Overview

1 Release published by1 person

• ray-2.47.1 Ray-2.47.1

  publishedJun 18, 2025

422 Pull requests merged by84 people

• [DOC][Core] fix typo in Anti-pattern.

  #54547 mergedJul 12, 2025

• use wait_condition for verifying http response

  #54522 mergedJul 12, 2025

• [serve] reorganize how we handle the http receive task

  #54543 mergedJul 12, 2025

• [deps] upgrade python protobuf to 4

  #54496 mergedJul 12, 2025

• cherrypick #54518

  #54561 mergedJul 12, 2025

• cherrypick #54386

  #54560 mergedJul 12, 2025

• cherrypick #54511

  #54559 mergedJul 12, 2025

• cherrypick #54544

  #54558 mergedJul 12, 2025

• [release] veresion change to 2.48.0

  #54557 mergedJul 12, 2025

• [Core] Add file_mounts to azure example-minimal config

  #54533 mergedJul 12, 2025

• [core][telemetry/10] support custom gauge+counter+sum metrics

  #53734 mergedJul 12, 2025

• [Serve.llm] Make llm serve endpoints compatible with vLLM serve frontend (5/N): Change source of dashboards from ray serve llm metrics to vllm metrics

  #54531 mergedJul 12, 2025

• [Data] Add Expression Support & with_columns API

  #54322 mergedJul 12, 2025

• [core][autoscaler] add the missing readonly/example.yaml to the build

  #54535 mergedJul 12, 2025

• document unexpected queuing behavior in handle

  #54542 mergedJul 12, 2025

• [train] TrainStateActor periodically checks controller status and sets aborted

  #53818 mergedJul 12, 2025

• Revert "[core] Default state API address when in a connected worker"

  #54549 mergedJul 12, 2025

• [Serve.llm] Make llm serve endpoints compatible with vLLM serve frontend (4/N): Refactor LLMServer

  #54484 mergedJul 12, 2025

• [ci] kick forge refresh

  #54544 mergedJul 11, 2025

• only print log line once during shutdown

  #54534 mergedJul 11, 2025

• concat: Handle mixed Tensor types for structs

  #54386 mergedJul 11, 2025

• [core][gpu-objects] garbage collection

  #53911 mergedJul 11, 2025

• [wheel] limit build artifacts duplicated in example directory for ray_cpp

  #54465 mergedJul 11, 2025

• [core] Default state API address when in a connected worker

  #54468 mergedJul 11, 2025

• [core] enable the v2 autoscaler by default when the cluster is managed by KubeRay

  #54518 mergedJul 11, 2025

• [Serve] Set the docs path after app is initialized on the replica

  #53463 mergedJul 11, 2025

• [Doc][Cluster] Update Azure cluster docs

  #54517 mergedJul 11, 2025

• [ci] bumping uv binary version

  #54514 mergedJul 11, 2025

• [serve] Fixtest_deploy on windows

  #54511 mergedJul 11, 2025

• [core][telemetry/09] record sum metric e2e

  #53512 mergedJul 11, 2025

• [core][telemetry/08-bis] api documentation + improvements

  #54472 mergedJul 11, 2025

• [release] remove dask from byod 3.9 deps

  #54521 mergedJul 11, 2025

• [serve] updatetest_request_timeout

  #54519 mergedJul 11, 2025

• [Serve.llm] Make llm serve endpoints compatible with vLLM serve frontend (3/N): Remove indirection layers of node initialization

  #54481 mergedJul 10, 2025

• [uv] Fix uv run parser for handling extra arguments

  #54488 mergedJul 10, 2025

• [core][autoscaler] fix: enable cloud_instance_id reusing in autoscaler v2

  #54397 mergedJul 10, 2025

• [core] Don't order retries for in-order actors to prevent deadlock

  #54034 mergedJul 10, 2025

• [serve] deflake test_e2e_preserve_prev_replicas

  #54513 mergedJul 10, 2025

• [Serve] Update timeout to 20 for test_deploy_bad_pip_package_deployment

  #54510 mergedJul 10, 2025

• increase timeout for wait condition

  #54503 mergedJul 10, 2025

• [serve] deflake test_replica_metrics_fields

  #54493 mergedJul 10, 2025

• Feat/add websocket support for di

  #54490 mergedJul 10, 2025

• [Serve.llm] Make llm serve endpoints compatible with vLLM serve frontend (2/N): Incremental refactor of LLMEngine

  #54443 mergedJul 10, 2025

• [serve] fixtest_standalone_2

  #54508 mergedJul 10, 2025

• Optimize get_live_deployments

  #54454 mergedJul 10, 2025

• Feat/fix callback tests

  #54507 mergedJul 10, 2025

• [Docs] Troubleshooting DeepSeek/multi-node GPU deployment on KubeRay

  #54229 mergedJul 10, 2025

• split _wrap_user_method_call into _wrap_request and _start_request

  #54485 mergedJul 10, 2025

• [core] Improve status messages and add comments about stale seq_no handling

  #54470 mergedJul 10, 2025

• [deps] Allow to call individual functions within install-dependencies

  #54502 mergedJul 10, 2025

• Updated stalebot to use unstale label instead of bounced.

  #54506 mergedJul 10, 2025

• [core][telemetry/12] record histogram metric e2e

  #53927 mergedJul 10, 2025

• [deps] core: drop opencensus-proto test dep

  #54497 mergedJul 10, 2025

• [runtime env]: Integrating ROCm Systems Profiler to Ray worker process

  #48525 mergedJul 10, 2025

• [release] update release test dependencies

  #54494 mergedJul 10, 2025

• [ci] stop using get.docker.com

  #54487 mergedJul 10, 2025

• [Serve] Add RouterConfig field to DeploymentConfig to configure RequestRouter

  #53870 mergedJul 10, 2025

• [serve] deflake test_metrics

  #54482 mergedJul 10, 2025

• [serve] deflake test_deployment_scheduler_with_comp_sched

  #54479 mergedJul 10, 2025

• [core] upgrade opentelemetry-sdk

  #53745 mergedJul 10, 2025

• [Serve.llm] Make llm serve endpoints compatible with vLLM serve frontend (1/N): Remove random v0 logic

  #54440 mergedJul 10, 2025

• [core] fix get_max_resources_from_cluster_config

  #54455 mergedJul 9, 2025

• [wheel] only call bazel once when building the wheel

  #54476 mergedJul 9, 2025

• Add __init__.py for prefix tree directories

  #54480 mergedJul 9, 2025

• [core][compiled graphs] Supporting allreduce on list of input nodes

  #51047 mergedJul 9, 2025

• [Data] Fixingmap_groups issues

  #54462 mergedJul 9, 2025

• [Data] Re-enable sorting Ray Data tests

  #54475 mergedJul 9, 2025

• [core][tune] fix RayTaskError (de)serialization logic

  #54396 mergedJul 9, 2025

• [Data] Prevent Op fusion for streaming repartition

  #54469 mergedJul 9, 2025

• [serve] tests for log formatter

  #54248 mergedJul 9, 2025

• [serve] update get_application_url and tests

  #54449 mergedJul 9, 2025

• [serve] deflaketest_autoscaling_policy_with_metr_disab

  #54458 mergedJul 9, 2025

• [Serve] Check multiple FastAPI ingress deployments in a single application

  #53647 mergedJul 9, 2025

• [serve.llm] Adaption of the change of vllm.PoolingOutput

  #54467 mergedJul 9, 2025

• [core][autoscaler] fix: use RAY_UP_enable_autoscaler_v2 instead of RAY_enable_autoscaler_v2 for ray up

  #54456 mergedJul 9, 2025

• migrate check_library_usage_telemetry to _common

  #54355 mergedJul 9, 2025

• [Core] Use std::move in cluster task manager constructor

  #54413 mergedJul 9, 2025

• [core] Delete old skipped tests

  #54427 mergedJul 9, 2025

• Add label selector observability to placement group tables and actor and task detail pages

  #54292 mergedJul 9, 2025

• [llm] bump vllm to 0.9.2

  #54407 mergedJul 9, 2025

• [ci] adding uv binary v0.7.19

  #54437 mergedJul 9, 2025

• [deps] upgrade datasets in release tests

  #54425 mergedJul 9, 2025

• [train][checkpoint] Populate checkpoint from before_init_train_context instead of workergroupcontext

  #54453 mergedJul 9, 2025

• [serve] skip test_proxy_disconnect_metrics on windows

  #54441 mergedJul 8, 2025

• Fix test_json CI failures

  #54352 mergedJul 8, 2025

• [Data] Avoid OOMs withread_json(..., lines=True)

  #54436 mergedJul 8, 2025

• update google tag container id

  #54444 mergedJul 8, 2025

• [train][doc] ray.train.report api docs should mention optional checkpoint_dir_name

  #54391 mergedJul 8, 2025

• [data] Extract backpressure-related code from ResourceManager as a policy

  #54376 mergedJul 8, 2025

• [serve] add request metadata to can_accept_request

  #54429 mergedJul 8, 2025

• [ci] add workspace status script

  #54398 mergedJul 8, 2025

• [deps] sync ml byod deps with global lock file

  #54424 mergedJul 8, 2025

• [serve] bump timeout for test_proxy

  #54426 mergedJul 8, 2025

• [wheel] remove pyarrow <18 restriction

  #54405 mergedJul 8, 2025

• [Autoscaler][V2] Use running node instances to rate-limit upscaling

  #50414 mergedJul 8, 2025

• [Refactor]Rename NCCL-related items to comm_backend

  #51061 mergedJul 8, 2025

• [core] Skiptest_owner_assign_inner_object on Windows

  #54383 mergedJul 8, 2025

• [Data] Addpin_memory toiter_torch_batches

  #53792 mergedJul 8, 2025

• [serve] Remove usage ofinternal_api.memory_summary()

  #54417 mergedJul 8, 2025

• [serve] Remove usage ofray._private.state

  #54140 mergedJul 8, 2025

• [core] Inject reorder_wait_seconds for scheduling queue test

  #54404 mergedJul 8, 2025

• [Core] Use smart pointer in logging.cc

  #54351 mergedJul 8, 2025

• [core] Delete skip_flaky_core_test_premerge

  #54382 mergedJul 8, 2025

• migrate signature from _private to _common

  #54357 mergedJul 8, 2025

• [ci] release tests: remove unused app configs

  #54410 mergedJul 8, 2025

• [data] dask: disable smoke test

  #54411 mergedJul 8, 2025

• [Data] Fixingsort_benchmark to avoid offsets overflows with Pyarrow

  #54390 mergedJul 8, 2025

• [ci] upgrade rayci version to 0.16.0

  #54392 mergedJul 8, 2025

• updating raydepsets test job

  #54387 mergedJul 8, 2025

• [release] Make KubeRay test run nightly

  #54243 mergedJul 8, 2025

• [core][GPU Objects] Disable tensordict tests in macos ci

  #54375 mergedJul 8, 2025

• [data.llm] Return a batch of rows in the udf instead of row by row

  #54329 mergedJul 7, 2025

• [Data] Revisit async UDF handling in Ray Data

  #54190 mergedJul 7, 2025

• [core] Unskiptest_placement_group_strict_spread

  #54381 mergedJul 7, 2025

• [core] Add deprecation warning and remove tests for_max_cpu_fraction_per_node

  #54380 mergedJul 7, 2025

• [core] Delete event_label

  #54378 mergedJul 7, 2025

• [core][autoscaler] make the autoscaler v2 work with the cluster launcher

  #54230 mergedJul 7, 2025

• [core] Removetest_get_locations_timeout

  #54367 mergedJul 7, 2025

• [data] dask: mark all dask-on-ray tests as manual

  #54371 mergedJul 7, 2025

• [train] Force abort on SIGINT spam and do not abort finished runs

  #54188 mergedJul 7, 2025

• [core] Deflaketest_reconstruction_suppression

  #54366 mergedJul 7, 2025

• [core] Deflaketest_actor_scheduling_not_block_with_placement_group

  #54368 mergedJul 7, 2025

• [train] Document ray.train.collective

  #54340 mergedJul 7, 2025

• [data.llm] Decouple max_tasks_in_flight from max_concurrent_batches

  #54362 mergedJul 7, 2025

• [ci] sync versions of pytest and pip-tools

  #54315 mergedJul 7, 2025

• reduce number of loops over request headers from 2 to 1

  #54326 mergedJul 7, 2025

• [data.llm] Log engine stats after each batch task is done.

  #54360 mergedJul 7, 2025

• [serve] deflaketest_autoscaling_policy

  #54336 mergedJul 7, 2025

• [core] Get cloud provider with ray on kubernetes

  #51793 mergedJul 7, 2025

• [deps] remove time series and ludwig dependencies

  #54316 mergedJul 7, 2025

• [Data] - write_parquet enable both partition by & min_rows_per_file, max_rows_per_file

  #53930 mergedJul 7, 2025

• [core][GPU Objects] Add related tests for tensordict

  #54286 mergedJul 7, 2025

• [core][refactor] move NodeManager::KillWorker to WorkerInterface::KillAsync for better testability

  #54068 mergedJul 7, 2025

• [core] Don't try to monitor zipped files

  #53151 mergedJul 7, 2025

• [cpp] add explicit files for deps

  #54311 mergedJul 6, 2025

• [wheel] mac: upgrade arm64 wheel to macos 12

  #54323 mergedJul 5, 2025

• raydepsets scaffolding (package management tool)

  #54265 mergedJul 5, 2025

• [Train] Remove the subclass relationship between RunConfig and RunConfigV1

  #54293 mergedJul 5, 2025

• [serve] configure http options in controller

  #54331 mergedJul 4, 2025

• [Data] Fix flakytest_shuffle

  #54339 mergedJul 4, 2025

• [train] Add broadcast_from_rank_zero and barrier collectives

  #54066 mergedJul 4, 2025

• [serve] deflake test_multiplex

  #54335 mergedJul 4, 2025

• Updated stalebot to add bounced label and exempt labels

  #54318 mergedJul 3, 2025

• [core][task-manager/02] consolidate TaskManager interface

  #54317 mergedJul 3, 2025

• Fixtest_runtime_env_container

  #54330 mergedJul 3, 2025

• [core] Fix gcs register actor callback check

  #53634 mergedJul 3, 2025

• [doc][kuberay] staterayStartParams is optional starting with KubeRay 1.4.0

  #53943 mergedJul 3, 2025

• [core] Skip generator reconstruction test

  #54320 mergedJul 3, 2025

• [core] UseSignalActor intest_hybrid_policy_threshold

  #54312 mergedJul 3, 2025

• [data] adapt dask on ray to the new dask task class

  #54108 mergedJul 3, 2025

• disable uvicorn message logger middleware

  #54309 mergedJul 3, 2025

• [serve] speed up CI tests

  #54303 mergedJul 3, 2025

• [serve] speed uptest_deploy_app

  #54304 mergedJul 3, 2025

• [ci] remove min install on aarch64 mac

  #54231 mergedJul 3, 2025

• [setup] remove invoker inbazel_invoke

  #54302 mergedJul 3, 2025

• [serve] rebalance serve CI tests

  #54296 mergedJul 3, 2025

• [train][tune] add support for dynamically loading callbacks by environement variables

  #54233 mergedJul 3, 2025

• [core] add digest for opentelemetry proto

  #54300 mergedJul 3, 2025

• modify example names to be modality based

  #54297 mergedJul 3, 2025

• [serve] Add arguments

  #54295 mergedJul 3, 2025

• [data.llm][Bugfix] Fix doc to only support intconcurrency

  #54196 mergedJul 2, 2025

• fix toctree for object detection README

  #54290 mergedJul 2, 2025

• add memory buffer logger to serve

  #54269 mergedJul 2, 2025

• [core][refactor/01] remove MarkTaskCanceled as a condition check

  #54283 mergedJul 2, 2025

• [core] Deflaketest_hybrid_policy_threshold

  #54271 mergedJul 2, 2025

• [doc][core] fix reStructuredText formatting on Resources page

  #53882 mergedJul 2, 2025

• [Data] Refactortest_json_read_partitioned_with_filter to avoid actors

  #54279 mergedJul 2, 2025

• [docs; RLlib] Remove "new API stack" banner from all RLlib docs pages as its now the default.

  #54282 mergedJul 2, 2025

• [Data] Fix IcebergDatasink to properly generate individual file uuids

  #52956 mergedJul 2, 2025

• [docs] move directives to bottom of README.ipynb

  #54250 mergedJul 2, 2025

• add a env var to disable forceful replica shutdown

  #54204 mergedJul 2, 2025

• feat(runtime_env): add Azure Blob Storage support

  #53135 mergedJul 2, 2025

• [Data] Update release test datasets to us-west-2 buckets

  #54258 mergedJul 2, 2025

• [serve] Movepickle.dumps to handle request methods

  #54259 mergedJul 2, 2025

• [deps] upgrade pytest-virtualenv

  #54214 mergedJul 2, 2025

• [core] Fix "Check failed: it->second.num_retries_left == -1"

  #54116 mergedJul 2, 2025

• Update azure.md - Missing azure dependency

  #49104 mergedJul 2, 2025

• [core] Fix sanitizers for actor manager test

  #54224 mergedJul 1, 2025

• [core] Add static type hints for Actor methods

  #54173 mergedJul 1, 2025

• [core] Removetest_schedule_many_actors_and_normal_tasks

  #54249 mergedJul 1, 2025

• [core] Add timeouts totest_scheduling.py::test_hybrid_policy

  #54176 mergedJul 1, 2025

• [RLlib; Offline RL] Implement Offline Policy Evaluation (OPE) via Importance Sampling.

  #53702 mergedJul 1, 2025

• [serve] optimize code when user code is on same event loop

  #54227 mergedJul 1, 2025

• [core][GPU objects] Attach tensor transport to task args protobuf

  #53935 mergedJul 1, 2025

• Cursor/update ray docs Twitter link to X

  #54238 mergedJul 1, 2025

• [RLlib] Enhance ConnectorV2ObservationPreprocessor APIs (add multi-agent support; addepisode arg).

  #54209 mergedJul 1, 2025

• Fix broken Ray Workflows documentation link in README.rst

  #53136 mergedJul 1, 2025

• [core][test] deflaky test_demand_report_when_scale_up by reducing workloads

  #54183 mergedJul 1, 2025

• [core] Deflaketest_node_affinity_scheduling_strategy_soft_spill_on_unavailable

  #54247 mergedJul 1, 2025

• remove flaky marker from test

  #44033 mergedJul 1, 2025

• [doc][kuberay] add version skew warning for plugin and RayCluster

  #53950 mergedJul 1, 2025

• [RLlib] Bug fix: Failed EnvRunners are not restored if there is no local EnvRunner.

  #54091 mergedJul 1, 2025

• [core] Delete asyncio actor logic in in-order scheduling code

  #54033 mergedJul 1, 2025

• move Collector class to _common

  #54180 mergedJul 1, 2025

• Revert "Revert "remove extraneous index.rst file for e2e examples (part 2)""

  #54234 mergedJul 1, 2025

• [docker] Update latest Docker dependencies for 2.47.1 release

  #54015 mergedJul 1, 2025

• [Doc][KubeRay] remove head pod trailing hash and adjust volcano output

  #53826 mergedJul 1, 2025

• [serve] support running user code on same event loop

  #54219 mergedJun 30, 2025

• [data] split dask doc tests out into its own jobs

  #54207 mergedJun 30, 2025

• [core] Recover intermediate objects if needed while generator running

  #53999 mergedJun 30, 2025

• [CI][KubeRay] Update KubeRay CI Tests branch for KubeRay v1.4.0 release

  #53984 mergedJun 30, 2025

• [serve] extract proxy args in replica

  #54216 mergedJun 30, 2025

• [data] run dask tests seperately

  #54195 mergedJun 30, 2025

• [serve] take scope, receive, send for call http entrypoint

  #54184 mergedJun 30, 2025

• Feat/update backpressure tests

  #54212 mergedJun 30, 2025

• [Core] Fix test_aggregator_agent Failing Test on MacOS

  #54217 mergedJun 30, 2025

• [ci] unify macos build script across platforms

  #54198 mergedJun 30, 2025

• [core] Removeray.wait fortest_locality_aware_leasing_borrowed_objects

  #54211 mergedJun 30, 2025

• [core] Removetest_scheduling_performance.py

  #54210 mergedJun 30, 2025

• [data] Add timeout fortest_arrow_block_scaling.py

  #54155 mergedJun 30, 2025

• [core] Fix race condition b/w object eviction & repinning for recovery.

  #53934 mergedJun 30, 2025

• [core] Manually run io service in actor manager tests to simulate different orderings

  #54203 mergedJun 30, 2025

• [core] Atomic num_tasks_submitted

  #54200 mergedJun 30, 2025

• [Doc][KubeRay] Kuberay gcs ft takes yaml file with version 1.4.0

  #54192 mergedJun 30, 2025

• [core] Task receiver cleanup

  #54205 mergedJun 30, 2025

• BLD: Automatically patch.bazelrc file for Windows 11 build

  #53586 mergedJun 30, 2025

• [Docs][KubeRay] Update all KubeRay version references for KubeRay 1.4.0 release

  #53884 mergedJun 30, 2025

• [serve] Move logic into user callable wrapper

  #54177 mergedJun 29, 2025

• [Doc] Convert configuring-autoscaling.ipynb back to markdown docs

  #54111 mergedJun 28, 2025

• [Docs][KubeRay] Convert rayservice-quick-start.ipynb back to markdown docs

  #54138 mergedJun 28, 2025

• [Doc][KubeRay] Convert raycluster-quick-start.ipynb back to markdown docs

  #54125 mergedJun 28, 2025

• [Doc][KubeRay] Add doc for running KubeRay dashboard

  #53830 mergedJun 28, 2025

• [ci] removeci/keep_alive

  #54079 mergedJun 28, 2025

• [data] gather dask tests into single test files

  #54163 mergedJun 28, 2025

• [Data] Add TooManyRequests catch to BQ writer

  #54000 mergedJun 28, 2025

• [Data] Fixtest_binary setup fixture that doesn't close file handles

  #54028 mergedJun 28, 2025

• [serve] Increase default uvicorn keep alive timeout

  #54127 mergedJun 27, 2025

• [doc] fix broken links in the vllm guide

  #54161 mergedJun 27, 2025

• [Docs][KubeRay] Delete KubeRay doctests

  #54080 mergedJun 27, 2025

• [Feat][Core] Implement Event Aggregator Agent

  #53182 mergedJun 27, 2025

• Feat/middleware callback support

  #54106 mergedJun 27, 2025

• [data] Handle HuggingFace parquet dataset resolve URLs

  #54146 mergedJun 27, 2025

• [data] Usewrite_dataset for partitioning & writing to file instead of custom implementation

  #54052 mergedJun 27, 2025

• Correct asyncio ref documentation for Python 3.11+

  #54157 mergedJun 27, 2025

• [core][test] fix flaky data races in NodeManagerTest

  #54129 mergedJun 27, 2025

• [Doc][KubeRay] verl example

  #54114 mergedJun 27, 2025

• [RLlib] Fix shapes inexplained_variance for recurrent policies.

  #54005 mergedJun 27, 2025

• [ci] add cibase tags for ci base envs

  #53755 mergedJun 27, 2025

• Removebotocore dependency in Ray Serve LLM

  #54156 mergedJun 27, 2025

• (serve.llm) Remove test leakage from placement bundle logic

  #53723 mergedJun 27, 2025

• [data] split dask and modin tests

  #54122 mergedJun 26, 2025

• [Data] Fixing PyArrow overflow handling

  #53971 mergedJun 26, 2025

• [serve] split call_user_method

  #54104 mergedJun 26, 2025

• [Data] Handle Huggingface Integration CI test failures

  #54128 mergedJun 26, 2025

• [Data] Fix ActorPool autoscaler to properly scale up

  #53983 mergedJun 26, 2025

• use gtm datalayer directly, fix format

  #54144 mergedJun 26, 2025

• [package] remove__api__ insetup.py

  #54143 mergedJun 26, 2025

• [Minor][Fix][Core/Test] Fix test_actor_restart_on_node_failure wrong test logic without waiting

  #54088 mergedJun 26, 2025

• [data] fix repartitioning empty datasets

  #54107 mergedJun 26, 2025

• [Doc][KubeRay] revert kuberay-gcs-ft.ipynb to markdown

  #54084 mergedJun 26, 2025

• Fix sort_benchmark release test arg

  #54145 mergedJun 26, 2025

• [Doc][KubeRay] Convert rayjob-quick-start.ipynb back to markdown docs

  #54093 mergedJun 26, 2025

• [core] split dask and modin tests

  #54121 mergedJun 26, 2025

• [Core] Remove Unnecessary Checks in GRPC Server Shutdown Process

  #53910 mergedJun 26, 2025

• [core] Delete unused env vars

  #54095 mergedJun 26, 2025

• [Doc][KubeRay] Removerayserve-dev-doc.md

  #54057 mergedJun 26, 2025

• [core] Bump timeout intest_ray_init

  #54136 mergedJun 26, 2025

• [core] Clean up unused FFs

  #54139 mergedJun 26, 2025

• [core] Fix GCS crash on duplicate MarkJobFinished RPCs due to network failures

  #53951 mergedJun 26, 2025

• [train] Remove usage ofray._private.state

  #54142 mergedJun 26, 2025

• [core] Deflaketest_scheduling.py in client mode

  #54137 mergedJun 26, 2025

• [core] Fixtest_basic_3.py in client mode

  #54135 mergedJun 26, 2025

• [serve] refactor _run_user_code

  #54103 mergedJun 26, 2025

• [Doc] vale ignores anchors of headers

  #53580 mergedJun 26, 2025

• set config for ua tag

  #54112 mergedJun 26, 2025

• [Serve.llm] Add a doc snippet to inform users about existing diffs between vllm serve and ray serve llm.

