Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 28 million developers.
Sign up
Pinned repositories
2,139 contributions in the last year
Contribution activity
November 2018
Created a pull request in ray-project/ray that received 27 comments
- Revert "[autoscaler] Also grant roles to worker nodes"
- Add warning about using latest wheels
- [rllib] Fix rllib rollouts script and add test
- [rllib] Decouple ape-x sampling and learning speed
- Add `ray stack` tool for debugging
- Set the process title in workers and actors
- [minor] Change chunk already exists to DEBUG, add flags for rllib multi node testing
- [minor] change dict serialization warning to debug
- Remove suppressing duplicate error message (missed a couple)
- Cap object store memory to 20GB when size is None
- [rllib] Add some debug logs during agent setup
- Better errors on process deaths
- [rllib] rollout.py should reduce num workers
- [autoscaler] Add option to allow private ips only
- [autoscaler] missing example-full.yaml file in the latest wheel for provider type "local".
- [rllib] Add test for multi-agent support and fix IMPALA multi-agent
- [rllib] Don't reset envs when possible
- [rllib] Add self-supervised loss to model
- [rllib] Clean up agent resource configurations
- Kill actor child processes on shutdown
- [tune] Track and warn on low memory
- [rllib] Raise worker TF intra_op threads to 2, lower driver intra_op threads to 8
- [rllib] Rename ServingEnv => ExternalEnv
- Add debug string to raylet
- Raise exception if the node is nearly out of memory
- Some pull requests not shown.
- Fix some paths /tmp/raylogs -> /tmp/ray.
- Increase timeout before reconstruction is triggered
- [xray] Implement actor reconstruction.
- [tune] Raise Error when overstepping
- Allow multiple threads to call ray.get and ray.wait
- Set the process title in workers and actors
- Expose internal config parameters for starting Ray
- Refactor ObjectDirectory to reduce and fix callback usage
- Suppress duplicate pre-emptive object pushes.
- [tune] Annotated Example Page and showcase Tutorials
- [tune] Support "None" for upload_dir
- [rllib] Add test for multi-agent support and fix IMPALA multi-agent
- [tune] Fix default handling for timesteps
- [rllib] Raise worker TF intra_op threads to 2, lower driver intra_op threads to 8
- Kill actor child processes on shutdown
- Don't transfer objects whose data is not needed
- [tune] Doc: Autofilled, StatusReporter
- Clean up Ray processes after cluster util exits
- Release sender connections as soon as WriteMessageAsync completes
- [autoscaler] Submit command
- [tune] Node Fault Tolerance
- KL Divergence Metrics
- support home path and relative path for temp-dir
- [WIP] Allow object managers to cancel unnecessary object pull requests.
- Don't setpgid() on actors
- Some pull request reviews not shown.
Created an issue in ray-project/ray that received 1 comment
[rllib] Perf regression in IMPALA in master
Describe the problem In master, IMPALA is noticeably slower in multi-node. This seems to be because broadcast is broken in the new backend (transfe…
1
comment
8
contributions
in private repositories
Nov 6 – Nov 15