124 Pull requests merged by 41 people

• [RLlib] Fix MultiAgentEnvRunner env check bug.

  #50891 merged Feb 25, 2025

• [data] add iceberg write support through pyiceberg

  #50590 merged Feb 25, 2025

• [ray.llm.batch] Ensure the type of prompt_token_ids

  #50865 merged Feb 25, 2025

• [serve.llm] remove backoff dependency

  #50822 merged Feb 25, 2025

• [core] Implement log rotation for worker

  #50759 merged Feb 25, 2025

• [docs][core] Improve Compiled Graphs docs

  #50627 merged Feb 25, 2025

• [llm.serving] Reconfigure router to better perform under high concurrency (#50876)

  #50884 merged Feb 25, 2025

• [core] Cover cpplint for ray/src/ray/internal

  #50662 merged Feb 25, 2025

• [ci] convert the if-else change in cond testing script into a config file

  #50776 merged Feb 25, 2025

• [llm.serving] Reconfigure router to better perform under high concurrency

  #50876 merged Feb 25, 2025

• [Dashboard] Added a proper message when Ray jobs submission is used a…

  #49409 merged Feb 25, 2025

• [core] [cgroup] Interface for cgroup setup and its fake implementation

  #50787 merged Feb 25, 2025

• [1/N] bazel lint C++ code

  #50869 merged Feb 25, 2025

• Change conda env validity checking to activate and run Python

  #48552 merged Feb 25, 2025

• [Serve] Faster bulk imperative Serve Application deploys

  #49168 merged Feb 25, 2025

• [data/preprocessors] feat: allow vectorizers to be executed in append mode

  #50847 merged Feb 24, 2025

• [data/preprocessors] feat: allow tokenizer to execute in append mode

  #50848 merged Feb 24, 2025

• cherrypick #50860

  #50867 merged Feb 24, 2025

• [CI] pass cloud into service.status calls

  #50860 merged Feb 24, 2025

• [ray.llm][Batch] Support LoRA

  #50804 merged Feb 24, 2025

• [train v2] Fix trainer import deserialization when captured within a Ray tasks

  #50862 merged Feb 24, 2025

• [core] Remove unused release_resources flag

  #50854 merged Feb 24, 2025

• [llm.serving] Fix using uni executor when world size == 1 (#50849)

  #50863 merged Feb 24, 2025

• [llm.serving] Fix using uni executor when world size == 1

  #50849 merged Feb 24, 2025

• [cleanup] Remove ray.storage._load_class indirection

  #50857 merged Feb 24, 2025

• Docs: add suspend to RayJob Configuration in RayJob QuickStart

  #50641 merged Feb 24, 2025

• [docs] add shell completion instructions for kubectl plugin

  #50808 merged Feb 24, 2025

• Add recommendations for when to use make local vs make develop

  #50773 merged Feb 24, 2025

• [core] Avoid creating a boost::asio::thread_pool with 0 threads

  #50837 merged Feb 24, 2025

• [data/preprocessors] feat: allow discretizers to be used in append mode

  #50584 merged Feb 24, 2025

• [core] Remove test_events_with_crash

  #50780 merged Feb 24, 2025

• [doc] unify the look and feel of anyscale buttons

  #50800 merged Feb 24, 2025

• Make core team codeowners for workflows

  #50855 merged Feb 24, 2025

• [Core] Add CancelTaskWithResourceShapes to Node Raylet Client and Node Manager

  #50200 merged Feb 24, 2025

• [ADAG]Redundant code optimization

  #50305 merged Feb 24, 2025

• [core][cgraph] Fix compiled graph buffer release issues

  #50434 merged Feb 24, 2025

• [serve.llm] remove asyncache and cachetools from dependencies.

  #50806 merged Feb 23, 2025

• [core] Guard concurrent access to generator IDs with a mutex

  #50845 merged Feb 23, 2025

• [core] Guard concurrent access to generator IDs with a mutex

  #50740 merged Feb 23, 2025

• [core] Cover cpplint for ray/tree/master/src/ray/rpc

  #50727 merged Feb 23, 2025

• [core] Cover cpplint for src/ray/pubsub

  #50732 merged Feb 23, 2025

• [ci] more project id fixes for release test launching

  #50841 merged Feb 23, 2025

• cherrypick #50841

  #50842 merged Feb 23, 2025

• cherrypick #50836

  #50840 merged Feb 23, 2025

• [ci] allow specifying project id for release test

  #50836 merged Feb 23, 2025

• [doc] replace PNGs

  #50810 merged Feb 22, 2025

• cherrypick #50816 and #50829

  #50834 merged Feb 22, 2025

• [ci] fix release test launching

  #50829 merged Feb 22, 2025

• [RLlib] APPO accelerate (vol 19): Torch clip util enhancements.