  #54042 mergedJun 26, 2025

• [ci][docs] Add test tag rule for Vale files

  #54118 mergedJun 26, 2025

• [train] update beginner pytorch example

  #54124 mergedJun 26, 2025

• [Data] Bumped latest PA version to 20.0

  #54123 mergedJun 26, 2025

• [ci] fix missingdask tag in all tags list

  #54113 mergedJun 26, 2025

• [core][test] fix data races in NodeManagerTest

  #54097 mergedJun 25, 2025

• [core] Remove experimental "array" library

  #54105 mergedJun 25, 2025

• [core] Clean uptest_locality_aware_leasing_borrowed_objects

  #54086 mergedJun 25, 2025

• [core][refactor] replace unnecessary shared_ptrs with unique_ptrs and references in raylet

  #54062 mergedJun 25, 2025

• [ci] fix mac ci by pinning cython version

  #54061 mergedJun 25, 2025

• [core] Deflaketest_basic_3.py

  #54083 mergedJun 25, 2025

• remove final references to plasma_event_handler

  #54085 mergedJun 25, 2025

• [core] Deflaketest_ray_init

  #54094 mergedJun 25, 2025

• [core] Deflaketest_actor_restart

  #54087 mergedJun 25, 2025

• Updated stalebot to run every 12 hours.

  #54041 mergedJun 25, 2025

• [serve] Prefer localhost instead of host ip for microbenchmarks

  #54092 mergedJun 25, 2025

• [train] Driver SIGINT calls controller abort

  #53600 mergedJun 25, 2025

• [data] Split out long running scaling test

  #54045 mergedJun 25, 2025

• [core] Deflaketest_actor_unavailable_conn_broken

  #54090 mergedJun 25, 2025

• [V2][Autoscaler] FixnumOfHosts > 1 slice termination logic

  #54063 mergedJun 25, 2025

• [V2][Autoscaler] Addcloud_instance_id to all V2 Austoscaler termination requests

  #53938 mergedJun 25, 2025

• Fix autoscaler recovery docker config to use node-specific settings

  #53992 mergedJun 25, 2025

• [data/preprocessors] Improve execution perf for One Hot encoding

  #54022 mergedJun 25, 2025

• [Docs][KubeRay] Update changes from KubeRay 1.3.2 to 1.4.0

  #53886 mergedJun 25, 2025

• [core] Fix comment

  #53853 mergedJun 25, 2025

• [RLlib] Fix env seeding in EnvRunners and rewritedeterministic_training example script (was still on old stack).

  #54039 mergedJun 25, 2025

• [ci] add-sSL for curl on node install

  #54060 mergedJun 25, 2025

• updating compile comment

  #54058 mergedJun 25, 2025

• Revert "remove extraneous index.rst file for e2e examples (part 2)"

  #54051 mergedJun 25, 2025

• [data] fix lint error in conftest.py

  #54053 mergedJun 25, 2025

• [serve] Useget_application_url in test_metrics

  #54050 mergedJun 24, 2025

• [ci] update anyscale layer

  #54043 mergedJun 24, 2025

• [serve.llm] Prefix aware router eviction thread improvements

  #53957 mergedJun 24, 2025

• [serve] Remove hardcoded urls from serve microbenchmarks

  #54026 mergedJun 24, 2025

• [core] fix detached actor being unexpectedly killed

  #53562 mergedJun 24, 2025

• [POC] fix test_metrics

  #54037 mergedJun 24, 2025

• [serve] Handle request with Semaphore

  #54019 mergedJun 24, 2025

• remove extraneous index.rst file for e2e examples (part 2)

  #54023 mergedJun 24, 2025

• [☀️] Fix repr for ray.ObjectRef, ray.ObjectRefGenerator types

  #54011 mergedJun 24, 2025

• [core][ci] Disable test db for container tests

  #54031 mergedJun 24, 2025

• [docker] Update latest Docker dependencies for 2.47.1 release

  #54016 mergedJun 23, 2025

• [core] improve assertion check in test_task_metrics

  #53958 mergedJun 23, 2025

• remove extraneous index.rst file for e2e-multimodal-ai-workloads

  #54017 mergedJun 23, 2025

• [Serve.llm] Remove ImageRetriever class and related tests from the LLM deployment module.

  #53980 mergedJun 23, 2025

• fix test_request_timeout timeout mismatch issue

  #54010 mergedJun 23, 2025

• fix gsat global

  #54012 mergedJun 23, 2025

• [train] Fix release test missing data key

  #53963 mergedJun 23, 2025

• [data] remove schema from release tests

  #53956 mergedJun 23, 2025

• [kuberay] log actionable err msg when required TPU node selectors missing

  #53914 mergedJun 23, 2025

• [core] Fix flakytest_state_api

  #53975 mergedJun 23, 2025

• [data] remove operator_fusion_benchmark

  #53962 mergedJun 23, 2025

• [Data] Add reading from Delta Lake tables and from Unity Catalog

  #53701 mergedJun 23, 2025

• test: refactortest_observability_helpers

  #53875 mergedJun 23, 2025

• [core] Remove actor task path in normal task submitter

  #53996 mergedJun 23, 2025

• [core] RenameGcsFunctionManager and use fake in test

  #53973 mergedJun 23, 2025

• [Serve.llm][P/D] Fix health check in prefill disagg

  #53937 mergedJun 22, 2025

• [Test][KubeRay] Update KubeRay version to v1.4.0 for autoscaler tests

  #53974 mergedJun 22, 2025

• [core] Fix ActorClass.remote return typing and expose Actor class methods to static analysis

  #53986 mergedJun 21, 2025

• [core] Use core worker client pool in GCS

  #53654 mergedJun 21, 2025

• [core] Revert container tests to medium size instance

  #53966 mergedJun 21, 2025

• Fix ray import error when both ROCR_VISIBLE_DEVICES and HIP_VISIBLE_DEVICES are set

  #53757 mergedJun 20, 2025

• [core] Making NodeManager use ILocalTaskManager instead of TaskManager.

  #53961 mergedJun 20, 2025

• defer loading csat so gtag loads first

  #53968 mergedJun 20, 2025

• fix ga4 events

  #53967 mergedJun 20, 2025

• [train][template] Remove clock emoji which does not always render well

  #53965 mergedJun 20, 2025

• [core][gpu-objects] Supportray.get on the driver process for GPU objects

  #53902 mergedJun 19, 2025

• [kuberay] Update helm install command in prometheus doc to set serviceMonitorrelease=prometheus

  #53952 mergedJun 19, 2025

• [Docs] Fix async code in serving notebook

  #53864 mergedJun 19, 2025

• [core][rocm] Allow CUDA_VISIBLE_DEVICS and HIP_VISIBLE_DEVICES

  #53531 mergedJun 19, 2025

• [train][template] Pip install with python block instead

  #53928 mergedJun 19, 2025

• [Data] RefactorPlanner to avoid storing plan-specific state

  #53955 mergedJun 19, 2025

• [core] Avoid unnecessary deserialization/serialization of CallerWorkerId

  #53939 mergedJun 19, 2025

• [serve] add ability to track child requests

  #53941 mergedJun 19, 2025

• [Doc][KubeRay] Add a doc for scheduler plugins

  #53846 mergedJun 19, 2025

• [core][telemetry/08] record counter metric e2e

  #53449 mergedJun 19, 2025

• [HashShuffle] - Add warnings for when there are insufficient resources for Aggregators

  #53705 mergedJun 19, 2025

• [Data] Join release tests

  #53903 mergedJun 19, 2025

• [docs][Serve] Add clarification for health check and FT of serve deployments

  #53944 mergedJun 19, 2025

• fastapi and streaming tests use get applications api

  #53949 mergedJun 19, 2025

• [RLlib; docs] Fix docstring example for custom MultiRLModule with shared encoder.

  #53912 mergedJun 19, 2025

• [Data] Prevent filename collisions on write

  #53890 mergedJun 19, 2025

• [data] fix flakey schema

  #53901 mergedJun 19, 2025

• [Data] FixedBlockMetadata derivation forRead operator

  #53908 mergedJun 19, 2025

• [core] Fix flakytest_worker_exit_intended_user_exit

  #53909 mergedJun 19, 2025

• fix the bash code run error in notebook

  #53900 mergedJun 19, 2025

• [Docs] Fix issues with e2e audio tutorial

  #53932 mergedJun 19, 2025

• [train] Cleanups for training ingest benchmark

  #53684 mergedJun 19, 2025

• [train] add proper filtering to metrics

  #53788 mergedJun 18, 2025

• [cgraph] Avoid depending on torch CPU module for CPU-only actor

  #53849 mergedJun 18, 2025

• [train] expose training input/output in callbacks

  #53869 mergedJun 18, 2025

• Skip test_metrics_agent_with_open_telemetry on mac

  #53917 mergedJun 18, 2025

• [Docs] Add ServiceMonitor section and make some step optional in Grafana & Promethus page

  #53474 mergedJun 18, 2025

• [Docs][KubeRay] Update KubeRay operator installation references for all docs

  #53885 mergedJun 18, 2025

• [Core] Support AMD GPU MI3xx product line

  #51802 mergedJun 18, 2025

• [Doc][KubeRay] Update KubeRay operator installation reference

  #53842 mergedJun 18, 2025

• [Docs][KubeRay] Fix RayJob quickstart doc step 9 error

  #53887 mergedJun 18, 2025

• [Core] Use fd instead of handle for windows log redirection

  #53852 mergedJun 18, 2025

• Add dashboard visualizations for TPU metrics

  #53898 mergedJun 18, 2025

• [ObjectStore] Warn if object store is allocated < 50% of total memory for data workloads

  #53857 mergedJun 18, 2025

• [Data] Deprecate use_polars flag

  #53867 mergedJun 17, 2025

• [data] split test_all_to_all.py

  #53865 mergedJun 17, 2025

• add missing configs for object detection template

  #53895 mergedJun 17, 2025

• [core] Remove hardcoded flaky tests

  #53888 mergedJun 17, 2025

• [Serve][LLM] Simplify _prepare_engine_config()

  #53704 mergedJun 17, 2025

• [core][gpu-objects] Fixtest_gpu_objects_nccl.py

  #53874 mergedJun 17, 2025

• [RLlib] MetricsLogger: Fixget/set_state to handle tensors inself.values.

  #53514 mergedJun 17, 2025

• [Data] Improve handling of mismatched columns

  #53861 mergedJun 17, 2025

• Fix pickle error with remote code models in vLLM Ray worker process

  #53815 mergedJun 17, 2025

• [train][template] Remove ineffective post build script and pip install instead

  #53822 mergedJun 17, 2025

• [core][gpu objects] Integrate single-controller collective APIs with GPU objects

  #53720 mergedJun 16, 2025

• [Data] Improve handling ofpandas.NA

  #53859 mergedJun 16, 2025

• [devx] Fix 'uv run' command line parsing

  #53838 mergedJun 16, 2025

• [Data] Improveread_text trailing newline semantics

  #53860 mergedJun 16, 2025

• [Serve.llm][P/D] Support separate deployment config for PDProxy in Prefill disagg

  #53821 mergedJun 16, 2025

• [Doc][KubeRay] Removevllm-rayservice.md and use Ray Serve LLM instead

  #53844 mergedJun 16, 2025

• add api to get application url

  #53796 mergedJun 16, 2025

• [Doc][KubeRay] Remove very old ResNet benchmark example

  #53839 mergedJun 16, 2025

• [release] Fix release tests

  #53855 mergedJun 16, 2025

• [Serve.llm] Disable TP=2 VLM batch test

  #53825 mergedJun 16, 2025

• [Doc][Fix] reveal the falsely hidden export command in the KubeRay GCS FT guide

  #53832 mergedJun 16, 2025

• [core][gpu-objects] Support intra-process communication

  #53798 mergedJun 16, 2025

• [Doc][KubeRay] Remove very old XGBoostTrainer example

  #53837 mergedJun 16, 2025

• [core] Release resources only after tasks have stopped executing

  #53660 mergedJun 16, 2025

• [core] Deflaketest_multiprocessing.py

  #53802 mergedJun 16, 2025

• [core] Fixtest_object_spilling.py on Windows

  #53851 mergedJun 16, 2025

• [KubeRay] Remove unused YAMLs

  #53840 mergedJun 16, 2025

• [chore] Change file mode ofrayservice-no-ray-serve-replica.md from 755 to 644

  #53843 mergedJun 16, 2025

• fixAggregateFnV2 doc to statefinalize instead of_finalize

  #53835 mergedJun 16, 2025

• [core] Fix GCS subscribers map race condition

  #53781 mergedJun 16, 2025

• [core] deleting unused code from plasma client

  #53814 mergedJun 16, 2025

• [core] Fix race condition in raylet graceful shutdown

  #53762 mergedJun 16, 2025

• [serve] Revert request timeout from serve instance fixtures

  #53809 mergedJun 16, 2025

• [Doc] Remove "Deploying a static Ray cluster without KubeRay"

  #53833 mergedJun 15, 2025

• [Doc] Small mistake in kuberay ingress

  #53834 mergedJun 15, 2025

• [ci] bazelizeget_contributors script

  #53743 mergedJun 14, 2025

• [ci] First release test on GKE

  #53390 mergedJun 14, 2025

• Replacepython setup.py bdist_wheel withpip wheel

  #53458 mergedJun 14, 2025

• [serve] Set route_prefix and docs_path when re-deploying app

  #53753 mergedJun 14, 2025

• [cherry-pick][dashboard] Fix retrieving IP address from the GPUProfilingManager on the dashboard agent

  #53817 mergedJun 14, 2025

• Add tpu usage metrics to reporter_agent

  #53678 mergedJun 14, 2025

• [data] Refactor interface for actor_pool_map_operator

  #53752 mergedJun 13, 2025

• ray-llm container cu124 -> cu128 update

  #53730 mergedJun 13, 2025

• [dashboard] Fix retrieving IP address from theGPUProfilingManager on the dashboard agent

  #53807 mergedJun 13, 2025

• [ci/release] Trigger Ray release by running a Bazel binary

  #52962 mergedJun 13, 2025

• version change for 2.47.1

  #53813 mergedJun 13, 2025

• cherrypick #53671

  #53812 mergedJun 13, 2025

• [core] Move dependencies of NodeManger to main.cc for better testability

  #53782 mergedJun 13, 2025

• [core] Deflaketest_object_spilling.py

  #53803 mergedJun 13, 2025

• [core] Deflaketest_state_api.py

  #53804 mergedJun 13, 2025

• [tune] update BlockMetadata args in tests

  #53791 mergedJun 13, 2025

• [serve] Fix autoscaling metrics

  #53778 mergedJun 13, 2025

• pass route prefix to replica

  #53777 mergedJun 13, 2025

• [Serve] Call shared long poll client router registration in event loop

  #53613 mergedJun 13, 2025

• [core] Add timeout toray.get call intest_update_object_location_batch_failure

  #53805 mergedJun 13, 2025

• [RLlib] Fix device check inLearner.

  #53706 mergedJun 13, 2025

• [core] Deflaketest_client_builder.py

  #53774 mergedJun 13, 2025

• [core] Increase instance sizes for wheel / HA tests

  #53783 mergedJun 13, 2025

139 Pull requests opened by80 people

• [Data] Avoid failing when no `batch_size` is specified while using GPUs as it's perfectly legitimate use-case

  #53810 openedJun 13, 2025

• Bump tqdm from 4.64.1 to 4.66.3 in /python

  #53820 openedJun 13, 2025

• [core] Ungracefully exit if the agent dies unexpectedly

  #53847 openedJun 16, 2025

• [RLlib] Mixin Layer Design Sketch Up

  #53850 openedJun 16, 2025

• [core] adding additional stats to the dump object store usage api.

  #53856 openedJun 16, 2025

• [core] Cleanup naming in core worker scheduling queues

  #53858 openedJun 16, 2025

• [core] Sleep to debug container test

  #53862 openedJun 16, 2025

• Feat/ray serve middleware support

  #53868 openedJun 17, 2025

• [dashboard] Support to overwrite the _client_max_size of http request entity

  #53880 openedJun 17, 2025

• [ci] add python 3.13 ray docker image build

  #53894 openedJun 17, 2025

• Bump gradio from 3.50.2 to 5.31.0 in /python/requirements

  #53899 openedJun 17, 2025

• python depsets tool

  #53904 openedJun 18, 2025

• [core] Move inner_publisher logic into gcsPublisher

  #53905 openedJun 18, 2025

• [RLlib] Add missing colon to CUBLAS_WORKSPACE_CONFIG

  #53913 openedJun 18, 2025

• [RLlib] Add missing documentation for SACConfig's training()

  #53918 openedJun 18, 2025

• Update deletion policy for rayjob quick start

  #53929 openedJun 18, 2025

• [serve] move test from test_grpc to test_proxy

  #53933 openedJun 18, 2025

• Bump urllib3 from 1.26.19 to 2.5.0 in /python

  #53936 openedJun 18, 2025

• [Data] Replaced `get_object_locations` with `get_local_object_locations`

  #53942 openedJun 19, 2025

• finishing commit for issue #52113

  #53964 openedJun 19, 2025

• tune: make Tune status/progress tables readable in dark mode

  #53969 openedJun 20, 2025

• docs(data): fix broken Parameters table

  #53972 openedJun 20, 2025

• [RLlib] Enhance SAC (new API stack) with discrete action support.

  #53982 openedJun 20, 2025

• [Core] Add AcceleratorManager implementation for Rebellions NPU

  #53985 openedJun 21, 2025

• [Doc] Update Istio service mesh graph

  #53988 openedJun 21, 2025

• [Serve] Make replica scheduler backoff configurable #52871

  #53991 openedJun 21, 2025

• Fixes default_dqn_torch_rl_module assuming the device is 'cpu'

  #54004 openedJun 23, 2025

• Added openssl support for PPC64LE.

  #54006 openedJun 23, 2025

• [dashboard] Clean up naming for GPU profiling module

  #54009 openedJun 23, 2025

• [DONOTMERGE] Proof-of-concept for GPU objects + NIXL

  #54024 openedJun 24, 2025

• Bump mlflow from 2.19.0 to 3.1.0 in /doc/source/ray-overview/examples/e2e-xgboost

  #54027 openedJun 24, 2025

• Multimodal ai

  #54029 openedJun 24, 2025

• [core][autoscaler][v1] add heartbeat timeout logic to determine node activity status

  #54030 openedJun 24, 2025

• Bump mlflow from 2.22.0 to 3.1.0 in /python

  #54032 openedJun 24, 2025

• gen test

  #54046 openedJun 24, 2025

• update all 'Run on Anyscale' buttons to redirect to respective template preview pages

  #54049 openedJun 24, 2025

• Add Azure Files support to persistent storage documentation

  #54055 openedJun 24, 2025

• Adapt to vLLM reducing exports from the top level

  #54099 openedJun 25, 2025

• [data] Remove asserts that test internal `ds._block_num_rows()`

  #54109 openedJun 25, 2025

• vLLM ZMQ KVEvent Router

  #54115 openedJun 25, 2025

• [core][cgraph] Export classes related to NCCL communicator

  #54117 openedJun 26, 2025

• [core] fix checking for uv existence during ray_runtime setup

  #54141 openedJun 26, 2025

• [RLlib] Fix checkpoints not having correct metrics

  #54148 openedJun 26, 2025

• [core] Deflake `test_spread_scheduling_overrides_locality_aware_scheduling`

  #54154 openedJun 26, 2025

• [Data] Fix examples in some Data user guides

  #54158 openedJun 27, 2025

• [test] fix test not ending cluster; spelling mistake: tearDow -> tearDown

  #54171 openedJun 27, 2025

• [Data] Fix the issue where data files written by different tasks have the same file name when using Iceberg sink.

  #54172 openedJun 27, 2025

• [Feat][Core] Don't count actor restarts due to node preemption towards max_restarts

  #54175 openedJun 27, 2025

• [Core] Use Factory method to create gcs KV Manager

  #54178 openedJun 27, 2025

• [core][raycheck/01] Fix "it != submissible_tasks_.end()"

  #54179 openedJun 27, 2025

• [Feat][Core] Don't count task retries due to node preemption

  #54182 openedJun 27, 2025

• [RLlib] - Increased default timesteps on two experiments.

  #54185 openedJun 27, 2025

• Token-split prefix router

  #54187 openedJun 27, 2025

• [Serve.llm][Prototype][WIP] Simplify LLMServer and inherit OpenAIServingChat behavior

  #54189 openedJun 28, 2025

• [data] allow custom batcher for dataset iteration

  #54193 openedJun 28, 2025

• [data.llm] Add release test to capture memory leak

  #54194 openedJun 28, 2025

• [core] Normal task submitter cleanup

  #54206 openedJun 30, 2025

• Feat/fix g rpc error code

  #54218 openedJun 30, 2025

• [core] [wip] lazy sub

  #54220 openedJun 30, 2025

• another round of mac debug

  #54232 openedJul 1, 2025

• [core] Introduce `ShutdownCoordinator` and unified core worker shutdown entry points

  #54244 openedJul 1, 2025

• [serve] refactor call_http_entrypoint

  #54253 openedJul 1, 2025

• [DOC-127] MVP for OSS Ray labels

  #54254 openedJul 1, 2025

• [core] Refactoring LocalObjectManager to have a cleaner API for pinning objects.

  #54255 openedJul 1, 2025

• [core][gpu-objects] Move data transfers to a background thread

  #54256 openedJul 1, 2025

• [train] Use FailurePolicy to handle resize failure

  #54257 openedJul 1, 2025

• [core] Always access task finisher from submitters without mutex

  #54262 openedJul 2, 2025

• [Core] Fixed the bug where the child process turned into a zombie process.

  #54266 openedJul 2, 2025

• [Core] Fixed the bug where the head was unable to submit tasks after redis is turned on.

  #54267 openedJul 2, 2025

• fix missing brace and add pytest

  #54270 openedJul 2, 2025

• [RLlib] Switch Offline Data iteration to `iter_torch_batches`.

  #54277 openedJul 2, 2025

• Run Docker builds as non-root user with scoped root access via `sudo`

  #54285 openedJul 2, 2025

• MCP Ray Serve End to End Example

  #54289 openedJul 2, 2025

• [data] Update Hudi integration to support incremental query

  #54301 openedJul 3, 2025

• Enable field documentation with Pydantic

  #54306 openedJul 3, 2025

• [RLlib; docs] Docs do-over (new API stack): `ConnectorV2` documentation (part II).

  #54313 openedJul 3, 2025

• [RLlib] Fix IndexError on peek on empty Stats object without reduction

  #54325 openedJul 3, 2025

• Updating Daft links in Ray documentation

  #54328 openedJul 3, 2025

• [WIP][Core] Deflake test_gcs_fault_tolerance

  #54334 openedJul 3, 2025

• [train] update `datasets` from `2.19.1` to `3.6.0`

  #54338 openedJul 3, 2025

• [Tests] Improve error message on some test-failures because of `IndexError` in `test_utils.py`

  #54343 openedJul 4, 2025

• [core] Correct bytes in flight when objects <5mb

  #54349 openedJul 5, 2025

• [Core] use RunFnPeriodically for metrics report in GCS server

  #54358 openedJul 6, 2025

• [data.llm] Make ray the distributed backend for vLLM stage

  #54361 openedJul 7, 2025

• [java] encapsulation + resource immutability for option classes

  #54370 openedJul 7, 2025

• [core] prevent sending SIGTERM after calling Worker::MarkDead

  #54377 openedJul 7, 2025

• [ci] raydepsets: adding compile operation

  #54389 openedJul 7, 2025

• [ci] raydepsets: adding config dataclass and config loading

  #54394 openedJul 8, 2025

• [train] fail fast if pg cannot be met

  #54402 openedJul 8, 2025

• Fix minor bug in sample code in documentation

  #54403 openedJul 8, 2025

• [RLlib] - Fix `TorchMultiCategorical.to_deterministic` when having different number of categories and logits with time dimension.

