug.
* Fix bug in client table ordering.
* Address comments.
* Kill raylet before plasma store in test.
* Minor
## What do these changes do?
Adds 2 commands to the CLI that take in an autoscaler config:
1. Kill a random ray node in the cluster.
2. Get all the worker node IP addresses.
These commands are both for testing and are not recommended for normal use.
## Related issue number
Closes#3685.
* Convert UniqueID::nil() to a constructor
* Cleanup actor handle pickling code
* Add new actor handles to the task spec
* Pass in new actor handles
* Add new handles to the actor registration
* Regression test for actor handle forking and GC
* lint and doc
* Handle pickled actor handles in the backend and some refactoring
* Add regression test for dummy object GC and pickled actor handles
* Check for duplicate actor tasks on submission
* Regression test for forking twice, fix failed named actor leak
* Fix bug for forking twice
* lint
* Revert "Fix bug for forking twice"
This reverts commit 3da85e59d401e53606c2e37ffbebcc8653ff27ac.
* Add new actor handles when task is assigned, not finished
* Remove comment
* remove UniqueID()
* Updates
* update
* fix
* fix java
* fixes
* fix
1. Fix the problem of duplicated stored logs.
2. Save log whose level is higher than severity_threshold, not only with severity_threshold.
3. Fix a `log_dir` bug: storing logs in a wrong path.
* Separate out functionality for querying client table and improve cluster.wait_for_nodes() API.
* Linting
* Add back logging statements.
* info -> debug
## What do these changes do?
This option goes along with `min_workers`, and `max_workers`. When the
cluster is first brought up (or when it is refreshed with a subsequent
`ray up`) this number of nodes will be started.
It's a workaround for issues of scaling (see related issues) where it
can take a long time (or forever in the case where the head node has
`--num-cpus 0`) to scale up a cluster in response to increasing demand.
## Related issue number
Workaround for https://github.com/ray-project/ray/issues/3339 and https://github.com/ray-project/ray/issues/2106
* Push a warning to all users when large number of workers have been started.
* Add test.
* Fix bug.
* Give warning when worker starts instead of when worker registers.
* Fix
* Fix tests
* Fix warning text in pbt logger
* Allow nested mutations in pbt by recursing explore function
* Add test for nested pbt mutation
* Update pbt explore to only call custom explore on top level
* fix test
* Limit Redis max memory to 10GB/shard by default.
* Update stress tests.
* Reorganize
* Update
* Add minimum cap size for object store and redis.
* Small test update.
This change ensures that Ray set up fault handlers only if it has not been enabled by other applications. Otherwise some applications could face strange issues when using Ray, and some unittests using xml runners will fail.
