ericl
Follow
Pinned
2,750 contributions in the last year
Contribution activity
December 2020
Created 20 commits in 2 repositories
Created a pull request in ray-project/ray that received 16 comments
Increase the number of unique bits for actors to avoid handle collisions
Why are these changes needed? Currently, actor ids are randomly generated, and there are only 32 bits of randomness available. If an actor id coll…
+54
−54
•
16
comments
Opened 22 other pull requests in 1 repository
ray-project/ray
8
closed
1
open
13
merged
- [WIP] ns test
- [WIP] ns test
- [WIP] Revert "Revert "Increase the number of unique bits for actors to avoi…
- [build] [hotfix] Revert "Increase the number of unique bits for actors to avoid handle…
- Revert "[Object Spilling] Implement level triggered logic to make str…
- Clean up block/unblock handling of resources in new scheduler
- Implement resource deadlock detection for new scheduler
- [WIP] Don't release resources on network IO stages of ray.get/wait
- [New scheduler] Also unsubscribe get dependencies on unblock
- Clip RLIMIT_NOFILE increase to avoid redis failing to start on Big Sur
- [flaky test hotfix] Revert "[PlacementGroup]Add PlacementGroup wait python api (#12601)"
- Add grace period before warning of resource deadlock
- Add automatic local GC and plasma debug logs every 10 minutes by default
- Enable by default new scheduler
- Add comments to clarify purpose of new scheduler queues
- Remove incorrect warning about process events being slow
- Revert "[RLlib] Batch-size for truncate_episode batch_mode should be …
- Disable ray debugger tests on windows
- Improve issue templates
- Revert "[autoscaler/k8s] Use ray node's HOME in Kubernetes command ru…
- [WIP] beam runner experiment
- Disable flaky test test_delete_objects_multi_node
Reviewed 54 pull requests in 1 repository
ray-project/ray 54 pull requests
- [ray_client] Support ray.get_actor()
- [ray_client] Implement optional arguments to ray.remote() and f.options()
- [ray_client] Integrate with test_basic, test_basic_2 and test_actor
- Enable by default new scheduler
- [Java] Fix output parsing in RunManager
- [autoscaler] Do not count head node with min_workers constraint.
- [Object Spilling] Implement level triggered logic to make streaming shuffle work + additional cleanup
- [Autoscaler] New output log format
- Clean up block/unblock handling of resources in new scheduler
- [ray_client]: Implement object retain/release and Data Streaming API
- Implement resource deadlock detection for new scheduler
- [New Scheduler/Old Scheduler] Implement WarnResourceDeadlock
- [RLlib] Attention Nets: tf
- [autoscaler] Fix flaky autoscaler test
- [rllib] Disable test_distributions due to failing in CI
- Fix pull manager retry
- [New scheduler] Object spilling
- Revert "[RLLib] Execution-Folder Type Annotations (#12760)"
- [Core] Fix for RLIMIT patch for Big Sur
- [Doc] Fix placement group doc
- Check for raylet PID as ppid in dashboard agent fate-sharing
- Revert heartbeat interval to keep ci stable
- [autoscaler] Fix flaky autoscaler test
- [tune] upgrade gpytorch, bump default pytorch to 1.7.0
- Hold GIL while submitting (actor) tasks
- Some pull request reviews not shown.
Created an issue in ray-project/ray that received 12 comments
[core] Support detached/GCS owned objects
Describe your feature request In the current ownership model, all created objects are owned by some worker executing the job. This means however t…
12
comments