  #54414 openedJul 8, 2025

• [release][ci] First test for kuberay release test trigger path

  #54415 openedJul 8, 2025

• [llm.data] Fix AttributeError for the shallow copy of data batch transfer

  #54419 openedJul 8, 2025

• Bump transformers from 4.30.1 to 4.52.1 in /doc/source/templates/testing/docker/03_serving_stable_diffusion

  #54423 openedJul 8, 2025

• [Core] Avoid copy deque in cluster task manager

  #54432 openedJul 8, 2025

• [WIP][Data] Fix Operators to make sure they doesn't produce empty blocks

  #54435 openedJul 8, 2025

• Remove CPU profiler from Ray Service Replica, Proxy & Controller

  #54438 openedJul 8, 2025

• [data] Allocate GPU resources in ResourceManager

  #54445 openedJul 8, 2025

• Add ray.dataset.write_delta for supporting writes to Delta Lake

  #54447 openedJul 8, 2025

• Give the option to make `target_max_block_size` nullable

  #54450 openedJul 8, 2025

• [Data] Limit operator push down

  #54457 openedJul 9, 2025

• WIP: Support nixl

  #54459 openedJul 9, 2025

• [Core] Remove ineffectual TODO comment

  #54464 openedJul 9, 2025

• Use upstream RayPrometheusStatLogger

  #54471 openedJul 9, 2025

• [Core] Minor fixes in GCS health check manager

  #54473 openedJul 9, 2025

• Train Benchmark: Add preserver_order

  #54474 openedJul 9, 2025

• Add optional APIType filter to /api/serve/applications/ endpoint

  #54478 openedJul 9, 2025

• Remove debug logs for when aggregators are ready

  #54483 openedJul 9, 2025

• [Data] Fixed chained inplace assignment to prevent FutureWarning from Pandas

  #54486 openedJul 9, 2025

• [data] Return schema of res in aggregations

  #54489 openedJul 9, 2025

• test basic functionality

  #54491 openedJul 10, 2025

• [train] add LightGBMTrainer user guide

  #54492 openedJul 10, 2025

• [Core] Fix the issue where multiple multithreaded calls to ray.get may cause hanging.

  #54495 openedJul 10, 2025

• [serve.llm] Pass dimensions of embedding request to vllm engine

  #54499 openedJul 10, 2025

• [Data] Add option for enabling out-of-order execution to optimize data processing performance

  #54504 openedJul 10, 2025

• [WIP] `TaskExecutionResult`

  #54505 openedJul 10, 2025

• [serve.llm] Remove upstreamed workarounds

  #54512 openedJul 10, 2025

• Fix bug in http_serve_head by using os.path.realpath instead of inval…

  #54523 openedJul 11, 2025

• Fix get actor timeout multiplier

  #54525 openedJul 11, 2025

• Handle missing 'chunks' key when Databricks UC query returns zero rows

  #54526 openedJul 11, 2025

• [core][raycheck/01] Fix "it != submissible_tasks_.end()"

  #54527 openedJul 11, 2025

• [Doc] Update deprecated `evaluation_strategy` parameter to `eval_strategy` in transformers examples

  #54528 openedJul 11, 2025

• [core] attempting streaming generator hanging fix

  #54529 openedJul 11, 2025

• [Serve] Fix windows test deploy apps flakiness

  #54530 openedJul 11, 2025

• Fix backpressure gRPC error code

  #54537 openedJul 11, 2025

• [Data] [Draft] introduce per-op config options to disable operator fusion

  #54539 openedJul 11, 2025

• [Doc] Make the wording more accurate since we not only have Python workers but also C++, Java workers

  #54546 openedJul 11, 2025

• [dashboard] fix typos

  #54550 openedJul 12, 2025

• [Serve.llm] Make llm serve endpoints compatible with vLLM serve frontend (6/N): Major LLMServer and LLMEngine refactor

  #54554 openedJul 12, 2025

• [train][checkpoint] CheckpointManager and Worker both count checkpoints

  #54555 openedJul 12, 2025

• debug

  #54556 openedJul 12, 2025

• [Core] Minor fixes in gcs job manager

  #54562 openedJul 12, 2025

• [ci] disable test db on release auto nightly run

  #54563 openedJul 12, 2025

• Revert "[serve] reorganize how we handle the http receive task"

  #54565 openedJul 12, 2025

• Feat/fix request replica context

  #54566 openedJul 12, 2025

• [Core] Core Worker GetObjStatus GRPC Fault Tolerance

  #54567 openedJul 12, 2025

• [ci] use compiled list for install

  #54568 openedJul 12, 2025

• Feat/remove cpu profiler

  #54569 openedJul 12, 2025

• [llm.serve] Add support of `list[int]` for `CompletionRequest.prompt`

  #54570 openedJul 12, 2025

1,410 Issues closed by70 people

• [Azure] Ray up for Azure fails

  #48976 closedJul 12, 2025

• CI test linux://python/ray/tests:test_metrics_agent is consistently_failing

  #48956 closedJul 12, 2025

• CI test linux://python/ray/air:test_integration_wandb is consistently_failing

  #54553 closedJul 12, 2025

• [Core][autoscaler] autoscaler v2 tries to load default configs that do not exist on the image.

  #54532 closedJul 12, 2025

• CI test linux://rllib:examples/connectors/flatten_observations_dict_space_impala is flaky

  #49754 closedJul 12, 2025

• [Ray Data] Filtering function is very slow

  #53493 closedJul 11, 2025

• [Ray Data] Ray doesn't correctly handle variable shaped tensors in .map after change in blocklevel metadata in ray>2.45

  #54186 closedJul 11, 2025

• [core][gpu-objects] Garbage collection for in-actor GPU objects

  #51262 closedJul 11, 2025

• [core][gpu-objects] Actor sends the same ObjectRef twice to another actor

  #51273 closedJul 11, 2025

• [Core] Ray Data job hanging with flooded Cancelling stale RPC with seqno 125 < 127 error

  #50814 closedJul 11, 2025

• [core][autoscaler] Enable autoscaler v2 by default when running on KubeRay

  #54226 closedJul 11, 2025

• [Serve] refactor serve code that sets `docs_path`

  #53023 closedJul 11, 2025

• ray azure does not work out of the box

  #52511 closedJul 11, 2025

• [<Ray component: Core|RLlib|etc...>] Anaconda free python in Docker images

  #51991 closedJul 11, 2025

• CI test windows://python/ray/serve/tests:test_standalone is flaky

  #48420 closedJul 11, 2025

• CI test windows://python/ray/serve/tests:test_logging is consistently_failing

  #46043 closedJul 11, 2025

• [Serve] FastAPI ingress does not work with composable routers

  #50373 closedJul 11, 2025

• [Serve] ingress decorator does not work with fastapi.APIRouter arg

  #50372 closedJul 11, 2025

• [Serve] Proxy issues: Request cancellation, intermittent 503 backpressure, and max_queued_requests configuration not applied

  #53794 closedJul 11, 2025

• CI test windows://python/ray/tests:test_node_labels is consistently_failing

  #52307 closedJul 10, 2025

• [autoscaler][v2] Autoscaler stops working after the head node recovers with enabled FT

  #54353 closedJul 10, 2025

• CI test linux://rllib:examples/metrics/custom_metrics_in_algorithm_training_step is flaky

  #51870 closedJul 10, 2025

• [Serve] DeepSeek-R1 mode load stuck in H20

  #50975 closedJul 10, 2025

• [data][bug] Dataset execution can be implicitly triggered when passing a dataset to an Actor.

  #52549 closedJul 10, 2025

• [Data] Refactor `ParquetDatasink._write_partition_files` to use `pyarrow.parquet.write_to_dataset`

  #50502 closedJul 10, 2025

• [Core] ray distributed debugger, always connecting to cluster..

  #50682 closedJul 10, 2025

• [Data] __repr__ shouldn't trigger execution

  #50361 closedJul 10, 2025

• [data] RefBundle doesn't always eagerly free data

  #37910 closedJul 10, 2025

• [data -- read_iceberg] pickling error on UDF for dataset.groupby.map_batches

  #54280 closedJul 10, 2025

• [Data] `test_hudi` flakes in CI 25% of the time

  #50463 closedJul 10, 2025

• [core] default uv integration breaks runtime env pip packages (when running in existing environments?)

  #54344 closedJul 10, 2025

• CI test windows://python/ray/serve/tests:test_deploy is consistently_failing

  #46033 closedJul 10, 2025

• CI test linux://python/ray/serve/tests:test_standalone_2_with_compact_scheduling is flaky

  #48338 closedJul 10, 2025

• CI test linux://python/ray/serve/tests:test_standalone_2 is flaky

  #48403 closedJul 10, 2025

• [Serve] Specify different images for each deployment

  #52994 closedJul 10, 2025

• [Serve] Optimize the _get_live_deployments function

  #45793 closedJul 10, 2025

• CI test windows://python/ray/serve/tests:test_grpc is flaky

  #46028 closedJul 10, 2025

• [Ray serve] Unable to serve meta-llama/Llama-3.1-8B-Instruct

  #53663 closedJul 10, 2025

• [Serve] Unable to load meta-llama/Llama-3.3-70B-Instruct

  #53571 closedJul 10, 2025

• [Dashboard] Refactor job / node / actor updating code

  #16243 closedJul 10, 2025

• [Dashboard][event] Event API in Python.

  #16250 closedJul 10, 2025

• [Dashboard][event] Event API in Java.

  #16251 closedJul 10, 2025

• Ray Data: AssertionError when using repartition

  #54434 closedJul 10, 2025

• [Data] repartition(target_num_rows_per_block=...) causes unclear zero_copy_map_fusion AssertionError in Ray Data

  #54431 closedJul 10, 2025

• [Train V2 + Tune] Exception in Tune while processing results from failed runs using TorchTrainer

  #54379 closedJul 9, 2025

• [data][bug] repartition(target_num_rows_per_block) should not be fused with downstream op

  #54448 closedJul 9, 2025

• [Serve] Multiple FastAPI ingress deployments in a single application are not disallowed

  #53024 closedJul 9, 2025

• Release test gcp_cluster_launcher_full failed

  #54460 closedJul 9, 2025

• Release test aws_cluster_launcher_full failed

  #54452 closedJul 9, 2025

• [llm.serve] vllm.PoolingOutput no more has attr embedding

  #54466 closedJul 9, 2025

• [Core] thread creation error, even with environment variables all set to 1

  #54225 closedJul 9, 2025

• [Core] Too many threads in single pod on large-CPU-cores machine

  #54422 closedJul 9, 2025

• Release test air_example_vicuna_13b_lightning_deepspeed_finetuning failed

  #54308 closedJul 9, 2025

• Release test air_example_gptj_deepspeed_fine_tuning failed

  #54133 closedJul 9, 2025

• Release test air_example_dolly_v2_lightning_fsdp_finetuning failed

  #54307 closedJul 9, 2025

• CI test windows://python/ray/serve/tests:test_proxy is consistently_failing

  #51464 closedJul 8, 2025

• Release test sort_fixed_size failed

  #52650 closedJul 8, 2025

• Release test sort_autoscaling failed

  #53546 closedJul 8, 2025

• [Autoscaler][V2] Autoscaler V2 does not honor 'Conservative' upscaling mode

  #50259 closedJul 8, 2025

• CI test windows://python/ray/tests:test_object_assign_owner_client_mode is flaky

  #54333 closedJul 8, 2025

• [core|serve] Migrate shared utilities from `ray._private` to `ray._common`

  #53478 closedJul 8, 2025

• Release test dask_on_ray_large_scale_test_spilling failed

  #54347 closedJul 8, 2025

• CI test linux://rllib:learning_tests_multi_agent_stateless_cartpole_ppo_multi_cpu is flaky

  #47313 closedJul 7, 2025

• CI test linux://python/ray/data:test_json is flaky

  #48150 closedJul 7, 2025

• Release test dask_on_ray_1tb_sort failed

  #54356 closedJul 7, 2025

• Release test chaos_dask_on_ray_large_scale_test_spilling.aws failed

  #54346 closedJul 7, 2025

• Release test chaos_dask_on_ray_large_scale_test_no_spilling.aws failed

  #54345 closedJul 7, 2025

• CI test windows://python/ray/tests:test_object_store_metrics is flaky

  #49514 closedJul 7, 2025

• [core][gpu-objects] Ability to register custom types for GPU data

  #52340 closedJul 7, 2025

• CI test linux://python/ray/data:test_block_sizing is flaky

  #54164 closedJul 5, 2025

• CI test linux://rllib:learning_tests_multi_agent_cartpole_dqn_multi_gpu is consistently_failing

  #47234 closedJul 4, 2025

• CI test darwin://python/ray/tests:test_gcs_fault_tolerance is flaky

  #43777 closedJul 4, 2025

• CI test linux://python/ray/tests:test_runtime_env_container is consistently_failing

  #54319 closedJul 4, 2025

• CI test windows://python/ray/tests:test_implicit_resource is flaky

  #43849 closedJul 3, 2025

• CI test darwin://python/ray/tests:test_streaming_generator_2 is consistently_failing

  #54239 closedJul 3, 2025

• CI test darwin://python/ray/tests:test_scheduling_client_mode is consistently_failing

  #54202 closedJul 3, 2025

• Release test sort.chaos failed

  #49765 closedJul 2, 2025

• [Data] `dataset.write_iceberg` error

  #52967 closedJul 2, 2025

• [Train] Support for Callbacks in Ray Train Training Loop

  #54268 closedJul 2, 2025

• [Core] Add Azure Blob Storage Support for Ray runtime_env

  #38316 closedJul 2, 2025

• CI test linux://src/ray/gcs/gcs_server/test:gcs_actor_manager_test is flaky

  #54221 closedJul 2, 2025

• [Docker] [CI] Bump the GPU base image to a newer version

  #54102 closedJul 2, 2025

• CI test linux://python/ray/tests:test_scheduling_client_mode is consistently_failing

  #54160 closedJul 2, 2025

• Ray component: Core ray.init() fails on windows since #51731

  #54151 closedJul 1, 2025

• ERROR services.py:1355 -- Failed to start the dashboard , return code 3221226505

  #54165 closedJul 1, 2025

• [Core] ASSERTION FAILED: queue.num_items() == 0

  #53510 closedJul 1, 2025

• [air/output] Jupyter notebook trial result table keeps swapping column order

  #35838 closedJun 30, 2025

• [RLlib] Make Learner more standalone with regards to LearnerHyperparameters

  #35788 closedJun 30, 2025

• [AIR] `on_trial_complete` callback hook happens before trial resources are freed

  #35721 closedJun 30, 2025

• [core] Failed to close sockets in CoreWorker when crash.

  #35681 closedJun 30, 2025

• Ray Data - Glob/wildcard in file path

  #35499 closedJun 30, 2025

• [serve] Document how to silence access logs from GradioIngress

  #35496 closedJun 30, 2025

• [RLlib] Windows CLI, cmd.exe, powershell parsing json arguments JSONDecodeError

  #35492 closedJun 30, 2025

• [RayClient]large object transfer failure

  #35448 closedJun 30, 2025

• [train] Simplify `test_transformers_trainer_steps::test_e2e_steps`

  #35424 closedJun 30, 2025

• [Core] Reducing scheduling fragmentation

  #35422 closedJun 30, 2025

• [Core, RLlib] Multi GPU RLlib experiment is unable to be scheduled.

  #35409 closedJun 30, 2025

• [Job] Failed to schedule supervisor actor leads to job failure

  #35387 closedJun 30, 2025

• [Job] Show submitter of a Job on the dashboard

  #35367 closedJun 30, 2025

• [Serve] Support sync function for multiplexing

  #35356 closedJun 30, 2025

• [AIR] [Train] train multiple instances simultaneously on machines with specified tags

  #35333 closedJun 30, 2025

• <RLlib> What is the cause of the low CPU utilization in rllib PPO?

  #35313 closedJun 30, 2025

• [Data] Infer the data schema in Ray Datasets

  #35230 closedJun 30, 2025

• [dashboard] how to adjust ray dashboard refresh rate?

  #35156 closedJun 30, 2025

• [KubeRay, dashboard] Clarify that the users can use persistent volumes for log_dir and ray dashboard can read from it.

  #35137 closedJun 30, 2025

• [RLlib] Better error handling when return shape from step() mismatch in utils._flatten_multidiscrete

  #35113 closedJun 30, 2025

• The ray rsync-up cli reports no issue, but actually file is absent on remote side (Ray AWS cluster)

  #35051 closedJun 30, 2025

• [Core] - GPU Support - Explanation of Results

  #35048 closedJun 30, 2025

• [Data] Optimize `read_datasource` setup

  #35029 closedJun 30, 2025

• [EC2 VM Cluster launcher] Document EC2 ssh key limit and workaround

  #35020 closedJun 30, 2025

• [VM launcher] Ran `Ray status` after I sshed in to the head node and it printed "No cluster status"

  #35017 closedJun 30, 2025

• [air/tune][multi-tenancy] Parallel runs can use the same experiment directory

  #35006 closedJun 30, 2025

• Issue on page /cluster/vms/examples/ml-example.html

  #34996 closedJun 30, 2025

• [AWS VM Cluster Launcher] AWS Cluster launcher installs nightly Ray by default

  #34991 closedJun 30, 2025

• [CI] Fix minimal-install python 3.11: build wheel with unsupported tags.

  #34980 closedJun 30, 2025

• [serve][docs] Add DAG building classes to the API reference

  #34953 closedJun 30, 2025

• [AIR output] Rich table gets truncated when the terminal height is smaller than it

  #34925 closedJun 30, 2025

• [AIR output] Format of trial table with Rich enabled.

  #34923 closedJun 30, 2025

• [AIR output] "iteration" is shown in the output for RL users

  #34918 closedJun 30, 2025

• [core] ray.kill doesn't guarantee resources are cleaned up

  #34917 closedJun 30, 2025

• [Data] Add `fn_kwargs` to `BatchMapper`

  #34852 closedJun 30, 2025

• Resource Allocation: Ray Core, Ray Client

  #34816 closedJun 30, 2025

• [Jobs] Job agent recovers all running jobs on restart, not just those monitored by that agent

  #34794 closedJun 30, 2025

• [Doc] Autogenerated "suggest an edit" link doesn't work

  #34751 closedJun 30, 2025

• [Tune] thread limit resulting in the job failure in multi-tenancy usage

  #34745 closedJun 30, 2025

• Ray Job

  #34710 closedJun 30, 2025

• [docs][infra] automate checks for common link errors

  #34681 closedJun 30, 2025

• [Ray Job] Auto-shutdown of the cluster when job finished

  #34672 closedJun 30, 2025

• [Core] Ray.wait should return if task throw exception

  #34653 closedJun 30, 2025

• [Core] ray2.3.1 gcs_server memory keeps increasing until OOM

  #34619 closedJun 30, 2025

• [Runtime Env/Ray Job] Job submission fails when specifing local zip file as working dir

  #34605 closedJun 30, 2025

• why ray.data.read_images cat not combine_chunks

  #34563 closedJun 30, 2025

• [Core] Add support for cancelling descendants of a completed task

  #34545 closedJun 30, 2025

• [Data] retrieve written paths from `Dataset.write_datasource`

  #34444 closedJun 30, 2025

• [Docs Infra] [RLLib] Remove "<<<" from code blocks

  #34439 closedJun 30, 2025

• [Serve] Production Guide: Add instruction for non-K8s on-premise clusters

  #34437 closedJun 30, 2025

• [Serve] Ray Serve hangs and becomes unresponsive when calling ffmpeg in deployment

  #34414 closedJun 30, 2025

• [Serve] Deployments page tasks history is full of system tasks. Not very useful

  #34386 closedJun 30, 2025

• [Core] serialisation of dataclass in separate module fails to recognise parameter change in child dataclass, but functions correctly if in the same module

  #34366 closedJun 30, 2025

• ImportError: cannot import name 'torch' from 'ray.rllib.train'

  #34354 closedJun 30, 2025

• [core][state] Include job info for placement group

  #34333 closedJun 30, 2025

• [Jobs] Use new API `is_head_node` to find head node

  #34317 closedJun 30, 2025

• [Core] RFC: simplify CI testing

  #34315 closedJun 30, 2025

• [air] Error while loading xgboost model in BatchPredictor

  #34307 closedJun 30, 2025

• [RLlib] Unity 3d env tests are broken

  #34290 closedJun 30, 2025

• [air/train] the logic to grab free ports for `tf_config` is potentially racy

  #34271 closedJun 30, 2025

• [Core][Object Store] Push Manager: round for object manager client and FIFO for object

  #34270 closedJun 30, 2025

• [air] xgboost/lightgbm trainer's validation result differ between online and offline

  #34211 closedJun 30, 2025

• [tune] support viewing partial experiment result as tuning goes on

  #34207 closedJun 30, 2025

• [Workflow] Improve efficiency of Ray Workflow by returning workflow metadata and completed task information in single API call

  #34158 closedJun 30, 2025

• Issue on page /rllib/package_ref/algorithm.html

  #34157 closedJun 30, 2025

• [Prometheus metrics util] Application level custom metrics aren't getting exported consistently

  #34145 closedJun 30, 2025

• [Core] Actors not cleaning up resources correct because `force_kill=true`.

  #34124 closedJun 30, 2025

• Ray Tune + ray xgboost running out of disk space

  #34118 closedJun 30, 2025

• [Core][Tune]Trials hang when using Pytorch

  #34028 closedJun 30, 2025

• [Data] `map_batches` hard to use and debug

  #34007 closedJun 30, 2025

• [Core] improve garbage collection after job go out of scope

  #34001 closedJun 30, 2025

• [Core] Timeout for unschedulable task due to unavailable workers

  #33954 closedJun 30, 2025

• [Observability] Programmatically fetch prometheus metrics

  #33940 closedJun 30, 2025

• [Ray AIR] Add more documentation about checkpointing

  #33932 closedJun 30, 2025

• Ray Workflow

  #33844 closedJun 30, 2025

• [Train] Intermittent `UnpicklingError` when loading estimator/preprocessor from checkpoint

  #33815 closedJun 30, 2025

• [AIR output] Warnings for AIR_VERBOSITY is confusing

  #33810 closedJun 30, 2025

• [air output] Aggregation of feedback for air output v2

  #33803 closedJun 30, 2025

• [Datasets] `FileBasedDataSource`s do not pass `filesystem` to `_read_stream()` methods' `reader_args`

  #33777 closedJun 30, 2025

• [Core][Runtime Env] Document how to write custom runtime env plugin

  #33746 closedJun 30, 2025

• Core: Can the ray core's scheduling mechanism support customized extensions?

  #33735 closedJun 30, 2025

• [Ray init] Ray init method does not support pathlib.Path

  #33672 closedJun 30, 2025

• [docs] improve user experience of the API ref

  #33645 closedJun 30, 2025

• [RLLib] Collecting external experience

  #33636 closedJun 30, 2025

• [Workflow] get_metadata(workflow_id)["status"] and get_status(workflow_id) not returning the same status

  #33633 closedJun 30, 2025

• [runtime_env] Actors always depend global `pip` field for `runtime_env`

  #33607 closedJun 30, 2025

• [Core] Raylet process not respecting `--node-ip-address`

  #33554 closedJun 30, 2025

• [Tune] Support ExperimentAnalysis.dataframe(mode='mean')

  #33540 closedJun 30, 2025

• [Train] `RunConfig` doesn't get propagated from the Tuner to the Trainer

  #33539 closedJun 30, 2025

• [Core] std::bad_alloc error using ray.init()

  #33525 closedJun 30, 2025

• [Core] `test_memory_deadlock` times out

  #33491 closedJun 30, 2025

• [Core] Support binding worker processes to NUMA nodes

  #33465 closedJun 30, 2025

• [Serve] Support for setting `working_dir` to a local directory in `RayService`

  #33456 closedJun 30, 2025

• RLLIB - RE3 Exploration Algorithm - No GPU support f0r Dynamic TF V2

  #33425 closedJun 30, 2025

• [client] kubernetes w ray client

  #33367 closedJun 30, 2025

• [Train] Reporting metrics/checkpoints from multiple workers

  #33360 closedJun 30, 2025

• [Data] `read_parquet` schema is incorrect (schema is a dict instead of a string)

  #33279 closedJun 30, 2025

• [Ray status] confusing output about gpus and accelerators

  #33272 closedJun 30, 2025

• [Tune] mlflow logger callback > log_trial_result fail (psycopg2.ProgrammingError) can't adapt type 'numpy.int64'

  #33233 closedJun 30, 2025

• [Serve] Enhance replica upgrade process.