  #50791 merged Feb 22, 2025

• cherrypick #50805

  #50832 merged Feb 22, 2025

• [release] change version to 2.43.0

  #50831 merged Feb 22, 2025

• [Core] Update Error Message and Anti-Pattern for the Case of Forking New Processes in Worker Processes

  #50705 merged Feb 22, 2025

• [release test] upgrade anyscale cli version

  #50816 merged Feb 22, 2025

• Fix minor typo in ray-client.rst

  #50809 merged Feb 21, 2025

• [deps] add anyscale CLI dependency into resolving

  #50700 merged Feb 21, 2025

• Various improvements to Serve Request Batching tutorial

  #50400 merged Feb 21, 2025

• [Core] Fix windows compilation error

  #50805 merged Feb 21, 2025

• [RLlib] Implement vectorization for MultiAgentEnv (new API stack).

  #50437 merged Feb 21, 2025

• [Release Tests][Data] Print Resource Manager stats in release tests

  #50801 merged Feb 21, 2025

• [Core] Split gcs into smaller targets

  #50764 merged Feb 21, 2025

• Fix indefinite article in CONTRIBUTING rst

  #50651 merged Feb 21, 2025

• [llm] removes unused aiobotocore dependency

  #50797 merged Feb 21, 2025

• [Core] Record core usage

  #50756 merged Feb 21, 2025

• [data] Adding in TaskDurationStats and on_execution_step callback

  #50766 merged Feb 21, 2025

• [llm.serving] add requirements sections to overview page

  #50788 merged Feb 21, 2025

• [train][ci] Bump up the timeout of test_data_parallel_trainer

  #50796 merged Feb 21, 2025

• [llm.serving] Update default batch timeout to 50ms

  #50786 merged Feb 21, 2025

• [core] Run redis tests on postmerge only

  #50795 merged Feb 21, 2025

• [core] Cover cpplint for ray/src/ray/scheduling

  #50686 merged Feb 21, 2025

• [core] [easy] [noop] Move pid_t alias to compatibility header

  #50789 merged Feb 21, 2025

• [llm.serving] Address dependencies related issues for llm serving

  #50785 merged Feb 21, 2025

• [llm.serving] fix telemetry test related to autoscaling changes

  #50779 merged Feb 21, 2025

• [LLM Serve] Fix docs example for vision language model

  #50782 merged Feb 21, 2025

• [docs][data/llm] Minor http docstring improvement

  #50777 merged Feb 21, 2025

• [serve.llm] Post dogfooding changes to the docs

  #50781 merged Feb 21, 2025

• [Data] Fixing aggregation protocol to be appropriately associative

  #50757 merged Feb 21, 2025

• [doc] Add Anyscale as an option to get started with Ray

  #50772 merged Feb 21, 2025

• [llm.serving] Refactor LLMConfig#get_serve_options

  #50753 merged Feb 21, 2025

• [serve.llm] Rename APIs to LLMRouter and VLLMService

  #50775 merged Feb 21, 2025

• [deps] remove indirect referencing

  #50742 merged Feb 21, 2025

• [ci] better line splitting on diff output

  #50765 merged Feb 20, 2025

• [llm.batching] fix doc

  #50774 merged Feb 20, 2025

• [Core] Add a Function to Get Per Node Infeasible Request Resource Shapes in GCS Autoscaling State Manager

  #50085 merged Feb 20, 2025

• [llm.serving] validate null response content

  #50771 merged Feb 20, 2025

• [ray.serve.llm][docs] Added llm serving docs

  #50675 merged Feb 20, 2025

• [core] Fix windows build from ray.wait race pr

  #50758 merged Feb 20, 2025

• [doc][core][cgraph] Passing filename to viz

  #50752 merged Feb 20, 2025

• [serve][tests] Print the wrk error to the driver logs if it exists

  #50755 merged Feb 20, 2025

• [RLlib] TorchLearner: Don't call no_sync (DDP/multi-GPU) on non-torch modules.