  #33192 closedJun 30, 2025

• [air output] Isolate/refactor/improve rllib related progress reporting logic

  #33150 closedJun 30, 2025

• [Tune][wandb] Report tune experiments as a wandb `sweep`

  #33142 closedJun 30, 2025

• [AIR][wandb] Add option to track artifact references in wandb if using cloud storage

  #33130 closedJun 30, 2025

• [AIR][Tune] Add an option in `WandbLoggerCallback` to group wandb runs by config

  #33084 closedJun 30, 2025

• [Serve] Support external storage for state

  #33059 closedJun 30, 2025

• [Serve] Use the namespace of context instead of "serve" when the Controller gets all running Actors

  #33057 closedJun 30, 2025

• [Serve] Specify replicas when scaling down

  #33056 closedJun 30, 2025

• [Serve] Restart a batch of replicas by Actor names or replica tags

  #33055 closedJun 30, 2025

• [Serve] Specify a batch of replicas to update their user_config

  #33054 closedJun 30, 2025

• Ray Core Runtime Environments with tea.xyz

  #33049 closedJun 30, 2025

• [Ray Tune] Support for continuing training when metrics are only reported from some of the workers

  #33042 closedJun 30, 2025

• [Data] Cannot get the length of a tf dataset created from `ray_ds.to_tf`

  #33004 closedJun 30, 2025

• [Data] Include image class id in the returned datasets of `ray.data.read_images()`.

  #32989 closedJun 30, 2025

• [Datasets] Raise descriptive error if `iter_torch_batches` can't convert data

  #32953 closedJun 30, 2025

• [Serve] Don't start Serve agent if Serve isn't installed

  #32920 closedJun 30, 2025

• [Data]: `ds.take()` and `ds.iter_batches()` have unexpected different behavior for pd.Series columns

  #32913 closedJun 30, 2025

• [Ray: Serve] Model Composition primitives should be part of Serve Core API docs.

  #32837 closedJun 30, 2025

• [core][state] Task backend : already submitted cancelled task showing up as finished

  #32826 closedJun 30, 2025

• [AIR][Tune] Make trial checkpoint + artifact upload happen atomically

  #32823 closedJun 30, 2025

• [Tune] During multi-GPU training (using mp.spawn), ray.tune.report does not take effect.

  #32810 closedJun 30, 2025

• [Tune] failure when using more than one GPU

  #32760 closedJun 30, 2025

• [Runtime Env] Add docstring for public class methods and attributes

  #32704 closedJun 30, 2025

• [tune] Add suggestions on when `reuse_actor` should be set to false.

  #32698 closedJun 30, 2025

• [serve] serve run doesn't restart app successfully in some environments

  #32633 closedJun 30, 2025

• [train] Big performance hit when TensorFlow trainer is not scheduled on head node

  #32509 closedJun 30, 2025

• [doc][tune] clarify `Stopper`, what is `training_iteration`

  #32497 closedJun 30, 2025

• [release] update our xgboost release test to catch issues like (see discription)

  #32491 closedJun 30, 2025

• [Core] The remote function in the worker no longer runs after the head crashes

  #32454 closedJun 30, 2025

• [RLlib] Special __common__ key in MultiAgent batches is not documented

  #32399 closedJun 30, 2025

• [tune] update how trainable reports result/checkpoint to driver

  #32380 closedJun 30, 2025

• [Datasets] The projection pushdown cannot work with hive style partitioning file path

  #32301 closedJun 30, 2025

• [Core][utilization] some anti-pattern that not well supported by Ray core.

  #32297 closedJun 30, 2025

• [tune/train] Provide actionable error messages for common thirdparty errors

  #32232 closedJun 30, 2025

• [ci] Mirror external dependenies in CI

  #32113 closedJun 30, 2025

• [Serve] ValueError: Message ray.serve.ReplicaConfig exceeds maximum protobuf size of 2GB

  #32049 closedJun 30, 2025

• [CLI] make `ray get-head-ip` and `ray get-worker-ips` work for kuberay clusters when run outside the cluster

  #32037 closedJun 30, 2025

• Serve build usage of click CLI library conflicts python argparse

  #32001 closedJun 30, 2025

• [Serve] Version Support in 2.X API

  #31928 closedJun 30, 2025

• [Train] User exceptions not propagated from remote cluster

  #31913 closedJun 30, 2025

• [RLlib] AlgorithmConfig() defaults not used by build_sac_model when implementing custom model

  #31783 closedJun 30, 2025

• [kubernetes/cluster] More guides on deployment

  #31623 closedJun 30, 2025

• [core][state] ray log supporting regex searching

  #31549 closedJun 30, 2025

• [Tune] Support NLopt search algorithms

  #31492 closedJun 30, 2025

• [Rllib] Possible Redudant Code

  #31463 closedJun 30, 2025

• [aws] ray submit --stop fails on aws

  #31380 closedJun 30, 2025

• [Tune] Avoid insufficient resources warning if cluster is autoscaling

  #31292 closedJun 30, 2025

• No worker logs in the dashboard after recreating the K8S Ray pods

  #31288 closedJun 30, 2025

• [core] Please improve warning message for ip mismatch

  #31264 closedJun 30, 2025

• [core][state] Refactor use of bounded LRU/FIFO buffer/map used in task backend

  #31158 closedJun 30, 2025

• [core] Ray resources should be case-insensitive

  #31087 closedJun 30, 2025

• [RayCluster]

  #31041 closedJun 30, 2025

• [Serve] gRPCis should not allow route_prefix set

  #30891 closedJun 30, 2025

• [General] Setup a "code walkthrough" meetup or tutorial

  #30852 closedJun 30, 2025

• [RFC][core] Option to avoid scheduling tasks to nodes with disk full

  #30843 closedJun 30, 2025

• [core] Enable greater control over log verbosity

  #30832 closedJun 30, 2025

• [Tune] ability to specify search algorithm when using tune.run_experiments()

  #30802 closedJun 30, 2025

• [RLlib] Deprecate the RLlib spaces that are duplications of gym spaces.

  #30800 closedJun 30, 2025

• [Tune] Guard against users overriding internal `Trainable` methods

  #30795 closedJun 30, 2025

• Ray Cluster Resources Issue

  #30780 closedJun 30, 2025

• [Core] Worker leak

  #30731 closedJun 30, 2025

• [RLlib] Default policy error in two trainer work flow

  #30676 closedJun 30, 2025

• [core] Can't set working directory for runtime env in actor definition

  #30666 closedJun 30, 2025

• [Tune] HeboSearch reproducible deterministic results

  #30661 closedJun 30, 2025

• [core] Memory changes are not as expected when using ray.get()

  #30615 closedJun 30, 2025

• [Tune] `fail_fast` marks all runs as terminated, making the experiment impossible to restore

  #30584 closedJun 30, 2025

• [RLLib] Custom model with LSTM causes the auto wrapping to be partially executed

  #30581 closedJun 30, 2025

• [Core|RayTrain] RuntimeError: Some workers returned results while others didn't

  #30545 closedJun 30, 2025

• [Core] Overriding the default logging format for Worker logs

  #30544 closedJun 30, 2025

• [AIR] Canonical way to determine whether the code is running in a Train/Tune session

  #30536 closedJun 30, 2025

• [client][runtime_env] Inconsistent runs on ray client

  #30518 closedJun 30, 2025

• [Core] ray.exceptions.RaySystemError: System error: buffer source array is read-only

  #30505 closedJun 30, 2025

• [Core] Access violation on windows 11 when running modin workload

  #30493 closedJun 30, 2025

• Critic Regularized Regression (CRR) model is getting error with Custom Environment (Offline RL)

  #30411 closedJun 30, 2025

• [Docs] [Jobs] Add pros and cons of different ways of submitting a job

  #30305 closedJun 30, 2025

• [air/horovod] horovod distributed worker creation may hang

  #30276 closedJun 30, 2025

• [<Ray component: Workflow>] module 'ray.workflow' has no attribute 'HTTPListener'

  #30248 closedJun 30, 2025

• [RLLIB][Torch] numerically unstable + mkl issue in torch.sqrt normc_initializer

  #30191 closedJun 30, 2025

• [RLlib] RuntimeError: invalid multinomial distribution (sum of probabilities <= 0)

  #30164 closedJun 30, 2025

• [AIR][Tune] Provide user guide on how to build active learning on AIR

  #30157 closedJun 30, 2025

• [AIR/Docs] Mention/warn that running a Trainer inside a custom Tune trainable is an anti-pattern

  #30153 closedJun 30, 2025

• [RLlib] (PPO) algo parameter "lambda_" never gets passed because `AlgorithmConfig` refractored "lambda_" to "lambda"

  #30143 closedJun 30, 2025

• [Core] Reference leakage somewhere after ray.shutdown()

  #30089 closedJun 30, 2025

• [Tune] Can't access all metrics for all trials

  #30004 closedJun 30, 2025

• [core][dashboard] state api on worker nodes can not connect to dashboard url

  #29959 closedJun 30, 2025

• [Jobs] Include requested and available resources in JobInfo status message

  #29921 closedJun 30, 2025

• [RLlib] Add some metric for aync algos (e.g. APPO) that shows the total number of gradient updates

  #29830 closedJun 30, 2025

• [AIR] Update pytorch training and prediction benchmark with numpy with updated metrics

  #29743 closedJun 30, 2025

• [RLlib] Undesired memory growing when using convolutional neural network

  #29699 closedJun 30, 2025

• [AIR] `XGBoostTrainer` gives misleading error if column missing

  #29695 closedJun 30, 2025

• [RLLib Tests] : Included pytests in package as well as basic commands fail with ValueError

  #29691 closedJun 30, 2025

• [air] GPU memory leak when using AIR trainer with torch dataloader when the latter uses multi-processing

  #29563 closedJun 30, 2025

• [RLlib] Benchmark bandit methods vs plain Thompson Sampling for a non-contextual MAB

  #29528 closedJun 30, 2025

• [RLlib] "model": {"free_log_std": True} generates Tensorflow Lambda layers warning with TF2 framework

  #29502 closedJun 30, 2025

• [Autoscaler] Delete AWS resources created when launching Ray cluster upon cluster termination

  #29499 closedJun 30, 2025

• [Core] util.multiprocessing.pool: imap and imap_unordered blocking on ray.wait even though processes are complete

  #29466 closedJun 30, 2025

• [Ray Log_monitor]: close_all_files ProcessLookupError

  #29452 closedJun 30, 2025

• Ray core: incorrect account of GPUs on ec2 ubuntu instance: g4dn.2xlarge

  #29420 closedJun 30, 2025

• [core] GCS segfaults under OOM

  #29336 closedJun 30, 2025

• [AIR] Add progress bar for training

  #29314 closedJun 30, 2025

• [CI] A simple way to reproduce osx/linux/windows CI run failure locally

  #29068 closedJun 30, 2025

• [Core] Is it possible to do asynchroneous task submission?

  #29039 closedJun 30, 2025

• [doc][core] multiprocessing.Pool should document resource usage with ray_remote_args

  #29004 closedJun 30, 2025

• [Train] Allow passing in placement group to trainer

  #28924 closedJun 30, 2025

• [<Algorithm overview>]

  #28915 closedJun 30, 2025

• Runtime Environment Dependencies- container per task

  #28875 closedJun 30, 2025

• [Ray component: Core] Returning to much data from ray remote fails with no error

  #28855 closedJun 30, 2025

• Issue on page /ray-core/examples/plot_parameter_server.html

  #28854 closedJun 30, 2025

• [Datasets] Why does pydantic make training slower？

  #28836 closedJun 30, 2025

• [Infra] Improve Ray client usability

  #28790 closedJun 30, 2025

• [Core] Download Logs from Ray Dashboard

  #28788 closedJun 30, 2025

• Ray Core: AttributeError: 'NoneType' object has no attribute 'enum_types_by_name'

  #28779 closedJun 30, 2025

• [Tune] HyperOptSearch fails with nested config dicts and points_to_evaluate

  #28753 closedJun 30, 2025

• Ray Deployment crashes in docker [<Ray component: Serve>]

  #28732 closedJun 30, 2025

• [Ray Serve]: Testing out on local using Docker container

  #28692 closedJun 30, 2025

• [core] Generator task that returns more values than specified by num_returns should throw error instead

  #28689 closedJun 30, 2025

• [Core] CloudPickle explain tool

  #28585 closedJun 30, 2025

• [dashboard] Dashboard randomly not showing the status of worker nodes.

  #28569 closedJun 30, 2025

• [AIR] Status updates still prints even with breakpoint

  #28554 closedJun 30, 2025

• [AIR/Tune] Session report does not show the key for those not included in the first metrics report

  #28549 closedJun 30, 2025

• [Core] dump the info and anaylze the data offline

  #28496 closedJun 30, 2025

• [Core] Document what are the generic python code that's easily scalable.

  #28487 closedJun 30, 2025

• [AIR] Refactor checkpoint encoding and decoding out of Backend to framework-specific Checkpoints

  #28462 closedJun 30, 2025

• [Core] [RLlib] RLlib on Ray 2.0 not easily working on Colab

  #28457 closedJun 30, 2025

• [Job Submission] Support env file input in ray.runtime_env.RuntimeEnv

  #28453 closedJun 30, 2025

• [Tune] Adding DEHB

  #28427 closedJun 30, 2025

• [serve] Gradio integration does surface error messages, runs indefinitely

  #28399 closedJun 30, 2025

• [Ray: Core] Ray can hang when getting an ObjectRef from an unknown environment

  #28341 closedJun 30, 2025

• [Core][RuntimeEnv]Make `job_submission_id` to a new index of GCS::JobTableData

  #28337 closedJun 30, 2025

• [Jobs] Run jobs tests on Windows

  #28316 closedJun 30, 2025

• [Core] Job stop should terminate runtime_env setup

  #28221 closedJun 30, 2025

• [Core] log_to_driver=False does not suppress worker errors in ipython

  #28216 closedJun 30, 2025

• [core][runtime envs] Ray should respect CUDA_VISIBLE_DEVICES if set in runtime env

  #28215 closedJun 30, 2025

• [Core] ray dashboard <rayhost>:8265/nodes?view=details cpuPercent should contains actor's subprocess

  #28100 closedJun 30, 2025

• [Core] Ray may hang if workers fail to start due to limited ports

  #28071 closedJun 30, 2025

• [<Ray component: Core|Cluster>] Documentation instructions for mounting AWS EFS Fails for Ray Cluster

  #28057 closedJun 30, 2025

• [Core] Support retry_delay option in Ray tasks

  #28015 closedJun 30, 2025

• [Core] Ray object primary copy transfer

  #27985 closedJun 30, 2025

• [Observability] ray timeline errors with ray.rpc.GetAllProfileInfoReply exceeded maximum protobuf size of 2GB

  #27952 closedJun 30, 2025

• [Core] allow customized error message for WorkerCrashedError

  #27947 closedJun 30, 2025

• tensorflow.python.framework.errors_impl.NotFoundError: ./multi_worker_model/variables/variables_temp/part-00000-of-00001.index; No such file or directory [Op:MergeV2Checkpoints]

  #27938 closedJun 30, 2025

• [observability] JSON file generated by Ray timeline doesn't render correctly in the new version of chrome tracing (perfetto)

  #27921 closedJun 30, 2025

• [RLlib] make policy evaluation support Attention nets

  #27909 closedJun 30, 2025

• [tune] allow using (nested) data classes for search space definition

  #27904 closedJun 30, 2025

• [Autoscaler][GCP] Autofill GCP node type resources

  #27888 closedJun 30, 2025

• [Core] The Idle worker killing feature slows down tasks

  #27863 closedJun 30, 2025

• [AIR] Support TorchRec trainer

  #27794 closedJun 30, 2025

• [Dashboard] Dashboard agent cannot be started because the port is still occupied

  #27736 closedJun 30, 2025

• [Tune/RLlib] log_to_file creates files, but doesn't write anything there

  #27702 closedJun 30, 2025

• [RLLib] global_timestep not monotonic when when running concurrent episodes with ExternalEnv

  #27669 closedJun 30, 2025

• [Dashboard] Ray Dashboard not showing the SpillWorker's actual memory usage

  #27591 closedJun 30, 2025

• [Core] The actors got distributed to just a few nodes even with spread scheduling

  #27577 closedJun 30, 2025

• [runtime_env] Add tests for all driver output (warnings, etc)

  #27566 closedJun 30, 2025

• [Tune] TuneReportCheckpointCallback causes two checkpoints to made every time it is called.

  #27524 closedJun 30, 2025

• [AIR] SettingWithCopyWarning for "A value is trying to be set on a copy of a slice from a DataFrame"

  #27352 closedJun 30, 2025

• [ray dashboard] profile button not working

  #27211 closedJun 30, 2025

• [Ray Train] Ray Train running slow when multiple workers executed

  #27107 closedJun 30, 2025

• [workflow] We should give the storage a default value if it's not set in some way.

  #27046 closedJun 30, 2025

• [State Observability][Log] Allow to ctrl + C when running logs API

  #27008 closedJun 30, 2025

• [runtime env] local `working_dir` doesn't work with strongly-typed `RuntimeEnv`

  #26984 closedJun 30, 2025

• [Core][State Observability] More fine-grained exceptions/error codes handling

  #26974 closedJun 30, 2025

• [Core] Typing for .options for Ray Tasks

  #26871 closedJun 30, 2025

• [State Observability] Support filter None value

  #26820 closedJun 30, 2025

• [Core] Batch PinObjectIDs requests from Raylet client

  #26796 closedJun 30, 2025

• [Train] feature request for catboost_ray

  #26687 closedJun 30, 2025

• [AIR/Tune] Add a `ScalingConfig`-based API to `ResourceChangingScheduler`

  #26538 closedJun 30, 2025

• [RLlib] CRR and CQL consume more cpus than reported

  #26533 closedJun 30, 2025

• [Job] Job submission not following convention for quote

  #26514 closedJun 30, 2025

• [RLlib] Unable to call ray.remote functions inside env/action dist

  #26468 closedJun 30, 2025

• [Core] Observing Multiple Exceptions When Using Different Python Patch Versions

  #26443 closedJun 30, 2025

• [RLlib] Use observations (input_dict) for exploration

  #26437 closedJun 30, 2025

• [core][c++ worker] RayClusterModeTest.DefaultActorLifetimeTest timed out in macOS

  #26435 closedJun 30, 2025

• [Ray component: Core] Enable better progress bar

  #26426 closedJun 30, 2025

• [RLlib] Issue Regarding Future Warnings

  #26424 closedJun 30, 2025

• [doc][Core | State Observability] Document usage of the rate limiting env variable in public doc

  #26370 closedJun 30, 2025

• [Tune] NevergradSearch Budget Exception

  #26305 closedJun 30, 2025

• How to color-code console output

  #26226 closedJun 30, 2025

• [Core] [Quality] Live handle raises unnecessary exception when script ends

  #26198 closedJun 30, 2025

• [RLlib]: SimpleQ TF2 is broken

  #26192 closedJun 30, 2025

• [State Observability] Raise an exception if the state schema contains predicates.

  #26125 closedJun 30, 2025

• [air] We should have a convenient method for user to interact with checkpoint file on driver when they checkpoint using other method in session

  #26082 closedJun 30, 2025

• [RLlib] server reports nan episodes and empty policy

  #26048 closedJun 30, 2025

• [test][autoscaler] ModuleNotFoundError: No module named 'ray.tests'

  #26023 closedJun 30, 2025

• [Tune] Ray Tune doesn't work inside Spark UDF

  #26002 closedJun 30, 2025

• [Serve] A Deployment Graph with unfulfilled demands fails to scale Pods in Kubernetes

  #25998 closedJun 30, 2025

• API server internal error message not useful

  #25986 closedJun 30, 2025

• [runtime env] Bad runtime env specified in ray.init() with eager install only raises error on task/actor invocation

  #25972 closedJun 30, 2025

• [Core][State Observability] Use a separate thread to run spill/restore

  #25960 closedJun 30, 2025

• [RLlib] KeyError: simple_list_collector.py, line 950, in postprocess_episode

  #25938 closedJun 30, 2025

• [RLLib] SampleBatch.update() doesn't update `added_keys`

  #25937 closedJun 30, 2025

• [runtime env] Use namespace for internal KV storage

  #25897 closedJun 30, 2025

• [Core?] Federation + data perimeters

  #25846 closedJun 30, 2025

• [Core] RBAC + auditability

  #25845 closedJun 30, 2025

• [Core] Arrow Flight Server doesn't work with Ray Actors due to two GRPC versions

  #25774 closedJun 30, 2025

• [Core | State Observability ] Refactor summary/log SDK to use StateApiClient

  #25746 closedJun 30, 2025

• [Serve] Deployment fails if name contains slashes

  #25714 closedJun 30, 2025

• [RLlib] ModelCatagolg Selects Wrong Model for Nested Complex Observations

  #25619 closedJun 30, 2025

• [Train] [Tune] When using Train with Tune, a `logdir` is created that's not the one specified by the user

  #25474 closedJun 30, 2025

• [Core][Observability] Ray memory should show more objects

  #25463 closedJun 30, 2025

• [Dashboard] Error during render node with gpu and 4 hdds

  #25437 closedJun 30, 2025

• [Ray Collective Lib] Enable CI

  #25396 closedJun 30, 2025

• Core: deamonset feature request

  #25334 closedJun 30, 2025

• [DeviceMesh][Collective] Support multiple tensors API

  #25129 closedJun 30, 2025

• [Ray Air] nan in the tensorflow_linear_dataset_example.py

  #25037 closedJun 30, 2025

• ray docker images do not have uvloop installed

  #25023 closedJun 30, 2025

• Ray Tune: No console output is logged to Wandb.

  #25011 closedJun 30, 2025

• [Core][RLlib][Tune] CUDA PTX error when training with Tune

  #25001 closedJun 30, 2025

• Ray component: Core: PoolActor processes hanging

  #24784 closedJun 30, 2025

• [RLlib] Duplicate custom metrics

  #24731 closedJun 30, 2025

• [Serve] Asynchronous inference best practices

  #24627 closedJun 30, 2025

• [tune] `progress_reporter.py` is messy and should be cleaned up

  #24604 closedJun 30, 2025

• [aws][autoscaler] AWS: When using spot instances, always single availability zone is selected

  #24310 closedJun 30, 2025

• [RLlib] PPO - ray.rllib.agents.ppo "Put Error"

  #24307 closedJun 30, 2025

• [Ray Collective] Remove Redis store and LocalFile store from gloo mode.

  #24288 closedJun 30, 2025

• [Autoscaler] upscaling_speed: 0 gets reset to 1

  #24177 closedJun 30, 2025

• [RLlib] Categorical action dist incorrectly uses tf.random.categorical

  #24055 closedJun 30, 2025

• [RLlib] [Bug] Inconsistent behavior between TFPolicy and TorchPolicy on `compute_actions_from_input_dict`

  #24007 closedJun 30, 2025

• [RLlib] Enable Training from Replay Buffer Larger than Memory

  #23816 closedJun 30, 2025

• [RLlib] [Bug] IMPALA causes an OOM after a long running.

  #23769 closedJun 30, 2025

• [BUG] Ray dashboard client failed to build

  #23548 closedJun 30, 2025

• [RFC][Feature][Autoscaler][Core]Graceful draining of nodes while scale-down

  #23522 closedJun 30, 2025

• [ml][Improvement] Improve messages to be “rank0, rank1” actors etc.

  #23310 closedJun 30, 2025

• [Feature] [tune] create a mlflow run name from config params

  #23228 closedJun 30, 2025

• [Feature][RLlib] Improve pytorch memory usage by disabling caching

  #23077 closedJun 30, 2025

• [tune][Bug] Worker doesn't sync the logs to HDFS at the given interval

  #23055 closedJun 30, 2025

• [Bug] AdaBelief optimizer crashes checkpoint restore

  #22976 closedJun 30, 2025

• [Bug] Custom model with R2D2

  #22747 closedJun 30, 2025

• [Bug] Resources displayed in Dashboard don't match cluster configuration

  #22548 closedJun 30, 2025

• [Bug] Deletion of Ray clusters hangs while Ray operator is still up

  #22505 closedJun 30, 2025

• Doing import ray breaks my logging [Bug]

  #22312 closedJun 30, 2025

• [Feature][Client] remove ray.disconnect() and ray.connect()

  #22125 closedJun 30, 2025

• [Bug] Detached actor exceptions are not logged.

  #21810 closedJun 30, 2025

• [Bug] Sometimes the worker node logs in the ray dashboard are empty

  #21785 closedJun 30, 2025

• [Core][Feature] Add checksum support for object store.

  #21782 closedJun 30, 2025

• Setting VF_SHARE_LAYERS to False and NO_FINAL_LINEAR to true leads to a bug

  #21756 closedJun 30, 2025

• [Feature] [runtime env] support using different python versions in Ray cluster

  #21597 closedJun 30, 2025

• [Feature] [Serve] Request Redistribution Among Replicas

  #21578 closedJun 30, 2025

• Put failed error occurred when shutdown and init again at client mode

  #21573 closedJun 30, 2025

• [Core] [Bug] No timeout or deadlock on scheduling job in remote cluster

  #21419 closedJun 30, 2025

• [Feature] [Autoscaler] Scaling Intelligently Based on Observed Resource Bottlenecks (related: task & actor profiling)

  #21301 closedJun 30, 2025

• [Core] [Bug] Failed to register worker to Raylet for single node, multi-GPU

  #21226 closedJun 30, 2025

• [Train] Port over `timm` example to Train

  #21020 closedJun 30, 2025

• [Bug] [RLlib] Custom metrics are not reported to Tune

  #20938 closedJun 30, 2025

• [Train] Deepspeed support

  #20648 closedJun 30, 2025

• [Bug] Cannot start cluster if other user is already running one

  #20634 closedJun 30, 2025

• [Bug] Excess memory usage when scheduling tasks in parallel?

  #20618 closedJun 30, 2025

• [Bug] Ray auto init interacts badly with allow_multiple=True and kills python shell

  #20355 closedJun 30, 2025

• [Bug] BasicVariantGenerator not compatible with Repeater

  #19879 closedJun 30, 2025

• [Feature] Support Sigopt for Tune standard space definitions

  #19018 closedJun 30, 2025

• [Bug] Re-enable Worker in Container Tests.

  #18787 closedJun 30, 2025

• [GCP][autoscaler] Rework ray TPU demos to create nothing but TPU VMs (no harddrives / `n2-standard-2` instances)

  #18645 closedJun 30, 2025

• [datasets] `random_shuffle` overspills objects on random node

  #17612 closedJun 30, 2025

• [Core] Ray Actor abnormal exit problem && Reproduction

  #17198 closedJun 30, 2025

• Support resizing placement groups

  #16403 closedJun 30, 2025

• [dashboard] Errors are not shown

  #15238 closedJun 30, 2025

• changing the docker image in consecutive `ray up` calls fails.

  #14990 closedJun 30, 2025

• [metrics] Add regression tests for Prometheus metrics

  #14614 closedJun 30, 2025

• [dashboard] Show more nodes at a time instead of paging through

  #14537 closedJun 30, 2025

• [autoscaler] Support Memory Aware Scheduling on a multi-node-type cluster.

  #14104 closedJun 30, 2025

• [RFC] k8s-native worker pool

  #14077 closedJun 30, 2025

• [dashboard] clicking on a column to sort makes the UI blank

  #13525 closedJun 30, 2025

• Autoscaler does not respect --num-cpus argument to `ray start`

  #13270 closedJun 30, 2025

• [core] Number of CPUs in ray.available_resources() does not match Dashboard's Machine View

  #13100 closedJun 30, 2025

• atexit handlers don't run when actor is terminated from going out of scope

  #12806 closedJun 30, 2025

• Task Cancellation is broken for queued tasks

  #12080 closedJun 30, 2025

• [logging] Use 'warnings.warn' appropriately

  #12060 closedJun 30, 2025

• [Dashboard] New dashboard port errors in a large cluster.

  #11638 closedJun 30, 2025

• ES Trainer does not support evaluation workers

  #10999 closedJun 30, 2025

• [Plasma] Improve plasma documentation on distributed storage

  #10858 closedJun 30, 2025

• Unable to connect to ray head running on linux from ray worker node on windows

  #10362 closedJun 30, 2025

• Ray log tracing

  #9786 closedJun 30, 2025

• [Core] Logging policy should be clearly defined and needs unit test coverage

  #9692 closedJun 30, 2025

• [dashboard] Error on Infinity values

  #9103 closedJun 30, 2025

• "ray timeline" command fails when RAY_ADDRESS is set

  #8951 closedJun 30, 2025

• [tune] [dashboard] Table formatting issues due to too many hparams

  #8667 closedJun 30, 2025

• [autoscaler] "Cannot perform an interactive login from a non TTY device" when trying to use a private docker registry

  #7339 closedJun 30, 2025

• Allowing multiple users to access a single ray cluster

  #6800 closedJun 30, 2025

• [Ray core & ray cluster] Add diagrams/architectures to explain how to run ray locally vs remotely

  #25663 closedJun 30, 2025

• [Ray Clusters] Remove nightly and latest images and wheels from all example configs.

  #25606 closedJun 30, 2025

• [air] Consider having a preprocessor for Feast integration

  #25559 closedJun 30, 2025

• [Core] Open telemetry Context pass from ray client to actors

  #25538 closedJun 30, 2025

• [dataset] Reduce tasks in push-based shuffle are not evenly distributed

  #25468 closedJun 30, 2025

• [Core] [State Observability] List all actor logs when actors are restarted.

  #25443 closedJun 30, 2025

• [air] Ordinal Encoder complains about None

  #25442 closedJun 30, 2025

• [Autoscaler] google-cloud-storage seems cannot read GOOGLE_APPLICATION_CREDENTIALS

  #25308 closedJun 30, 2025

• [Serve] Dynamically move models between CPUs and GPUs

  #25295 closedJun 30, 2025

• [RLlib][Doc] Add documentation for `ModelCatalog.get_model_v2()`

  #25186 closedJun 30, 2025

• [AIR] MLflow integration polish

  #25156 closedJun 30, 2025

• [AIR] TensorFlow warns to use `distribute.MultiWorkerMirroredStrategy` when I'm already using it

  #25140 closedJun 30, 2025

• [air] Have a default column for not frequent enough categories for OHE

  #25096 closedJun 30, 2025

• [Core] Make NodeManager unit testable

  #25095 closedJun 30, 2025

• [AIR] Improve logging for train

  #25088 closedJun 30, 2025

• [RLlib] Hope RLlib can support DQfD & POfD

  #25058 closedJun 30, 2025

• [AIR] Support postprocessing in Predictors

  #24979 closedJun 30, 2025

• [AIR] Add a `TorchVision` preprocessor

  #24976 closedJun 30, 2025

• [AIR/Train] Torch: Automatically unpack model when checkpointing state dicts

  #24975 closedJun 30, 2025

• [AIR/Train] Automatically return the framework specific dataset in `train_loop_per_worker`

  #24974 closedJun 30, 2025

• [<Ray component: RLlib>] ppo error when not using critic

  #24907 closedJun 30, 2025

• [RLlib]: Add tabular models to ModelV2

  #24882 closedJun 30, 2025

• [RLlib] Error when converting GYM Robotics env to Multi-agent Env with the make_multi_agent wrapper

  #24881 closedJun 30, 2025

• [tune] SigOptSearch suggester is not serialisable

  #24864 closedJun 30, 2025

• [core] Add basic metrics for lineage reconstruction

  #24855 closedJun 30, 2025

• [Core] Enhance runtime env state when `ray list runtime-env` is used.

  #24838 closedJun 30, 2025

• [Core] Refactor Ray memory codepath to follow same pattern as `ray list tasks`.

  #24836 closedJun 30, 2025

• [Core] Reach parity of task status for `ray memory` and `ray list tasks`

  #24835 closedJun 30, 2025

• [Tune] `MedianStoppingRule` mishandles `nan`s

  #24809 closedJun 30, 2025

• [Serve] Simplify json_serde of deployment graph

  #24620 closedJun 30, 2025

• [RLlib] Metrics not reported with Client/Server and env=None

  #24601 closedJun 30, 2025

• [Core] In C++, there are D_GLIBCXX_USE_CXX11_ABI settings conflicts when both Ray and Arrow are used.

  #24566 closedJun 30, 2025

• [Serve] `Deployment.url` not updated after options changing name or prefix.

  #24548 closedJun 30, 2025

• [Rllib] Lack validation for "num_workers" parameter in DDPGTrainer.

  #24536 closedJun 30, 2025

• [doc] Update instructions for wheel installation

  #24533 closedJun 30, 2025

• [RLlib] Simplex action space shape

  #24529 closedJun 30, 2025

• [Tune] Make it easy to configure logger level

  #24447 closedJun 30, 2025

• [tune] improve documentation around "resource exhausted error"

  #24439 closedJun 30, 2025

• [Core] Unify RegisterClient and AnnounceWorkerPort

  #24432 closedJun 30, 2025

• [core] Annotation and docstring for ray.remote wrapped functions

  #24411 closedJun 30, 2025

• [AIR] `Result` object doesn't work with Ray Client

  #24396 closedJun 30, 2025

• [RLlib] Current Implementation of Replay Buffer is not a True Circular Buffer

  #24393 closedJun 30, 2025

• [RLlib] wrong env step counting when train multi-agent with shared default policy

  #24340 closedJun 30, 2025

• [Core][observability] Enable observability features built in gRPC

  #24327 closedJun 30, 2025

• [Autoscaler][Docs] Add up-to-date docs on how the autoscaler works.

  #24323 closedJun 30, 2025

• [Ray Serve Autoscaling] Add release test that checks that nodes scale down when there are no requests

  #24315 closedJun 30, 2025

• Received message larger than max (105683136 vs. 104857600)

  #24286 closedJun 30, 2025

• [Serve] [Doc] HTTP Adapters Cookbooks

  #24245 closedJun 30, 2025

• [Serve] Default DAGDriver implementation cannot serve.run() or serve.build() twice

  #24122 closedJun 30, 2025

• [AIR] Support functionality to stitch Preprocessor with Keras model

  #24023 closedJun 30, 2025

• [Core] Log propagation between actor exit called and process terminated

  #24020 closedJun 30, 2025

• [<Ray component: Serve] Improve access by index/key on intermediate result in Serve deployment graph

  #23987 closedJun 30, 2025

• [Serve] [Docs] Improve architectural diagrams

  #23956 closedJun 30, 2025

• [Runtime Env] Dependency Installation private git repositories via ssh

  #23768 closedJun 30, 2025

• [ray client] ray.wait timeout is not respected when connection is interrupted

  #23694 closedJun 30, 2025

• [Feature] [Tune] Trial-wise dependencies

  #23654 closedJun 30, 2025

• [Bug] `policies_to_train` throws incorrect/confusing error message when passed an empty list.

  #23646 closedJun 30, 2025

• [Feature] support of complicated action space in QMix algorithm in Rllib.

  #23634 closedJun 30, 2025

• [runtime env] Deflake `test_runtime_env_working_dir_2`

  #23569 closedJun 30, 2025

• [runtime env] [Feature] Make Internal KV operations async

  #23567 closedJun 30, 2025

• [Feature] .bind() on function does not take pre-bind value from upstream DAGNode

  #23511 closedJun 30, 2025

• [RLlib][Bug] RLLib Dreamer tuned example requesting unreasonable amount of GPU memory

  #23479 closedJun 30, 2025

• [Core] Add a warning message if options / arguments differ for Actor.options(get_if_exists=True)

  #23455 closedJun 30, 2025

• [RLlib][Feature] Feature Importance Plots

  #23447 closedJun 30, 2025

• [air] If you kill train via control C, a bunch of random error messages show up next time you run Train.

  #23431 closedJun 30, 2025

• [air] Logging message is not relevant to user

  #23430 closedJun 30, 2025

• [RLlib][docs] Adding more flow charts to RLlib components docs

  #23393 closedJun 30, 2025

• [runtime env] Warn user if pip check fails

  #23335 closedJun 30, 2025

• [runtime env] Refactor packaging code

  #23257 closedJun 30, 2025

• [Feature] Cleanup current use of `other_args_to_resolve` that passes deployment object into ClassNode

  #23243 closedJun 30, 2025

• [runtime env] Improve tracking of URI size

  #23186 closedJun 30, 2025

• [updater][Bug] update fails on preempted node and autoscaler stops scheduling

  #23182 closedJun 30, 2025

• Pipeline ingress requires trailing /

  #23048 closedJun 30, 2025

• Shouldn't require `PipelineInputNode` to build a pipeline DAG

  #23037 closedJun 30, 2025

• Pipeline DAG sanity check for model wrappers fields

  #23019 closedJun 30, 2025

• Pipeline doesn't accept importable class as arguments

  #23016 closedJun 30, 2025

• [tune][Bug] 'tune.report( mean_accuracy=sklearn.metrics.accuracy_score(test_y, pred_labels), done=True)'where can i get the mean_accuary result?

  #22992 closedJun 30, 2025

• [Train] add logging to `finish_training` for existing `Callback`s

  #22754 closedJun 30, 2025

• [Bug] [serve] Accessing shared objects within a deployment

  #22751 closedJun 30, 2025

• [Feature] Client version check on commit

  #22675 closedJun 30, 2025

• [Jobs] run all doc examples in CI

  #22487 closedJun 30, 2025

• Some tests misusing assertTrue for comparisons

  #22395 closedJun 30, 2025

• [Enhancement][client] Move synchronous GetObject calls to datapath

  #22357 closedJun 30, 2025

• Enhance state notification pattern in Ray pubsub

  #22340 closedJun 30, 2025

• [Core] Avoiding subscribing to all logs by each log subscriber

  #22274 closedJun 30, 2025

• [Train] TPU support

  #22251 closedJun 30, 2025

• [train] support per epoch shuffling with `prepare_dataloader`

  #22108 closedJun 30, 2025

• [runtime env] Refactor `pip` protobuf to store a single str (`requirements.txt` contents) instead of list of "packages"

  #22097 closedJun 30, 2025

• [runtime_env] Remove `.lock` files after URI garbage collection

  #22062 closedJun 30, 2025

• [runtime env] Use LRU cache for URIs instead of random eviction

  #22060 closedJun 30, 2025

• [runtime env] Use single URI for `py_modules` field

  #22059 closedJun 30, 2025

• [Train] Add callback preprocessor that smoothly tracks values

  #21989 closedJun 30, 2025

• [Bug] Policy - ActionDistribution Type

  #21973 closedJun 30, 2025

• [runtiime env] Use coroutine to create runtime envs in `runtime_env_agent`

  #21950 closedJun 30, 2025

• [Train] Add support for Bagua

  #21934 closedJun 30, 2025

• [Bug] "The kernel has died..." during Ray tune.run

  #21917 closedJun 30, 2025

• [Jobs] Backwards compatibility tests for REST API

  #21915 closedJun 30, 2025

• [Jobs] Make jobs work out-of-the-box with cluster YAML

  #21911 closedJun 30, 2025

• [Train] Support for averaging results

  #21849 closedJun 30, 2025

• AttributeError raised when using response_model in FastAPI route decorator

  #21744 closedJun 30, 2025

• [Feature] [runtime env] [C++] support a strong-typed API in C++

  #21733 closedJun 30, 2025

• [runtime env] Cross-language runtime env

  #21731 closedJun 30, 2025

• [Testing] multi fake node set up doesn't work under non ray client mode

  #21653 closedJun 30, 2025

• [Bug] "Sent message larger than max" error with dask

  #21601 closedJun 30, 2025

• [runtime env] Can we avoid merging two runtime envs?

  #21494 closedJun 30, 2025

• [Tune] [Bug] Ray checkpoint sync can sometimes fail to upload checkpoints to s3, plus log spew about sync client observed

  #21469 closedJun 30, 2025

• [runtime env] raise exception for unsupported runtime_env features on Windows

  #21435 closedJun 30, 2025

• [train] fix scalability of `JsonLoggerCallback`

  #21416 closedJun 30, 2025

• [Feature] [runtime env] [java] select jdk version

  #21239 closedJun 30, 2025

• [Feature][Tune] Trial status based Stopper

  #21222 closedJun 30, 2025

• [Train][Tune] Unify Train and Tune Callbacks

  #21065 closedJun 30, 2025

• [Bug] rsync_filter isn't used in hash_runtime_conf

  #20878 closedJun 30, 2025

• [autoscaler] Persistent problems encountered during autoscaling can lead to driver log spam

  #20855 closedJun 30, 2025

• [Feature] Autoscaler should understand AWS availability and act accordingly

  #20774 closedJun 30, 2025

• [Bug] Test placement group chaos testing

  #20716 closedJun 30, 2025

• [GCP][autoscaler] Scale down is slow and Ray status doesn't show pending nodes

  #20695 closedJun 30, 2025

• Support snappy compression for spilled objects

  #20575 closedJun 30, 2025

• Sparse object reads - read part of an object, without downloading the entire object

  #20500 closedJun 30, 2025

• [core] Scale shuffle to 200+ nodes

  #20499 closedJun 30, 2025

• Memory-aware task scheduling to avoid OOMs under memory pressure

  #20495 closedJun 30, 2025

• [Feature] [Placement Group] Add timeout mechanism when scheduling placement group

  #20477 closedJun 30, 2025

• [job submission] Add RAY_ADDRESS or --address to suggested commands for logs/status

  #20441 closedJun 30, 2025

• [Bug] [Ray Autoscaler] [Core] Ray Worker Node Relaunching during 'ray up'

  #20402 closedJun 30, 2025

• [workflow] Fail to construct workflow within a workflow

  #20381 closedJun 30, 2025

• [Feature] [Serve] Threading for Ray Serve

  #20169 closedJun 30, 2025

• [Feature] [Serve] Support Sticky Sessions for Stateful Workflows Deployed via Ray Serve

  #20107 closedJun 30, 2025

• [runtime env] Remove filelock dependency

  #20083 closedJun 30, 2025

• [Bug] Potential deadlock in task scheduling algorithm for placement group resources.

  #20051 closedJun 30, 2025

• [Feature] rllib + tune metric logging selection

  #19816 closedJun 30, 2025

• [RLlib] [documentation] clarify postprocess_fn usage in our doc

  #19648 closedJun 30, 2025

• [Feature] [runtime env] Clean up the command arguments in raylet args

  #19448 closedJun 30, 2025

• [client] better error message when failing to connect with client

  #19371 closedJun 30, 2025

• [SGD] Document best practices for Pipeline epochs

  #19323 closedJun 30, 2025

• [workflow] scan_prefix with pages/as geneartor

  #19234 closedJun 30, 2025

• [Core][usability] Improve Ray cluster start up time

  #19215 closedJun 30, 2025

• [Serve] Don't use `ray.wait()` to drain tracking refs in handle

  #19158 closedJun 30, 2025

• Unify internal configs & common datastructures

  #19152 closedJun 30, 2025

• Clean up EndpointState

  #19148 closedJun 30, 2025

• [Core][Feature] use clang-tidy/format to block usage of std::getenv

  #18894 closedJun 30, 2025

• [Feature][workflow] Namespace for workflow

  #18818 closedJun 30, 2025

• [Feature][workflow] Resource limit for workflow job

  #18780 closedJun 30, 2025

• [Bug] Exception in task leads to truncated error message

  #18699 closedJun 30, 2025

• [Bug] Logging config is not propagated to driver

  #18660 closedJun 30, 2025

• Enable copy/paste to get correct command for connecting to Ray client

  #18513 closedJun 30, 2025

• Ray client suppresses error messages

  #18512 closedJun 30, 2025

• [serve] Feature request: timeout or max_retries to limit the time spent waiting for a deployment to complete

  #18432 closedJun 30, 2025

• Add workflow.current_step_uuid() function

  #18356 closedJun 30, 2025

• [Shuffle] non streaming shuffle 5000 partitions seem to reach the scalability limit

  #18333 closedJun 30, 2025

• [tune] atari-impala-large.yaml does not finish gracefully

  #18325 closedJun 30, 2025

• [runtime env] eagerly install for task/actor level

  #18160 closedJun 30, 2025

• [C++ API] Support cross-lang API with Python/Java

  #18149 closedJun 30, 2025

• [helm][kubernetes][test] Add formatting tests for Helm chart

  #18125 closedJun 30, 2025

• [workflows] Better message when not init'ed

  #18121 closedJun 30, 2025

• resource config is not respected in head_start_ray_commands in cluster.yaml

  #18097 closedJun 30, 2025

• [Dask-on-Ray] Propagate Dask-on-Ray scheduler config to (rest of) cluster

  #17943 closedJun 30, 2025

• [core] PlacementGroup should be no op for local_mode=True

  #17937 closedJun 30, 2025

• Enhance document on Java API

  #17820 closedJun 30, 2025

• [Object Spilling] Remove the spilled directory upon Sigterm for ray start

  #17790 closedJun 30, 2025

• [C++ API] Support non-global named actor

  #17734 closedJun 30, 2025

• Cleanup stats/metrics.h

  #17679 closedJun 30, 2025

• workflow cli to manage all jobs

  #17672 closedJun 30, 2025

• [docs] Tutorial on Pytorch Lightning needs rearranging

  #17611 closedJun 30, 2025

• [Serve] Helper functions that are written below the actor class don't work

  #17590 closedJun 30, 2025

• Fix circular dependence in workflow's code

  #17445 closedJun 30, 2025

• [lineage] Support lineage reconstruction for borrowed ObjectRefs

  #17380 closedJun 30, 2025

• Errors during scaling cluster

  #17292 closedJun 30, 2025

• Trial is being repeated with the exact same results

  #17257 closedJun 30, 2025

• [RFC][Placement groups] Allow tasks to acquire resources in addition to placement group bundle

  #17229 closedJun 30, 2025

• runtime env in workflow

  #16992 closedJun 30, 2025

• [autoscaler][core] Safe node termination

  #16975 closedJun 30, 2025

• [Ray Client] [Usability] Help users spot bandwidth bounded workload

  #16966 closedJun 30, 2025

• [docker][Clusters][autoscaler][local] Can't connect to cluster when using docker with ray cluster launcher

  #16961 closedJun 30, 2025

• [cli] Support redis password for all ray commands

  #16921 closedJun 30, 2025

• [Core] [runtime env] Use portable hash function for runtime_env_hash

  #16821 closedJun 30, 2025

• [runtime env] Support rescheduling tasks when runtime env creation failed.

  #16800 closedJun 30, 2025

• Priority scheduling of jobs

  #16782 closedJun 30, 2025

• [C++ API] Completed object reference counting support

  #16702 closedJun 30, 2025

• [Core] Programmatic way to access pending tasks for an actor?

  #16641 closedJun 30, 2025

• [Core] Erroneous check for size_t underflow

  #16626 closedJun 30, 2025

• [Core] Standardize Timestamps across codebase

  #16510 closedJun 30, 2025

• [test][MLDataset] Fix test_from_modin

  #16357 closedJun 30, 2025

• Example for tuning layer count, dropout probabilities with Transformers

  #16340 closedJun 30, 2025

• Ray started in local mode doesn't restore environment variables after shutdown

  #16132 closedJun 30, 2025

• [core] ray.remote hides the docstring of the decorated class

  #15877 closedJun 30, 2025

• [autoscaler] support rsync option `--include`

  #15859 closedJun 30, 2025

• [Placement Group] The bundle_reservation_check_func breaks load code from local

  #15840 closedJun 30, 2025

• Contributor docs don't mention running tests via bazel

  #15833 closedJun 30, 2025

• [docs] should actor methods always have num_returns value?

  #15818 closedJun 30, 2025

• [rfc] Support `ray[aws,gcp,azure]` as an install target

  #15725 closedJun 30, 2025

• [rllib] Error while using "count_steps_by": "agent_steps" and misleading documentation

  #15708 closedJun 30, 2025

• Ray duplicate data from GPU to CPU when placing an actor on GPU

  #15692 closedJun 30, 2025

• [kubernetes] ModuleNotFoundError when executing a task on a remote cluster

  #15668 closedJun 30, 2025

• [cross_language] Support Python dictionaries

  #15569 closedJun 30, 2025

• [core] detached actor logs are not streamed to successive clients

  #15549 closedJun 30, 2025

• Serve Deployment with Reload Option

  #15505 closedJun 30, 2025

• [client][core] Have Unified `register_serializer` interface

  #15486 closedJun 30, 2025

• [Job submission] Monitor driver

  #15480 closedJun 30, 2025

• [Job submission] Java support

  #15479 closedJun 30, 2025

• [Job submission] Basic drop job feature

  #15478 closedJun 30, 2025

• Ray memory size and object store size not correct on k8s

  #15463 closedJun 30, 2025

• Ray status not report correctly after node crashed

  #15459 closedJun 30, 2025

• Async actor method hang

  #15437 closedJun 30, 2025

• [client] python packages version mismatch fail silently

  #15407 closedJun 30, 2025

• [cluster] Make node_ip_address work throughout

  #15239 closedJun 30, 2025

• [autoscaler][docs] Explain how the `ray_bootstrap_config` is generated

  #15232 closedJun 30, 2025

• [autoscaler] Don't autofill `setup_commands` if head/worker `setup_commands` are used

  #15231 closedJun 30, 2025

• [Core] Add gRPC streaming support.

  #15219 closedJun 30, 2025

• Optimise for num_workers stucks in the infinite loop

  #15168 closedJun 30, 2025

• Ray dies without a proper error message - "Killed", might have to do with pandas

  #15165 closedJun 30, 2025

• [autoscaler] Simplify Custom ObjectStore Size

  #15147 closedJun 30, 2025

• [Core] Periodical runner can cause heap-use-after-free

  #15141 closedJun 30, 2025

• Metric tag keys type inference (Tuple To String)

  #15130 closedJun 30, 2025

• Actor task hangs after actor crashes with max_task_retries=0

  #15045 closedJun 30, 2025

• AlphaZero torch model doesn't support cuda, only cpu

  #14970 closedJun 30, 2025

• [Autoscaler] AWS setup commands hardcodes pip

  #14963 closedJun 30, 2025

• Support "dry runs" for deploy() operations

  #14936 closedJun 30, 2025

• [Object Spilling] Failing objects that fail to restore many times.

  #14921 closedJun 30, 2025

• num_cpus not handled correctly when function has a Queue argument

  #14863 closedJun 30, 2025

• Make rolling update batch size configurable

  #14853 closedJun 30, 2025

• Typed handle to deployments

  #14810 closedJun 30, 2025

• [Core] Docs - run data processing examples in CI

  #14769 closedJun 30, 2025

• [core] The remote function has been exported 100 times..

  #14730 closedJun 30, 2025

• Support `ray status CLUSTER.YAML`

  #14549 closedJun 30, 2025

• Support decoupling task/actor interfaces from implementation

  #14529 closedJun 30, 2025

• Support specifying container images in runtime_env

  #14528 closedJun 30, 2025

• [autoscaler] Error message not being cleared when autoscaler recovers

  #14494 closedJun 30, 2025

• [Docs] [tune] WanDB + Ray Integration a bit unclear from the docs

  #14478 closedJun 30, 2025

• [tune] TBXLoggerCallback not creating necessary directory

  #14437 closedJun 30, 2025

• [autoscaler][interface] Per-node-type docker configs

  #14418 closedJun 30, 2025

• [metrics] Add metrics for debugging Dask-on-Ray

  #14372 closedJun 30, 2025

• [metrics] Report metrics to be used for debugging load balancing issues

  #14369 closedJun 30, 2025

• [metrics] Remove unused or unnecessary metrics.

  #14366 closedJun 30, 2025

• When the node is crashed, logs are not accessible.

  #14307 closedJun 30, 2025

• [autoscaler] SSH command errors aren't written to monitor.out

  #14298 closedJun 30, 2025

• [dashboard] Add resource usage/availability to the dashboard

  #14292 closedJun 30, 2025

• [Core] Fix ray::Status <--> gRPC status interplay.

  #14278 closedJun 30, 2025

• updating worker nodes show as healthy

  #14232 closedJun 30, 2025

• [Object Spilling] Use subdirectories to avoid large top level inodes for file spilling

  #14166 closedJun 30, 2025

• [tune] Stack Traces with Function API are really hard to parse

  #14162 closedJun 30, 2025

• [Object Spilling] Plasma store probably doesn't respect the max shm size.

  #14145 closedJun 30, 2025

• Latent bugs in command_runner.py

  #14139 closedJun 30, 2025

• [rllib] undocumented behavior of timers/* in progress.csv

  #14052 closedJun 30, 2025

• Graceful Placement Group Removal

  #14045 closedJun 30, 2025

• Improve Docker manual setup document

  #14030 closedJun 30, 2025

• [UX] Allow passing CPU and GPU to actor and task resources.

  #13996 closedJun 30, 2025

• Remove cluster_synced_files and file_mounts_sync_continuously

  #13967 closedJun 30, 2025

• [Object Spilling] Allow to specify max_disk_usage for file system spilling.

  #13960 closedJun 30, 2025

• [Dashboard] add actor detail to experimental dashboard

  #13875 closedJun 30, 2025

• ray.put() slows down over time.

  #13612 closedJun 30, 2025

• [rllib]Action masking with tuple action space

  #13592 closedJun 30, 2025

• [dask-on-ray] Remove internal Dask API dependencies from the Dask-on-Ray scheduler.

  #13560 closedJun 30, 2025

• [core] GCS doesn't always cancel worker leases for killed actors

  #13545 closedJun 30, 2025

• test_autoscaling_policy.py prints out huge pile of JsonErrors

  #13433 closedJun 30, 2025

• Remove the RAY_CLIENT_MODE flag now that we don't need it

  #13279 closedJun 30, 2025

• [Core] Make CoreWorker more unit-testable

  #13268 closedJun 30, 2025

• Test S3 object spilling on multiple nodes with big data (streaming shuffle)

  #13222 closedJun 30, 2025

• [core] RAY_HOME path is hardcoded

  #13168 closedJun 30, 2025

• [Plasma Store]PlasmaClient::Get() return Status::OK() when timeout

  #12995 closedJun 30, 2025

• Add dashboard to bazel target to avoid running manual build commands

  #12956 closedJun 30, 2025

• Improve dashboard not found exception

  #12955 closedJun 30, 2025

• Cannot save training episodes: "TypeError: Object of type ndarray is not JSON serializable"

  #12951 closedJun 30, 2025

• [Object Spilling] Improve Read throughput

  #12950 closedJun 30, 2025

• Startup log use autoscaler_log.out / err instead of monitor.log

  #12884 closedJun 30, 2025

• [New scheduler] Don't assume 1-CPU tasks are feasible

  #12870 closedJun 30, 2025

• Turn on Test_reference_counting

  #12849 closedJun 30, 2025

• [Core] Locality-aware leasing: Milestone 3 - Spillback

  #12815 closedJun 30, 2025

• [Autoscaler] Refactor bin packing routines in autoscaler for code clarity

  #12723 closedJun 30, 2025

• [Core] Ray.get(timeout=0) doesn't work

  #12680 closedJun 30, 2025

• [core] Is starvation possible for multi-driver on the same cluster?

  #12667 closedJun 30, 2025

• GCS server ip error

  #12639 closedJun 30, 2025

• [core] Support detached/GCS owned objects

  #12635 closedJun 30, 2025

• [autoscaler] respect max_workers per node type when terminating nodes

  #12634 closedJun 30, 2025

• [Cluster launcher] Command runner logs are improperly quoted when logged

  #12631 closedJun 30, 2025

• permissions on rsync'd files are incorrect on worker nodes, results in inability to update workers

  #12630 closedJun 30, 2025

• [tune] Full experiment checkpointing doesn't work with PBT

  #12558 closedJun 30, 2025

• New workers are started slowly on a node if running workers >= `num_cpus`

  #12525 closedJun 30, 2025

• [tune] get_checkpoint_paths fails due to glob command for .tune_metadata file

  #12453 closedJun 30, 2025

• [New scheduler] Implement dynamic resources

  #12433 closedJun 30, 2025

• [metrics] Investigate tracing visualization tools

  #12314 closedJun 30, 2025

• [metrics] Utility to easily configure logging for a Ray job/actor/task

  #12306 closedJun 30, 2025

• `ray dashboard` throws bad exception

  #12246 closedJun 30, 2025

• [Object Spilling] Tune S3 performance + Add unit tests with moto3

  #12232 closedJun 30, 2025

• Duplicated IDs are generated

  #12197 closedJun 30, 2025

• [tune/logging] Warning for Tune

  #12140 closedJun 30, 2025

• [tune] Restarted Trials Use Incorrect Command When Multiple Commands Run on Cluster/Runtime

  #12048 closedJun 30, 2025

• [Object spilling] Move LocalObjectManager into the plasma store

  #12042 closedJun 30, 2025

• [Object spilling] Improve OutOfMemory handling through better memory bookkeeping in plasma store

  #12040 closedJun 30, 2025

• [Object Spilling] Use compression to reduce IO cost.

  #11992 closedJun 30, 2025

• [Tune] Add more custom Error Types

  #11871 closedJun 30, 2025

• Tune report histograms

  #11797 closedJun 30, 2025

• [Feature Request] [Tune] Add a special 'evaluation-step' flag to avoid unnecessary lengthy evaluations

  #11725 closedJun 30, 2025

• Unable to create ActorHandle for already created inherited classes object list [java][ray]

  #11715 closedJun 30, 2025

• Socket connections from GCS stuck in TIME_WAIT after actor death

  #11713 closedJun 30, 2025

• [docs] tutorial for autoscaling (really basic version)

  #11680 closedJun 30, 2025

• [flaky] test_multi_node/2 is flaky

  #11663 closedJun 30, 2025

• [flaky] test_object_manager is flaky

  #11661 closedJun 30, 2025

• [Core] Reduce the Redis connection per worker.

  #11655 closedJun 30, 2025

• [flaky] gcs_server test is flaky

  #11640 closedJun 30, 2025

• Use Pathlib instead of strings in Autoscaler

  #11633 closedJun 30, 2025

• [tune] PopulationBasedTraining and Tensorboard HPARAMS

  #11612 closedJun 30, 2025

• AWS Security group rule issue

  #11601 closedJun 30, 2025

• [dask] Parquet write fails if directory does not exist in advance

  #11566 closedJun 30, 2025

• [dask] Object store fills up too quickly in simple processing script

  #11565 closedJun 30, 2025

• [dask/tune] Provide an example of using Dask on Ray with Tune

  #11564 closedJun 30, 2025

• [tune] tutorial should indicate specific library version that we've tested against.

  #11540 closedJun 30, 2025

• [Core] Raylet can schedule tasks from a dead driver.

  #11520 closedJun 30, 2025

• `ray stop` should not kill all redis-server processes

  #11513 closedJun 30, 2025

• [core] Track the number of connection and use shared pool whenever possible for grpc clients.

  #11445 closedJun 30, 2025

• ray commandline tools raise exceptions if you forget the YAML config file

  #11396 closedJun 30, 2025

• [Autoscaler] Placement group rescheduling over-allocates resources

  #11372 closedJun 30, 2025

• [autoscaler] request_resources with partial instance availability leads to workers never shutting down

  #11367 closedJun 30, 2025

• how to add two-timescales Learning rate schedule in coustom policy?

  #11328 closedJun 30, 2025

• [Autoscaler] Add additional gpu types to util.accelerators

  #11160 closedJun 30, 2025

• [autoscaler] Worker node container is not removed after ray down?

  #11098 closedJun 30, 2025

• `ray stop` should wait for processes to exit

  #10955 closedJun 30, 2025

• [autoscaler] node type preferences

  #10929 closedJun 30, 2025

• Private/onprem clusters always need explicit ssh_private_key in docker

  #10838 closedJun 30, 2025

• [docs] Add examples for using custom resources

  #10808 closedJun 30, 2025

• Autoscaler should set RAY_ADDRESS environment variable

  #10752 closedJun 30, 2025

• Stop using `file_mounts` for ray_bootstrap_config & ray_bootstrap_key

  #10743 closedJun 30, 2025

• [Java] Remove Java 9/10/11 warnings

  #10673 closedJun 30, 2025

• [Documentation] need for default_resource_requests when using custom train function

  #10572 closedJun 30, 2025

• [rllib] action from policy with Tuple action space has wrong shape

  #10516 closedJun 30, 2025

• [tune] String summarization/representations for user objects

  #10489 closedJun 30, 2025

• [tune] Add regression test for avoiding extraneous output

  #10485 closedJun 30, 2025

• [GCS]Remove tightly coupled Redis code path from Python

  #10359 closedJun 30, 2025

• [GCS]Support Sharding GCS server

  #10358 closedJun 30, 2025

• [GCS]Support Multi-threaded GCS server.

  #10357 closedJun 30, 2025

• [GCS]Support different backend for GCS instead of Redis

  #10356 closedJun 30, 2025

• [metrics] Better way of grouping metric definitions

  #10341 closedJun 30, 2025

• [Core] WorkerThreadContext semantics are incorrect for async Python actors.

  #10324 closedJun 30, 2025

• [ray] Programatically expose the amount of memory available in the object store

  #10278 closedJun 30, 2025

• [tune] Improve the serialization diagnoser by providing deeper introspection

  #10263 closedJun 30, 2025

• [tune] Usability issues

  #10248 closedJun 30, 2025

• [ray] Support mypy

  #10244 closedJun 30, 2025

• Removed the following hyperparameter values when logging to tensorboard: ... [tune]

  #10166 closedJun 30, 2025

• [dask-on-ray] ValueError on read-only memory

  #10124 closedJun 30, 2025

• [cli/docs] Provide example commands in the CLI docstrings.

  #10079 closedJun 30, 2025

• [Placement Group] Placement group dashboard

  #9775 closedJun 30, 2025

• Ray issue with serializing pytorch objects only when running on 40+ cores

  #9752 closedJun 30, 2025

• Ray typing IDE code completion support

  #9623 closedJun 30, 2025

• [Cluster][Task Schedule] Remote function is not executing without any errors

  #9598 closedJun 30, 2025

• [core] RayConfig does not get set properly after multiple `ray.init` calls

  #9545 closedJun 30, 2025

• [New scheduler] Performance optimization

  #9487 closedJun 30, 2025

• [New scheduler] Release testing

  #9486 closedJun 30, 2025

• Specify network interface to use / RuntimeError: Redis has started but no raylets have registered yet.

  #9456 closedJun 30, 2025

• [autoscaler] Autoscaler prints (harmless) errors every 30 mins and then kills the workers in GCP cluster

  #9368 closedJun 30, 2025

• [Core] Core Worker Actor Handle GC.

  #9342 closedJun 30, 2025

• Graph related applications

  #9324 closedJun 30, 2025

• Options Support for Actor Methods

  #9296 closedJun 30, 2025

• [ray] constant memory usage increase of actor using actor handle.

  #9232 closedJun 30, 2025

• Invalid memory access in RedisAsioClient/RedisAsyncContext on shutdown

  #9074 closedJun 30, 2025

• Performance issue with many large tasks on 10 node cluster.

  #8950 closedJun 30, 2025

• Ray Dashboard Head-node CLI [autoscaler]

  #8450 closedJun 30, 2025

• Support TPUs across all of Ray

  #8260 closedJun 30, 2025

• [core] ray.init does not work if run in a node with external ip while the cluster is started internally

  #8244 closedJun 30, 2025

• [docs][autoscaler] additional dependencies needs to be mentioned to build your own autoscaler image

  #8235 closedJun 30, 2025

• [autoscaler] Api instead of CLI to interact with cluster.

  #8036 closedJun 30, 2025

• Incorrect unreconstructable error message and raise different exception.

  #7804 closedJun 30, 2025

• ray.wait hangs with no warning or error when local object store is too small to receive object

  #7802 closedJun 30, 2025

• Segmentation Fault when using multiprocessing.Queue

  #7793 closedJun 30, 2025

• ray.wait with local_mode=True blocks for a very long time

  #7741 closedJun 30, 2025

• Awesome: algorithm selection helper & diagrams

  #7722 closedJun 30, 2025

• Ray hangs when machine is disconnected from network

  #7696 closedJun 30, 2025

• [docs] Clarify that in K8s the jobs need to be launched from the workers

  #7188 closedJun 30, 2025

• [ray] tasks running in docker containers are not stopped on local cluster

  #6898 closedJun 30, 2025

• [dist] Release notes for Java And other Languages

  #6608 closedJun 30, 2025

• Ray.wait causes node to hang if there are too many object ids

  #6403 closedJun 30, 2025

• Performance issues with defining remote functions and actor classes from within tasks.

  #6240 closedJun 30, 2025

• TypeError: can't pickle CudnnModule objects

  #5947 closedJun 30, 2025

• Profiling ray tasks includes ray initialization time

  #5832 closedJun 30, 2025

• Make it possible to see resource deadlocks through web UI.

  #5789 closedJun 30, 2025

• [autoscaler] Raise better error message if `ssh_user` is not correct

  #5772 closedJun 30, 2025

• Code coverage tracker

  #5473 closedJun 30, 2025

• [ray] ray misuse gpu in docker container

  #5245 closedJun 30, 2025

• On a background thread, `ray.wait` doesn't timeout until another method on the actor is called

  #4934 closedJun 30, 2025

• Ray is not propagating variable types correctly

  #4463 closedJun 30, 2025

• [tune] Support nesting grid_search in lambdas

  #3466 closedJun 30, 2025

• Retry policy when a worker crashes: a hook missing?

  #2635 closedJun 30, 2025

• Task introspection

  #2617 closedJun 30, 2025

• ray start does not restart failed processes

  #2587 closedJun 30, 2025

• [rllib] flattening error in gym.spaces.Sequence

  #45563 closedJun 30, 2025

• cannot import name 'EPISODE_RETURN_MEAN' from 'ray.rllib.utils.metrics'

  #45453 closedJun 30, 2025

• error: No such option: --torch

  #45452 closedJun 30, 2025

• [Core] Unable to run worker with virtual environment without installing dashboard

  #45410 closedJun 30, 2025

• [RLlib] How to support gymnasium graph obs space?

  #45290 closedJun 30, 2025

• Ray Cluster does not work across multiple docker containers

  #45252 closedJun 30, 2025

• [Core] Worker crashes unexpectedly due to frequent triggering of OOM

  #45244 closedJun 30, 2025

• Ray Cluster: Failed to create a ray cluster using running container

  #45148 closedJun 30, 2025

• [Rllib] Rllib provides wrong state batch size during "bug check" batches on torch custom model

  #45131 closedJun 30, 2025

• [RLlib] ValueError in initialization of ImpalaTF2Policy

  #45050 closedJun 30, 2025

• [core] GcsSubscriber hangs in shutdown if the connection broke on MacOS

  #45044 closedJun 30, 2025

• Workflow: Reading workflow status can lead to corrupted json reads.

  #45027 closedJun 30, 2025

• [Core] `ray.wait` not actually wait until ready when the task is longer than 12 days

  #44909 closedJun 30, 2025

• [Data] Add `delete_dir_contents` parameter to `FileDatasink`

  #44794 closedJun 30, 2025

• [RLlib] PPO and framework=tf / issue with latest tensorflow 2.16.1

  #44675 closedJun 30, 2025

• [RLlib] PPO reset_config() AttributeError: 'dict' object has no attribute '_enable_new_api_stack'

  #44506 closedJun 30, 2025

• [RLlib] ReplayBuffer doesnt work with zero_init_states False when store rnn sequence

  #44383 closedJun 30, 2025

• [Cluster, YARN with Skein] Ray cluster keeps crashing when running on YARN via Skein

  #44112 closedJun 30, 2025

• [Ray Core] Ray nightly GPU docker image broken on NVIDIA V100 GPUs on AWS

  #43565 closedJun 30, 2025

• Using RNN for RL

  #43420 closedJun 30, 2025

• from ray.rllib.agents.registry import get_trainer_class ModuleNotFoundError: No module named 'ray.rllib.agents'

  #43310 closedJun 30, 2025

• WARNING deprecation.py:50 -- DeprecationWarning: `ray.rllib.execution.train_ops.multi_gpu_train_one_step` has been deprecated. This will raise an error in the future!

  #43250 closedJun 30, 2025

• [RLlib] Build_for_inference() in env_runner_v2.py created empty state_out_1 and lead to failure of initiation

  #42978 closedJun 30, 2025

• Core: ray.remote raises ValueError when used on torch IterableDataset

  #42914 closedJun 30, 2025

• Core: Join zombie subprocesses after task completion

  #42913 closedJun 30, 2025

• [Core] SIGSEGV when running Ray

  #42868 closedJun 30, 2025

• [Core] Serialisation does not work with classes with `__init_subclass__`

  #42823 closedJun 30, 2025

• Problem with YOLOv8 Hyperparameters tuning

  #42770 closedJun 30, 2025

• [RLLIB] Passing configuration to Custom Environment in rllib is giving an error

  #42753 closedJun 30, 2025

• [RLlib] Algorithms ES, A3C are deprecated and replacement does not exist in python package

  #42579 closedJun 30, 2025

• [<Ray component: Core|RLlib|etc...>] Inite state of attention_net.py is empty

  #42569 closedJun 30, 2025

• [<Ray component: Core|RLlib|etc...>] KeyError with RNN

  #42501 closedJun 30, 2025

• [RLlib] gpu cannot enable

  #42388 closedJun 30, 2025

• [<Ray component: Core|RLlib|etc...>] reslink in model

  #42333 closedJun 30, 2025

• [RLlib] shape [] in Box action space not supported.

  #42199 closedJun 30, 2025

• Building an executable using Ray and Cx_freeze

  #42101 closedJun 30, 2025

• RichProgressBar in PyTorch Lightning only show progress at the very end

  #42091 closedJun 30, 2025

• [<Ray component: Core|RLlib|etc...>] Channel errore

  #42089 closedJun 30, 2025

• [RLlib] When setting `config.environment(normalize_actions=False)` and using CQL, it raises an error: `AttributeError: 'TorchDiagGaussian' object has no attribute 'sample_logp'`.

  #42064 closedJun 30, 2025

• [Workflow] get_metadata() returns RUNNING instead of RESUMABLE status

  #41980 closedJun 30, 2025

• Ray IDs vs endianness?

  #41961 closedJun 30, 2025

• "RaySystemError: System error: Unknown error"

  #41786 closedJun 30, 2025

• [RLlib] Value error while running DQN

  #41559 closedJun 30, 2025

• Is there an error in the interface name " cdef cppclass CFunctionDescriptorInterface "ray::CFunctionDescriptorInterface "?

  #41553 closedJun 30, 2025

• Saving XGBoost model with json extension

  #41374 closedJun 30, 2025

• [data] date32 and datetime64 handling should be the same

  #41358 closedJun 30, 2025

• [RLlib] User guides are not ordered

  #41340 closedJun 30, 2025

• error installing library

  #41223 closedJun 30, 2025

• [core][state][dashboard] Better tasks info GC control at GCS

  #41142 closedJun 30, 2025

• [RLLib] External simulator: mean episode reward is NaN due to done not set

  #40954 closedJun 30, 2025

• [Core] - Cannot install in tiny core linux

  #40832 closedJun 30, 2025

• [Tune|RLlib] Add error-tolerant version of PB2

  #40787 closedJun 30, 2025

• NCCL Proxy Call to rank 1 failed - on Cloud VM Docker setup for huggingface distributed ray train script

  #40758 closedJun 30, 2025

• [Ray Train] - Add Options to Save Last checkpoint in Ray Train Checkpointing Config

  #40503 closedJun 30, 2025

• ray.init() can sometimes hang with a limited range specified for --worker-port-list

  #40497 closedJun 30, 2025

• [Core] Dead session not closed

  #40482 closedJun 30, 2025

• [RLlib][MBMPO] The algorithm does not learn as intended.

  #40400 closedJun 30, 2025

• [Tune] Support for new algorithm: Cost-Aware Pareto Region Bayesian Search (CARBS).

  #40356 closedJun 30, 2025

• [Workflow] Incorrectly set max_calls in options

  #40252 closedJun 30, 2025

• [PPOConfig] Utilising new API/models without matching documentation

  #40201 closedJun 30, 2025

• [Rllib] Tune locks up when attempting to create an rllib algorithm in a trainable

  #40015 closedJun 30, 2025

• [Tune/Air] Memory Leak when using WandbLoggerCallback with Population Based Tuning

  #40014 closedJun 30, 2025

• [RLlib] TD3/DDPG doesn't seem to respect action space bounds (at least initially)?

  #40002 closedJun 30, 2025

• [RLLIB] Issue with AlphaZero algorithm Stateless CartPole

  #39937 closedJun 30, 2025

• [RLLIB] Error in executing StatelessCartPole environment with AlphaZero

  #39862 closedJun 30, 2025

• Allow train_loop_config to be a dataclass / pydantic model

  #39824 closedJun 30, 2025

• [Core] ResolutionImpossible - Test requirements appear to not fit versions

  #39782 closedJun 30, 2025

• Job history is lost when Ray cluster is restarted (via kuberay)

  #39764 closedJun 30, 2025

• Ray::Tune::Logger::Tensorboardx

  #39741 closedJun 30, 2025

• [Core] Upgrading grpc to 1.57.0 causes perf regressions

  #39679 closedJun 30, 2025

• ray failed to register worker when I used vllm

  #39618 closedJun 30, 2025

• [rllib] Action space MultiDiscrete([11 5 1 2]) is not supported for DQN

  #39571 closedJun 30, 2025

• [RLlib] Support JAX-(numpy)-based envs.

  #39528 closedJun 30, 2025

• [RLlib] Ray RLLib Dependencies Version Information

  #39405 closedJun 30, 2025

• [RLlib] dreamerv3 causes debug code to be executed when running tune

  #39302 closedJun 30, 2025

• [Core] CPP Interface crashes on Ray.Init()

  #39252 closedJun 30, 2025

• ValueError: Must set agent_id on policy config

  #39246 closedJun 30, 2025

• [Core] Actor retry count is consumed because the task is retried when actor is still alive.

  #39110 closedJun 30, 2025

• [Core] Memory Leak

  #38877 closedJun 30, 2025

• [Tune] Leaky core concepts in Ray Tune documentation

  #38781 closedJun 30, 2025

• latest ray microbenchmark fails

  #38758 closedJun 30, 2025

• Ray Memory Usage Keeps Increasing even after Manual Garbage Collection

  #38730 closedJun 30, 2025

• [docs] Document Tune/Train placement group

  #38706 closedJun 30, 2025

• ray/RLlib/offline/estimators

  #38357 closedJun 30, 2025

• [Core] gcs_server Failed accept4: Too many open files

  #38248 closedJun 30, 2025

• [Core] Segfault from fibers when using streaming/dynamic generator (only happening from test_streaming_generator_exception)

  #38167 closedJun 30, 2025

• [RLlib] RL module and PPO implementation

  #38012 closedJun 30, 2025

• [RLlib] ray 2.6 relies on tf.bool which does not exist in tensorflow 2.13

  #37895 closedJun 30, 2025

• [RLlib] Sampler takes first step before next batch is requested

  #37893 closedJun 30, 2025

• [Data] Ray 2.6 created a breaking change in the index of a Modin DataFrame

  #37771 closedJun 30, 2025

• [Ray-Java client] Call actor report 'No module named' with py script

  #37600 closedJun 30, 2025

• [Core] Ray cpp example, if not call ray::Shutdown when exit, will cause segment fault.

  #37596 closedJun 30, 2025

• RLLib: Training Rllib-DDPG with custom environment leads error in Inference.

  #37242 closedJun 30, 2025

• [<Ray component: autoscaler>] _load_kubernetes_defaults_config function is not yet made

  #37033 closedJun 30, 2025

• [Core] Activate Ray tracing casue error when calling actor method with decorator with wraps. (i propose the possible solution)

  #36891 closedJun 30, 2025

• [Core] No dependency on setuptools results in broken build

  #36742 closedJun 30, 2025

• [data] Report actual task time and object sizes in Dataset.stats()

  #36671 closedJun 30, 2025

• [CI][Docs] Example in Train FAQ is flakey

  #36399 closedJun 30, 2025

• Be consistent on whether or not you include a dot at the end of a bullet list element.

  #36308 closedJun 30, 2025

• [Core] ray.put and ray.get extremely slow with polars frames

  #36068 closedJun 30, 2025

• [Core] DecodeError when `ray.put` a large (2GB) object

  #35976 closedJun 30, 2025

• [Ray Core] There is a Exception error message bug which convert byte array to String.

  #35880 closedJun 30, 2025

• System error: Ray has not been started yet. You can start Ray with 'ray.init()'

  #35592 closedJun 30, 2025

• [Client] Dataset write_csv AttributeError: ‘Worker’ object has no attribute 'core_worker'

  #35537 closedJun 30, 2025

• Ray: Data - Cannot read json its written to s3

  #35501 closedJun 30, 2025

• [Core] `OwnerDiedError` if dataset owner actor handle get out of scope

  #35262 closedJun 30, 2025

• [VM launcher] Automtically shut down the ec2 machine when I stop ray up in the middle

  #35013 closedJun 30, 2025

• [Core] Incorrect detection of cpus

  #34846 closedJun 30, 2025

• [Clusters] - Cannot switch off rsync during Cluster Launch with `ray up`

  #34390 closedJun 30, 2025

• Azure autoscaler cannot create additional nodes

  #34198 closedJun 30, 2025

• [Core] Error in external storage writing for object spilling

  #33913 closedJun 30, 2025

• get_node_to_storage_syncer has an empty docstring

  #33841 closedJun 30, 2025

• [ Core ] Correct usage of min/max-worker-port arguments

  #33749 closedJun 30, 2025

• Core: nightly builds for macos only include an x86 _raylet.so even though they claim to be universal

  #33720 closedJun 30, 2025

• [Core] The resources have minus values in ray status output

  #33569 closedJun 30, 2025

• [tune] tqdm/Hyperopt-style TuneReporter for Databricks notebooks

  #33519 closedJun 30, 2025

• [Core] Ray client doesn't support `should_capture_child_tasks_in_placement_group` API

  #33513 closedJun 30, 2025

• [RLlib] DictFlatteningPreprocessor order is inconsistent leads to invalid mapping of OBS

  #33327 closedJun 30, 2025

• [runtime env] Raise warning when using `runtime_env` with `local_mode=True`

  #33260 closedJun 30, 2025

• [Core] `get_runtime_context()` in task fails with unhelpful error "cannot pickle '_thread.lock' object"

  #32987 closedJun 30, 2025

• [Train] Benchmark testing on Mosaic Composer with Ray

  #32946 closedJun 30, 2025

• [Ray Core] Actor Handles not properly passed to Actors created by other Actors

  #32848 closedJun 30, 2025

• [RLlib] A3C has problems with the horizon option removed

  #32812 closedJun 30, 2025

• [Core][Object Store] Object Store to manage files in the cluster

  #32694 closedJun 30, 2025

• [Clusters] [KubeRay] problem with pending actors' pods in Kubernetes

  #32651 closedJun 30, 2025

• [core] Lock contention when submitting actor task on the client queue

  #32595 closedJun 30, 2025

• [Core] Install via `pip` fails, install with `conda` crashes worker and exits

  #32423 closedJun 30, 2025

• [Core] "ImportError: No module named ray" when using `ray submit`

  #31924 closedJun 30, 2025

• [workflow] memory leakage

  #31819 closedJun 30, 2025

• [Clusters] [RLlib] Trainer Object running on Worker node & RolloutWorker running on Head node

  #31808 closedJun 30, 2025

• [runtime envs] Ray Client Server failed when starting

  #31622 closedJun 30, 2025

• Huge numbers of "deleted" files with open processes left after Ray Tune run

  #31556 closedJun 30, 2025

• [Tune] Reenable `zoopt` searcher test after fixes for handling invalid results are included in its next release

  #31439 closedJun 30, 2025

• [RLlib] Pytorch multiple optimizers

  #31428 closedJun 30, 2025

• In the docker bridge mode, pulling the actor on a non head node fails.

  #31308 closedJun 30, 2025

• [Dashboard] Head node exited unexceptly because of dashboard process exited

  #31261 closedJun 30, 2025

• [CORE] Unable to run celery task containing ray tasks

  #31157 closedJun 30, 2025

• [core] Segfaults when restarting Ray multiple times in unit tests with background threads running

  #31145 closedJun 30, 2025

• [core] Error with Slurm: No available node types can fulfill resource request {'node:<ip>': 0.01}.

  #31135 closedJun 30, 2025

• [RLlib] Not able to save evaluation recording videos

  #30949 closedJun 30, 2025

• [Ray Job] SchedulingCancelled for JobSupervisor Actor

  #30898 closedJun 30, 2025

• [Ray client] Ray Zombie Process Issue

  #30894 closedJun 30, 2025

• [Devprod] Bazel reports an error when compiling as a non-root user

  #30885 closedJun 30, 2025

• [release tests] Prometheus metrics collection sometimes takes 15min to run for long_running_node_failures

  #30859 closedJun 30, 2025

• [core] Disk full error logging is verbose

  #30833 closedJun 30, 2025

• [RLlib] Error when running RLlib

  #30412 closedJun 30, 2025

• [Cluster Launcher] `ray dashboard` CLI command does not stop port-forwarding after Ctrl+C

  #30385 closedJun 30, 2025

• [autoscaler] AWS Single Sign-On support

  #30064 closedJun 30, 2025

• [Autoscaler][GCP] Autoscaler crashing on GCP with error 404.

  #30050 closedJun 30, 2025

• Setting some system configs causes Ray to fail to start

  #29841 closedJun 30, 2025

• [Ray Cluster] Assigning all host GPUs into head node without nvidia.com/gpu present

  #29753 closedJun 30, 2025

• [gcp] "No such container" error after ray up

  #29671 closedJun 30, 2025

• [Tune] Passing a handle to grid search cause trials to get stuck in running and pending mode

  #29545 closedJun 30, 2025

• [Serve] `ServeHandles` fail if GCS crashes before first request

  #29539 closedJun 30, 2025

• [Core] util.multiprocessing.pool scheduling inefficiencies, blocking behavior in imap and imap_unordered

  #29453 closedJun 30, 2025

• [Core] inspect_serializability bug - parent object serializable but bound method not

  #29423 closedJun 30, 2025

• [Core] Ray doesn't shutdown properly on KeyboardInterrupt

  #29384 closedJun 30, 2025

• [Serve] Unable to upload current working directory

  #29354 closedJun 30, 2025

• [core][observability] Improving reliability of memory_summary API call

  #29329 closedJun 30, 2025

• InvalidLocationConstraint Message: The specified location-constraint is not valid for storage option

  #29309 closedJun 30, 2025

• [Core] Worker pool didn't prestart num_cpus workers

  #29162 closedJun 30, 2025

• [core] use proto for oom error / node died error in the frontend

  #28907 closedJun 30, 2025

• [ray client] surface ray client logs better

  #28890 closedJun 30, 2025

• [Backlog][Collective] Facilitate NCCL test in ray cluster

  #28860 closedJun 30, 2025

• [core/k8s/GKE] Ray schedules actors on pods/nodes that are shutting down

  #28852 closedJun 30, 2025

• [Core] Cannot 'ray list nodes' after setting the environmental variable 'export RAY_ADDRESS="http://127.0.0.1:8265" '

  #28847 closedJun 30, 2025

• [AIR] [Tune] Don't add random hash to trial id for single trial

  #28830 closedJun 30, 2025

• [core] Object returned by a generator with num_returns="dynamic" should throw an error if reconstruction fails

  #28688 closedJun 30, 2025

• [P0] test_submit_cpp_job failed in osx

  #28592 closedJun 30, 2025

• [Ray: Core] - Unable to enable TLS on the ray head node

  #28534 closedJun 30, 2025

• Dashboard / Jobs RegexMatcher ignores "includes".

  #28502 closedJun 30, 2025

• [Core, RLlib] RLlib uses Metal GPU even when told not to

  #28385 closedJun 30, 2025

• [Core] Actor methods will be modified for tracing even if tracing is not enabled.

  #28293 closedJun 30, 2025

• [Runtime] Improve runtime environment error message when virtualenv version is too old

  #28232 closedJun 30, 2025

• [Core] Multi-Threaded Actors are Un-Killable

  #28086 closedJun 30, 2025

• [Autoscaler] Assigning None to optional keys leads to failure

  #28012 closedJun 30, 2025

• [Core] Can't pickle objects defined in top-level environment

  #28000 closedJun 30, 2025

• [Doc] [Serve] Serve Loki monitoring tutorial screenshot has outdated API

  #27453 closedJun 30, 2025

• [core] Very slow task scheduling during Dataset.sort on 100TB

  #27410 closedJun 30, 2025

• Is Ray going to support Weighted Quantile Sketches or Quantile Sketches?

  #27363 closedJun 30, 2025

• [Core] Raylet continually exiting on worker in docker

  #26576 closedJun 30, 2025

• Tensorboard with Docker from Ray dashboard, tune tab cannot be accessed

  #26325 closedJun 30, 2025

• [RLlib] Eval episode runs forever if Env doesn't terminate properly

  #26241 closedJun 30, 2025

• [Ray Client] Using many concurrent client connections results in deadlock/hanging

  #26144 closedJun 30, 2025

• [Core] worker died randomly and unexpectedly under heavy workload (Check failed: inner_it->second.mutable_nested()->contained_in_borrowed_ids.erase(id))

  #26128 closedJun 30, 2025

• [Core][HA] Actor entries are not deleted from the storage permanently if GCS is crashed.

  #26114 closedJun 30, 2025

• Unclear error when using generator tasks

  #25836 closedJun 30, 2025

• [Core] SIGSEGV when I run experimental shuffle command.

  #25650 closedJun 30, 2025

• [Core][Metrics] Prometheus-client not working with the latest version.

  #25523 closedJun 30, 2025

• [core] Scheduler stalls during shuffle reduce stage with 100k concurrent tasks or more

  #25412 closedJun 30, 2025

• [AIR] Utilities to go from Predictor to `BatchPredictor` and `ModelWrapperDeployment`

  #24977 closedJun 30, 2025

• [Train/AIR] Ray Train actors still use up resources after Notebook cell is stopped

  #24947 closedJun 30, 2025

• [Core] Failed to delete named actor in client mode

  #24906 closedJun 30, 2025

• [AIR] Add a `reconfigure` option to `ModelWrapperDeployment`

  #24869 closedJun 30, 2025

• [core] Uninformative error for unserialisable objects

  #24863 closedJun 30, 2025

• [Serve] Prototype C++ Worker in Serve

  #24738 closedJun 30, 2025

• [Core] Spilling performance regression in large-scale shuffle

  #24667 closedJun 30, 2025

• [Core] Restore objects directly from S3

  #24581 closedJun 30, 2025

• [Ray component: Core] Dask on Ray - Worker processes go to idle state and not garbage collected when used with RayProgressBar()

  #24556 closedJun 30, 2025

• [runtime env] `serialized_env` used as ID, but identical envs can produce different `serialized_env`

  #24515 closedJun 30, 2025

• [Core] No overloads for "remote" match the provided arguments

  #24371 closedJun 30, 2025

• Workflows: Type stubs are incorrect: argument missing for parameter status_filter

  #24367 closedJun 30, 2025

• [Core] /api/cluster_status treats placement groups differently than ray status

  #24309 closedJun 30, 2025

• [Core] Restore worker silently fails and the program is stuck

  #24248 closedJun 30, 2025

• [RLlib][Bug] duplicate action unsquashing in DDPG / TD3 policy

  #24213 closedJun 30, 2025

• [Tune] support for FIRE PBT

  #24137 closedJun 30, 2025

• [Tune] Tune Job hangs out and can't finish the tune job

  #23858 closedJun 30, 2025

• [Workflows] Cant use custom storage backends

  #23831 closedJun 30, 2025

• [RLlib] Add Option for Custom Sample Preprocessing when Sampling from Replay Buffer

  #23815 closedJun 30, 2025

• [Core][Bug] global-scoped actor handles/Ray objects prevents Ray workers from being destructed.

  #23677 closedJun 30, 2025

• [runtime env] `zip_directory` `excludes` parameter doesn't work with absolute paths

  #23473 closedJun 30, 2025

• [Train] [Feature] Print useful traceback on SIGINT

  #23148 closedJun 30, 2025

• [Train] [Docs] Document how to change logging verbosity

  #23147 closedJun 30, 2025

• [docs][Bug] Workflow docs have few typos and type issue

  #23113 closedJun 30, 2025

• [tune][Feature] add tune.choices to select multiple values from a search space

  #23001 closedJun 30, 2025

• [Bug] An exception in a task cannot be caught with ActorPool.map_unordered making restarting meaningless

  #22978 closedJun 30, 2025

• Ray Train / Tune - W&B logger documentation

  #22881 closedJun 30, 2025

• [Train] update `logdir` relative path

  #22753 closedJun 30, 2025

• [Bug][placement groups] Actor scheduling does not respect placement_group=None

  #22742 closedJun 30, 2025

• [Cluster snapshot] [Bug] `runtime_env` fields in cluster snapshot are converted to camelcase when they should not be

  #22565 closedJun 30, 2025

• [Train] Add flags to disable creating log directories

  #22261 closedJun 30, 2025

• [RLlib] [Feature] Support for having parametric action spaces/action masking for continuous action space models

  #22259 closedJun 30, 2025

• [RLLib] Workers died at the initialization stage when the observation space is a 3D shape

  #22033 closedJun 30, 2025

• [Train] Automatically choose number of workers

  #21987 closedJun 30, 2025

• [Serve] The adjustment about Ray Serve Java Proxy and Java Replica

  #21694 closedJun 30, 2025

• [C++] Cluster Mode Tests Should have 1 test per feature tested

  #21454 closedJun 30, 2025

• [Tune] [Bug] lazily expand directories for client compatibility

  #21408 closedJun 30, 2025

• [Tune] Issue on page /tune/tutorials/tune-pytorch-lightning.html

  #21354 closedJun 30, 2025

• [Bug] Got stucked when running python script from a shell script

  #21298 closedJun 30, 2025

• [Bug] [Tune] pbt run_experiments not stable, some trial will error.

  #21259 closedJun 30, 2025

• [Train] Document Callbacks

  #21066 closedJun 30, 2025

• [Feature] Single source of truth for Ray version in Java `pom.xml` and `pom_template.xml` files

  #21059 closedJun 30, 2025

• [Test Bug] Matching `psutil.Process.name()` doesn't work on macOS

  #20982 closedJun 30, 2025

• [Bug] Incorrect promise usage that causes infinite blocking calls

  #20899 closedJun 30, 2025

• We encountered the cast exception after we got result from ray actor task

  #20369 closedJun 30, 2025

• [Train] Refactor `TrainingIterator` result processing logic

  #20330 closedJun 30, 2025

• [tsan] Add TSAN CI build that runs basic Python tests

  #20080 closedJun 30, 2025

• [tsan] Race in census SetGlobalTags

  #20079 closedJun 30, 2025

• [tsan] Race accessing global stats objects

  #20078 closedJun 30, 2025

• [tsan] Several global config variables accessed unsafely

  #20077 closedJun 30, 2025

• Support working_dir=None for skipping packaging upload/download

  #19962 closedJun 30, 2025

• [Bug] Placement group removal refinement

  #19937 closedJun 30, 2025

• [Feature] Able to access objects put in cross language

  #19873 closedJun 30, 2025

• [Bug] Improve RuntimeEnvSetupError message

  #19824 closedJun 30, 2025

• [RLlib] Deprecate Internally Maintained Probability Distributions In Favor Of Native TFP And torch.distributions Solutions

  #19725 closedJun 30, 2025

• [Serve] Test KVStore early in constructor init.

  #19714 closedJun 30, 2025

• [Bug] [Workflow] ray.wait on workflow result doesn't work as expected

  #19295 closedJun 30, 2025

• [tune] MLFlowLogger doesn't save artifacts for remote mlflow tracking_uri

  #19263 closedJun 30, 2025

• [Bug] [XLang] Segfault when Java returns void

  #18837 closedJun 30, 2025

• [Bug] tensorboardX vs tensorboard?

  #18727 closedJun 30, 2025

• Dashboard exposes redis PW on the command line

  #18491 closedJun 30, 2025

• Race condition of grpc backpressure

  #18439 closedJun 30, 2025

• [Core] Task spec including inlined objects can crash lease request RPCs.

  #18194 closedJun 30, 2025

• [Runtime Env] Setup process doesn't have CPU limit

  #18137 closedJun 30, 2025

• ray.init with address crashes process outside of cluster

  #17769 closedJun 30, 2025

• new dashboard agent port conflict issues

  #17498 closedJun 30, 2025

• [Core] Unable to get actor handle of global named actor created in java from python in Ray 1.4.0

  #16436 closedJun 30, 2025

• [serve] java api

  #16393 closedJun 30, 2025

• [serve] java serve handle

  #16392 closedJun 30, 2025

• [serve] java http proxy

  #16391 closedJun 30, 2025

• [Shuffle] non-streaming consumed bytes are too low compared to spilled / restored bytes.

  #16149 closedJun 30, 2025

• [ray] Multiple concurrent requests to create a named actor crash GCS

  #15941 closedJun 30, 2025

• Remove unused util functions for conda environments

  #15912 closedJun 30, 2025

• [core] Zero-gpu node shouldn't be marked with accelerator_type resource.

  #15878 closedJun 30, 2025

• Cannot using external model with cuda when using ray

  #15869 closedJun 30, 2025

• [wheel][doc] Make it easier to access Ray wheels for specific commits

  #15765 closedJun 30, 2025

• [rllib]Update the docs about Variable-length / Parametric Action Space

  #15710 closedJun 30, 2025

• Odd task scheduling behavior on same node

  #15602 closedJun 30, 2025

• Averaging learning curves over repetitions + plotting confidence intervals [Tune]

  #15400 closedJun 30, 2025

• AssertionError when using pyinstaller with ray

  #15396 closedJun 30, 2025

• [core] Memory leak when using local simulated cluster (long_running_tests/workloads/apex.py)

  #15305 closedJun 30, 2025

• [Core] Bad traceback on failure to reconnect to GCS server.

  #15235 closedJun 30, 2025

• [metrics] Custom sum metrics have type comment "gauge"

  #15150 closedJun 30, 2025

• [core] Actor restart does not work when owner dies and constructor task has dependencies

  #15076 closedJun 30, 2025

• [k8s] ray down command does not remove pods which are in evicted state

  #14958 closedJun 30, 2025

• [Tune] [Ray Client] tune_cifar10_gluon example fails with Ray Client

  #14946 closedJun 30, 2025

• [ray white paper] broken links

  #14897 closedJun 30, 2025

• Fix Asyncio Event Metrics on Java

  #14715 closedJun 30, 2025

• Add ray.__wheel__ with a link to the wheel to install the same version

  #14623 closedJun 30, 2025

• Pre-push hooks allow code to be pushed that fails LINT

  #14367 closedJun 30, 2025

• __del__ magic method can't access class properties

  #14285 closedJun 30, 2025

• Failed to load actor due to dependencies not being pickled

  #14284 closedJun 30, 2025

• optimization: Client blocks on releasing references due to detached actor race condition

  #14137 closedJun 30, 2025

• [autoscaler] request resources doesn't work with multiple jobs

  #13534 closedJun 30, 2025

• [Metrics] Custom metrics don't work after calling `ray.shutdown()` followed by `ray.init()`

  #13532 closedJun 30, 2025

• Unify linting of clang-format and *.proto files

  #13465 closedJun 30, 2025

• Hang or Deadlock when calling ray.get() inside pytorch Dataset when DataLoader with num_workers >0

  #13407 closedJun 30, 2025

• [core] Unwanted pickling behaviour when starting remote actor with @propery

  #13365 closedJun 30, 2025

• Explore Protos as the Ray Client pickle transport (instead of namedtuples)

  #13280 closedJun 30, 2025

• SIGKILL generates core dumps on some systems

  #13221 closedJun 30, 2025

• Object store thrashing if it runs ray.get in a non-main thread.

  #12906 closedJun 30, 2025

• Canonicalize the python lint options

  #12801 closedJun 30, 2025

• [autoscaler] refactor duplicate code for handling request_resources().

  #12699 closedJun 30, 2025

• [Dashboard]Profile Actor Button Not Working

  #12668 closedJun 30, 2025

• [core] bytearray is parsed as bytes in remote function

  #12648 closedJun 30, 2025

• ValueError: After taking into account object store and redis memory usage, the amount of memory on this node available for tasks and actors (-0.01 GB) is less than 0% of total. You can adjust these settings with ray.init(memory=<bytes>, object store memory=<bytes>

  #12561 closedJun 30, 2025

• Ray grinds to a halt if both PyTorch and TensorFlow are installed

  #12467 closedJun 30, 2025

• Ray does not handle MIG devices

  #12413 closedJun 30, 2025

• [tune] progress reporter should limit table to 80char

  #12374 closedJun 30, 2025

• [serve] Distributed Tracing Support in Serve

  #12320 closedJun 30, 2025

• [metrics] Replace ray timeline with distributed tracing

  #12315 closedJun 30, 2025

• [metrics] Support filtering logs streamed to driver by actor/task

  #12305 closedJun 30, 2025

• [serve] Support more expressive policies for choosing replicas

  #12296 closedJun 30, 2025

• [Tune] [PBT] Automatic experiment restart for synch=True

  #12122 closedJun 30, 2025

• [tune] [wandb] Experiment checkpointing fails with `WandbTrainableMixin`

  #11917 closedJun 30, 2025

• [tune] quniform distribution

  #11879 closedJun 30, 2025

• [docs] improve tune distributed tuning guide

  #11681 closedJun 30, 2025

• [tune] doc should indicate print output

  #11679 closedJun 30, 2025

• [cli] attach `--tmux` should show parallel command output

  #11678 closedJun 30, 2025

• [tune] Client API improvements

  #11676 closedJun 30, 2025

• [cloudpickle] Too much override for cloudpickle, breaks scikit-learn usage

  #11547 closedJun 30, 2025

• [docs] search results don't link to correct tab

  #11288 closedJun 30, 2025

• [Autoscaler] Prioritize infeasible bundles and placement group rescheduling

  #11259 closedJun 30, 2025

• Remove the `remove_after_get` flag

  #10977 closedJun 30, 2025

• [placement groups] Feasibility Check

  #10913 closedJun 30, 2025

• [autoscaler] Add unit tests for sdk.py

  #10903 closedJun 30, 2025

• Cannot call remote instance method of a superclass from within a different instance method of the superclass

  #10899 closedJun 30, 2025

• Add testing to `commands.py`/`NodeUpdaterThread` level

  #10846 closedJun 30, 2025

• Treat CPUs as abstract resources

  #10818 closedJun 30, 2025

• Installing ray on powerpc

  #10774 closedJun 30, 2025

• [core] [docs] use-cases for Ray's async support

  #10688 closedJun 30, 2025

• Exceptions and ResourceWarnings on ray.init (Jupyter+offline)

  #10279 closedJun 30, 2025

• Can CPU resource scheduling be scheduled through Cgroup?

  #10037 closedJun 30, 2025

• Windows debugging on gdb does not work

  #9827 closedJun 30, 2025

• [util.multiprocessing] Support generators

  #9712 closedJun 30, 2025

• [Core] A ray.remote flag for nested object ID gathering in task arguments.

  #9489 closedJun 30, 2025

• [docs] ray up <config.xml> --help does not show help

  #9455 closedJun 30, 2025

• [docs] Document how to use conda environments with the autoscaler

  #9199 closedJun 30, 2025

• [ray] Visualize Ray dashboard locally/offline

  #9095 closedJun 30, 2025

• Confusing RedisError when many threads are used

  #9083 closedJun 30, 2025

• [autoscaler] Check failed: _s.ok() Heartbeat failed: NotImplemented

  #8883 closedJun 30, 2025

• Can't parallelize non-pickable function with initializer in Pool

  #8876 closedJun 30, 2025

• DQN Minibatch Option

  #8870 closedJun 30, 2025

• tune: module 'tensorflow' has no attribute __version__ in Ray Trainable since v0.7.7

  #8729 closedJun 30, 2025

• Blank redis-password gives wrong message to add node

  #8629 closedJun 30, 2025

• absl.logging inside remote tasks does not get printed

  #8625 closedJun 30, 2025

• Invalid iterator dereference in TestReconstructionChain (fails in debug mode)

  #8587 closedJun 30, 2025

• Can't pickle CudnnModule objects

  #8569 closedJun 30, 2025

• Reducing unnecessary process overhead in practice

  #8522 closedJun 30, 2025

• [tune]Error in BOHB perhaps caused by different trainable instances running in the same Trial ???

  #8455 closedJun 30, 2025

• incompatible with 'msgpack_numpy.patch()' function

  #8409 closedJun 30, 2025

• Error connecting to Redis server at 127.0.0.1:35709

  #8389 closedJun 30, 2025

• Error while shutting down Ray

  #8385 closedJun 30, 2025

• [ray] Pyarmor compatibility

  #8365 closedJun 30, 2025

• [ray] Can RAY pause and continue tasks distributed to the cluster's nodes?

  #8263 closedJun 30, 2025

• [tune] unify run() and run_experiments()

  #8127 closedJun 30, 2025

• [ui] More metadata for the task timeline

  #8050 closedJun 30, 2025

• [tune] Support for config to (optionally) be an argparse.Namespace?

  #8006 closedJun 30, 2025

• [tune] Resource Allocation UX

  #7968 closedJun 30, 2025

• `pandas has no attribute 'compat'` Deserialization bug when running tasks very rarely

  #7879 closedJun 30, 2025

• "Lost reference to actor" when returning actor handle from actor

  #7815 closedJun 30, 2025

• Ray has both ray.util and ray.utils, which is confusing.

  #7787 closedJun 30, 2025

• Provide more scheduling algorithms for actors/tasks

  #7723 closedJun 30, 2025

• [ray] Object store shared memory numpy leak in worker loop

  #7653 closedJun 30, 2025

• Ray processes on slave node become defunct when the head node is restarted/stopped

  #7651 closedJun 30, 2025

• Relax python version match requirement when joining a cluster

  #7648 closedJun 30, 2025

• Does ray workers could share the same tf.sess?

  #7646 closedJun 30, 2025

• About model configuration.

  #7644 closedJun 30, 2025

• Probable race condition

  #7617 closedJun 30, 2025

• Recursion with pickling in ray.init with py3.5

  #7605 closedJun 30, 2025

• Is it possible to create process inside ray Actor?

  #7578 closedJun 30, 2025

• Why seems getting from local object store not faster than getting from remote object store?

  #7575 closedJun 30, 2025

• [util.multiprocessing] Unable to pass Queue to pool.apply_async

  #7561 closedJun 30, 2025

• Keyword arguments should be keyword only arguments in the Ray API

  #7548 closedJun 30, 2025

• [Pool] About using ray.util.multiprocessing import Pool

  #7542 closedJun 30, 2025

• Reporting Reward Breakdowns

  #7518 closedJun 30, 2025

• [config] Introduce a configuration library for unified configuration code

  #7485 closedJun 30, 2025

• Proper way of calling a class method in another method

  #7450 closedJun 30, 2025

• [autoscaler] Provide ability to provide elastic ip when launching cluster

  #7446 closedJun 30, 2025

• What are system requirements for building on Mac OSX

  #7430 closedJun 30, 2025

• Ray dashboard integration

  #7383 closedJun 30, 2025

• Do not suggest calling __ray_terminate__ directly

  #7382 closedJun 30, 2025

• ray.services.get_node_ip_address doesn't work well if there is a local proxy

  #7316 closedJun 30, 2025

• Provide abstraction/interface to implement resource isolation for custom resources

  #7204 closedJun 30, 2025

• [cross-language]Problem about cross language data layout

  #7191 closedJun 30, 2025

• Documentation for connecting to ray cluster could be improved

  #7186 closedJun 30, 2025

• ray.experimental.queue is very slow

  #7172 closedJun 30, 2025

• Using asserts for argument checks is probably a bad idea

  #7171 closedJun 30, 2025

• Ray Issue: The class state is never hold by passing it to the remote function/actors if the class is defined in separate files

  #7160 closedJun 30, 2025

• [core] Gets timeout on randomly generated ObjectIDs

  #7074 closedJun 30, 2025

• Allow remote functions to require running on a fresh worker

  #7059 closedJun 30, 2025

• How to use Ray with closures?

  #7055 closedJun 30, 2025

• The project `setup.py` script doesn't install tools needed by `ci/travis/format.sh`

  #6999 closedJun 30, 2025

• Don't run Java or sanitizer tests when only Python changes.

  #6992 closedJun 30, 2025

• ray plasma object store connection refused after 24hrs

  #6988 closedJun 30, 2025

• Sharing in memory

  #6976 closedJun 30, 2025

• [ray] ray on slurm not respecting memory limits

  #6968 closedJun 30, 2025

• Unable to override ray's default logging format

  #6965 closedJun 30, 2025

• MADDPG used onto a MultiEnv does not show learning.

  #6949 closedJun 30, 2025

• How to throttle process to avoid "UnreconstructableError"

  #6892 closedJun 30, 2025

• pip install from source requires --editable/-e flag

  #6845 closedJun 30, 2025

• [scheduling] Default actor lifetime resources (0 CPUs) cause cluster not to be saturated

  #6814 closedJun 30, 2025

• How to Reduce Memory Usage for Creating Actor?

  #6778 closedJun 30, 2025

• Reconstruction semantics around failing actor constructor.

  #6768 closedJun 30, 2025

• [Deploy]Ray on Yarn Deployment

  #6753 closedJun 30, 2025

• failed on virtualnevironment

  #6735 closedJun 30, 2025

• Managing memory during long loops

  #6717 closedJun 30, 2025

• Not able to reproduce speed performance improvements using ray on my machine

  #6716 closedJun 30, 2025

• [tune] Logs don't sync up to workers on restore

  #6702 closedJun 30, 2025

• The remote_function.options is not documented.

  #6699 closedJun 30, 2025

• [tune] More robust checkpoint garbage collection

  #6697 closedJun 30, 2025

• Fault tolerance to dead actors

  #6670 closedJun 30, 2025

• ray.wait's num_returns should not fail if num_returns > len(results)

  #6667 closedJun 30, 2025

• Parallel execution of multiple dataframes by dividing them into sub-frames

  #6640 closedJun 30, 2025

• Batch Norm example failing under APEX

  #6638 closedJun 30, 2025

• limiting tensorflow memory failed in actor or function

  #6633 closedJun 30, 2025

• Remote function is executed in python `exec` with empty local/global will fails

  #6620 closedJun 30, 2025

• [tune] Estimate timing

  #6618 closedJun 30, 2025

• [streaming] Add micro batching feature

  #6607 closedJun 30, 2025

• Package reference should include task & actor APIs

  #6566 closedJun 30, 2025

• Serialization is 20% slower from 0.7.6 -> 0.7.7

  #6551 closedJun 30, 2025

• [ray] How to write into numpy arrays in shared memory with Ray?

  #6507 closedJun 30, 2025

• Support for mxnet.ndarray?

  #6494 closedJun 30, 2025

• [ray] Handle memory pressure more gracefully

  #6458 closedJun 30, 2025

• Reloading module changes in workers

  #6449 closedJun 30, 2025

• [tune] [serve] Don't use daemon threads

  #6421 closedJun 30, 2025

• Terminal freezes after setting @ray.remote(num_gpu=2)

  #6418 closedJun 30, 2025

• Ray does not preserve requires_grad attribute

  #6405 closedJun 30, 2025

• Ray over mpi for supercomputers

  #6344 closedJun 30, 2025

• Support of Ray Decorator for Built in Functions

  #6308 closedJun 30, 2025

• [docs] Issue on `tune-schedulers.rst`

  #6063 closedJun 30, 2025

• Can I set priority for my tasks

  #6057 closedJun 30, 2025

• Avoid putting the redis password in plain text in processlist

  #5872 closedJun 30, 2025

• Handling `use_pickle=True` with pickle5 serializer and performance regression

  #5856 closedJun 30, 2025

• Install ray with conda but not pip

  #5511 closedJun 30, 2025

• [tune] saving mechanism and PBT

  #5312 closedJun 30, 2025

• Feature request: An API to wait until there are are X resources available

  #5243 closedJun 30, 2025

• [Feature request] Also expose python function after decorating with ray.remote

  #4981 closedJun 30, 2025

• Creative action space support: contains method, action interpoalation.

  #4837 closedJun 30, 2025

• __module__ can be None

  #4758 closedJun 30, 2025

• [autoscaler] Autoscaler UX Issues

  #4656 closedJun 30, 2025

• [autoscaler] Add tests that mock endpoints for AWS, GCE

  #4303 closedJun 30, 2025

• Python Worker class should have proper constructor and destructor.

  #3961 closedJun 30, 2025

• Should not ignore "AttributeError"

  #3820 closedJun 30, 2025

• Backend timing statements should be made type safe.

  #3341 closedJun 30, 2025

• Make it possible to limit memory usage of processes

  #3055 closedJun 30, 2025

• Task submission from local scheduler client is blocking

  #2940 closedJun 30, 2025

• Add test for numpy array alignment.

  #2937 closedJun 30, 2025

• Allow ray.get and ray.wait to take in additional argument types

  #2126 closedJun 30, 2025

• Remove the import thread from the workers and driver.

  #951 closedJun 30, 2025

• Remote decorator fails on jitted function.

  #593 closedJun 30, 2025

• Actors do not work properly with subclasses that call super.

  #449 closedJun 30, 2025

• Methods on actors inherited from built-in classes are not visible

  #278 closedJun 30, 2025

• CI test darwin://python/ray/dashboard:modules/aggregator/tests/test_aggregator_agent is consistently_failing

  #54191 closedJun 30, 2025

• Release test random_shuffle_fixed_size failed

  #53806 closedJun 30, 2025

• Assessment of the difficulty in porting CPU architecture for Ray

  #54162 closedJun 30, 2025

• CI test linux://python/ray/data:test_arrow_block_scaling is flaky

  #54110 closedJun 30, 2025

• [Epic][Docs/KubeRay] Convert doctests back to normal markdown docs

  #54072 closedJun 28, 2025

• [Docs][KubeRay] Convert configuring-autoscaling.ipynb back to markdown docs

  #54077 closedJun 28, 2025

• [Docs][KubeRay] Convert rayservice-quick-start.ipynb back to markdown docs

  #54076 closedJun 28, 2025

• [Docs][KubeRay] Convert raycluster-quick-start.ipynb back to markdown docs

  #54074 closedJun 28, 2025

• CI test linux://rllib:examples/connectors/multi_agent_with_different_observation_spaces is flaky

  #53473 closedJun 28, 2025

• [Data] When writing on BigQuery, Google's "TooManyRequests" exceptions is not retried

  #53997 closedJun 28, 2025

• [RayLLM] RayLLM / vLLM production stack integration

  #53331 closedJun 27, 2025

• [Core] Ray fails to fulfill request due to node being annotated by IP address

  #54152 closedJun 27, 2025

• [Docs][KubeRay] Delete KubeRay doctests

  #54073 closedJun 27, 2025

• CI test linux://:local_object_manager_test is flaky

  #54131 closedJun 27, 2025

• [Data] `ArrowInvalid` during `ray.data.from_huggingface`: Parquet magic bytes not found in footer

  #54101 closedJun 27, 2025

• CI test linux://rllib:examples/algorithms/vpg_custom_algorithm is flaky

  #53925 closedJun 27, 2025

• CI test linux://rllib:examples/algorithms/appo_custom_algorithm_w_shared_data_actor is flaky

  #53176 closedJun 27, 2025

• CI test linux://rllib:examples/evaluation/evaluation_parallel_to_training_multi_agent_duration_auto is flaky

  #53255 closedJun 27, 2025

• CI test linux://:node_manager_test is flaky

  #54059 closedJun 27, 2025

• [Serve] UnboundLocalError: local variable 'stopped' in deployment state

  #54169 closedJun 27, 2025

• [Core] Exiting because this node manager has mistakenly been marked as dead by the GCS

  #54035 closedJun 27, 2025

• [bug][serve.llm] AssertionError: failed to get the hash of the compiled graph (VLM, batch, TP=2)

  #53824 closedJun 27, 2025

• [Serve, LLM] missing botocore dependency!

  #53052 closedJun 27, 2025

• Error Handling Large Pyarrow Chunk

  #53536 closedJun 26, 2025

• CI test linux://python/ray/train/v2:test_controller is consistently_failing

  #54147 closedJun 26, 2025

• [Serve][LLM] Qwen3 models “enable_thinking: False” still returns thinking process

  #52979 closedJun 26, 2025

• [Core] Ray fails to fulfill request due to node being annotated by IP address

  #54150 closedJun 26, 2025

• [Docs][KubeRay] Convert kuberay-gcs-ft.ipynb back to markdown docs

  #54078 closedJun 26, 2025

• [Docs][KubeRay] Convert rayjob-quick-start.ipynb back to markdown docs

  #54075 closedJun 26, 2025

• CI test darwin://python/ray/tests:test_basic_3_client_mode is consistently_failing

  #54126 closedJun 26, 2025

• CI test windows://python/ray/tests:test_basic_3_client_mode is consistently_failing

  #54132 closedJun 26, 2025

• [Core] Transient network failure on RPC `MarkJobFinished` causes node crash

  #53645 closedJun 26, 2025

• CI test linux://python/ray/tests:test_basic_3_client_mode is consistently_failing

  #54119 closedJun 26, 2025

• [Doc] The anchors of headers doesn't follow Vale rules.

  #53516 closedJun 26, 2025

• CI test linux://:local_object_manager_test is flaky

  #54130 closedJun 26, 2025

• [Core] Could not connect to socket

  #54067 closedJun 26, 2025

• [core] TSAN failing on `node_manager_test`

  #54096 closedJun 25, 2025

• CI test linux://python/ray/data:test_arrow_block is flaky

  #48859 closedJun 25, 2025

• CI test linux://python/ray/data:test_huggingface is consistently_failing

  #44516 closedJun 25, 2025

• CI test linux://python/ray/train:accelerate_torch_trainer_no_raydata is consistently_failing

  #48939 closedJun 25, 2025

• CI test linux://python/ray/train:deepspeed_torch_trainer is consistently_failing

  #44517 closedJun 25, 2025

• Release test training_ingest_benchmark-task=image_classification.full_training.jpeg failed

  #53953 closedJun 25, 2025

• [Core] Autoscaler Node Recovery Ignores Node-Specific Docker Config

  #53987 closedJun 25, 2025

• [Air] WandB logger / _WandbLoggingActor crashes silently when logging a video with relative path or if it cannot find the given file.

  #50307 closedJun 25, 2025

• [Doc][KubeRay] Run doctest `user-guides/configuring-autoscaling.ipynb` in CI

  #53989 closedJun 25, 2025

• CI test windows://python/ray/tests:test_basic is consistently_failing

  #51497 closedJun 25, 2025

• [CI] Migrate from flake8 to ruff

  #34889 closedJun 25, 2025

• [Docker] Upgrade the base image from ubuntu:focal to ubuntu:22.04LTS

  #35514 closedJun 25, 2025

• CI test linux://python/ray/data:test_backpressure_e2e is flaky

  #49963 closedJun 25, 2025

• CI test linux://python/ray/tests:test_runtime_env_complicated is consistently_failing

  #49674 closedJun 25, 2025

• CI test linux://python/ray/data:test_execution_optimizer is consistently_failing

  #44410 closedJun 25, 2025

• [Dashboard] Decorator that exposes attribute to dashboard for display in grid

  #33188 closedJun 24, 2025

• [serve] AttributeError when attempting to use serve with cluster and FastAPI

  #54008 closedJun 24, 2025

• [gcp] Node mistakenly marked dead: increase heartbeat timeout?

  #16945 closedJun 24, 2025

• Docs on Cython extensions and install requirements

  #7094 closedJun 24, 2025

• [core] Detached actor being killed when its parent actor crashes

  #40864 closedJun 24, 2025

• CI test linux://doc:doctest[data] is consistently_failing

  #54036 closedJun 24, 2025

• CI test linux://python/ray/data:doctest is consistently_failing

  #44570 closedJun 24, 2025

• [data/proprocessors] Support flattening vector features in concatenator

  #51757 closedJun 24, 2025

• [Docs][KubeRay] Don't sleep for a long time in `kuberay-gcs-ft.ipynb`

  #54040 closedJun 24, 2025

• Release test many_nodes_actor_test_on_v2.aws failed

  #53990 closedJun 24, 2025

• CI test linux://doc/source/train/examples/lightning:lightning_cola_advanced is consistently_failing

  #44545 closedJun 24, 2025

• CI test linux://python/ray/train:accelerate_torch_trainer is consistently_failing

  #44513 closedJun 24, 2025

• CI test linux://doc/source/train/examples/transformers:huggingface_text_classification is consistently_failing

  #44546 closedJun 24, 2025

• CI test linux://python/ray/train:deepspeed_torch_trainer_no_raydata is consistently_failing

  #44932 closedJun 24, 2025

• CI test windows://python/ray/serve/tests:test_request_timeout is flaky

  #48417 closedJun 24, 2025

• [old]

  #54020 closedJun 23, 2025

• CI test windows://python/ray/serve/tests:test_batching is consistently_failing

  #46016 closedJun 23, 2025

• CI test linux://python/ray/tests:test_runtime_env_container is consistently_failing

  #45223 closedJun 23, 2025

• [CI] `linux://python/ray/tests:test_state_api` is failing/flaky on master.

  #54001 closedJun 23, 2025

• Ability to select a disk for ray workers

  #8607 closedJun 23, 2025

• Conflict between ROCR_VISIBLE_DEVICES and HIP_VISIBLE_DEVICES environment variables causes Ray import error

  #53737 closedJun 21, 2025

• CI test linux://python/ray/serve/tests:test_multiplex is flaky

  #48378 closedJun 21, 2025

• [RLlib] MAML does not work with TF2 in Ray 2.3.1

  #34620 closedJun 20, 2025

• [RayData|RayServe] Does RayData/RayServe support multi-node vllm inference

  #53192 closedJun 20, 2025

• [Core] Core Worker crashing

  #49088 closedJun 20, 2025

• [core][gpu-objects] Driver tries to get the data from in-actor store

  #51272 closedJun 19, 2025

• [Core][ROCm] Setting CUDA_VISIBLE_DEVICES leads to an assertion

  #52701 closedJun 19, 2025

• [Autoscaler][V2] Autoscaler fails to delete idle KubeRay Pod

  #52264 closedJun 19, 2025

• CI test linux://python/ray/data:test_consumption is flaky

  #48163 closedJun 19, 2025

• CI test windows://python/ray/tests:test_actor_state_metrics is consistently_failing

  #46303 closedJun 19, 2025

• [data] ray.data.read_images is slower than reading images manually

  #37499 closedJun 19, 2025

• [RFC] Q2 Ray Data Roadmap

  #51808 closedJun 19, 2025

• [RFC] LLM APIs for Ray Data and Ray Serve

  #50639 closedJun 19, 2025

• [<Ray component: Core|RLlib|etc...>] ValueError: There was an error while reducing the Stats object under key=('actual_n_step',)!

  #53947 closedJun 19, 2025

• CI test windows://python/ray/serve/tests:test_standalone_3 is flaky

  #44003 closedJun 19, 2025

• Release test compiled_graphs failed

  #53716 closedJun 18, 2025

• CI test darwin://python/ray/tests:test_metrics_agent_open_telemetry is consistently_failing

  #53828 closedJun 18, 2025

• [RLlib] ActionMaskingTorchRLModule can't set up `conv_filters`

  #53325 closedJun 18, 2025

• [Core] `ray.init()` and `ray start` fails on Windows 11 in ray 2.45+

  #52739 closedJun 18, 2025

• CI test windows://python/ray/tests:test_object_spilling_debug_mode is flaky

  #43796 closedJun 18, 2025

• [core] support S3 path style access in runtime_env download_and_unpack_package()

  #53893 closedJun 17, 2025

• How to transfer tensors stored in GPU in actor with NCCL?

  #53816 closedJun 17, 2025

• [Data] PyArrow 20.0.0 Backward Incompatability (`unexpected keyword argument 'maps_as_pydicts'`)

  #52685 closedJun 17, 2025

• CI test linux://python/ray/tests:test_gpu_objects_nccl is consistently_failing

  #53871 closedJun 17, 2025

• [RLlib] Headnode without GPU triggers torch/CUDA de-serialization error

  #53467 closedJun 17, 2025

• [Core] Ray Autoscaler does not restart a worker node on setup failure

  #29127 closedJun 17, 2025

• Release test llm_batch_vllm failed

  #53827 closedJun 17, 2025

• [Serve] Add timeout parameter for `deploy`

  #25433 closedJun 17, 2025

• [Core] Read-only buffer error in some scikit-learn models

  #52571 closedJun 17, 2025

• [core] ray stop --force doesn't kill processes on worker node

  #28038 closedJun 17, 2025

• [core][gpu-objects] Support TensorDict

  #51550 closedJun 17, 2025

• [core][gpu-objects] Allocate placeholder tensor on corresponding devices

  #53622 closedJun 17, 2025

• [core][gpu-objects] Driver should order all collective calls to avoid deadlock

  #51264 closedJun 17, 2025

• CI test windows://python/ray/tests:test_object_spilling_asan is consistently_failing

  #45962 closedJun 17, 2025

• CI test windows://python/ray/tests:test_object_spilling is consistently_failing

  #45961 closedJun 16, 2025

• [RLlib] Add syntax checking to configuration string literals or migrate to enums.

  #39384 closedJun 16, 2025

• [Ray Core] Ray error causes the Python interpreter to terminate without failing

  #28211 closedJun 16, 2025

• [CI] Test GPU training tutorial with Ray Release tests

  #28902 closedJun 16, 2025

• [core][gpu-objects] intra-process communication

  #51685 closedJun 16, 2025

• CI test windows://python/ray/tests:test_basic_client_mode is flaky

  #52117 closedJun 13, 2025

• [Serve] check_health with custom exception does not enter failed state, infinite retries

  #53742 closedJun 13, 2025

• [core][gpu-objects] Object contains multiple tensors and/or mix of CPU data and GPU tensors

  #51274 closedJun 13, 2025

• CI test windows://python/ray/serve/tests:test_standalone_with_comp_sche is flaky

  #48425 closedJun 13, 2025

• CI test linux://python/ray/tune:test_tuner is consistently_failing

  #53786 closedJun 13, 2025

• Release test serve_autoscaling_load_test.aws failed

  #53760 closedJun 13, 2025

• Release test llm_serve_llama_3dot1_8B_quantized_tp1_2p6d failed

  #53769 closedJun 13, 2025

• Release test llm_serve_llama_3dot1_8B_quantized_tp1_1p1d failed

  #53768 closedJun 13, 2025

• Release test llm_serve_llama_3dot1_8B_quantized_tp_1 failed

  #53764 closedJun 13, 2025

• Release test llm_serve_llama_3dot1_8B_lora failed

  #53766 closedJun 13, 2025

• Release test llm_serve_llama_3dot2_1B_s3 failed

  #53767 closedJun 13, 2025

• Release test llm_serve_llama_3dot2_1B_no_accelerator failed

  #53765 closedJun 13, 2025

• Release test llm_serve_llama_3dot1_8B_tp_2 failed

  #53763 closedJun 13, 2025

• Release test serve_scale_replicas.aws failed

  #53761 closedJun 13, 2025