  #50760 merged Feb 20, 2025

• [core] Fix missing type alias on windows

  #50716 merged Feb 20, 2025

• [ray.llm] Minor improvments

  #50748 merged Feb 20, 2025

• Revert "[core] Implement redirection for core worker (#50398)"

  #50738 merged Feb 20, 2025

• [CI] Enable check-json hooks in pre-commit

  #50366 merged Feb 20, 2025

• [llm.serving] fix more diff on test_usage

  #50744 merged Feb 20, 2025

• [data/llm/docs] LLM Batch API documentation improvements

  #50747 merged Feb 20, 2025

• [core][cgraph] Revive max_buffered_result arg

  #50725 merged Feb 20, 2025

• [data/preprocessors] feat: allow simple imputer to execute on append mode

  #50713 merged Feb 20, 2025

• [deps] upgrade certifi

  #50701 merged Feb 19, 2025

• [deps] fix bash styling in install-deps script

  #50734 merged Feb 19, 2025

• [data] Simplify Operator.__repr__

  #50620 merged Feb 19, 2025

• [RLLib] Pass large AlgorithmConfig by reference to RolloutWorker

  #50688 merged Feb 19, 2025

• [deps] explicitly add dl-cpu requirements in ray-ml docker build file

  #50733 merged Feb 19, 2025

• [data] Make iter_torch_batches release test only use GPU for the head node

  #50712 merged Feb 19, 2025

• [deps] compile dependencies in the repo root dir

  #50702 merged Feb 19, 2025

• [doc][cgraph] Reference compiled graph in Ray docs

  #50654 merged Feb 19, 2025

• [data/preprocessors] feat: allow normalizer to be used in append mode

  #50714 merged Feb 19, 2025

• [llm.serving] fix usage test

  #50730 merged Feb 19, 2025

• [core] Avoid logger flush at every write

  #50722 merged Feb 19, 2025

• [ray.serve.llm] Fix setting up AutoProcessor

  #50715 merged Feb 19, 2025

• Update working-with-llms.rst

  #50717 merged Feb 19, 2025

• [core] Remove unused mock class

  #50719 merged Feb 19, 2025

• [release] Try fix data loss

  #50709 merged Feb 19, 2025

• [core] Implement redirection for core worker

  #50398 merged Feb 19, 2025

• [ray.llm] Support assistant prefill in chat template stage

  #50628 merged Feb 19, 2025

• [data/llm/docs] Initial draft of user guide for Data LLM APIs

  #50674 merged Feb 19, 2025

• [LLM] OSS LLM Serving

  #50643 merged Feb 19, 2025

• [ray.llm][Batch] Add __init__.py

  #50699 merged Feb 18, 2025

• [deps] add more dependencies for ray[llm]

  #50692 merged Feb 18, 2025

• [Data] Replace AggregateFn with AggregateFnV2, cleaning up Aggregation infrastructure

  #50585 merged Feb 18, 2025

• [core][cgraph] Only show type on visualize with channel_details

  #50689 merged Feb 18, 2025

• [data][dashboard] allow reusing of dashboard conftest

  #50614 merged Feb 18, 2025

• [deps] upgrade h11, httpcore and httpx

  #50691 merged Feb 18, 2025

• Add nightly Python 3.13 wheels to documentation and build aarch64 wheels

  #50667 merged Feb 18, 2025

• [core] Add back workflow test

  #50660 merged Feb 18, 2025

39 Pull requests opened by 24 people

• minor improvements to hyperopt tutorial

  #50697 opened Feb 18, 2025

• various improvements to lightgbm tutorial:

  #50704 opened Feb 19, 2025

• [Autoscaler][V2] Check IM instance_status before terminating nodes

  #50707 opened Feb 19, 2025

• [doc] Add documentation for Asynchronous HyperBand Example in Tune

  #50708 opened Feb 19, 2025

• [core] Adding warnings for non zero-copy serialization

  #50731 opened Feb 19, 2025

• [WIP][core] Use `DestroyWorker` to disconnect raylet client when killing leased workers

  #50736 opened Feb 19, 2025

• [doc][core][cgraph] Add Compiled Graph API

  #50754 opened Feb 20, 2025

• [core] [wip attempt] StatusOr union construction sometimes breaks windows build

  #50761 opened Feb 20, 2025

• docs: add quickstart button to examples

  #50763 opened Feb 20, 2025

• [data] Add dataset/operator state, progress, total metrics

  #50770 opened Feb 20, 2025

• [core] Unit tests for tensor serialization

  #50778 opened Feb 21, 2025

• [WIP] Ray Collective Communication Lib Support HCCL Backend

  #50790 opened Feb 21, 2025

• [docs] add missing step to install KubeRay in gke-gcs-bucket.md

  #50811 opened Feb 21, 2025

• [wip] Detect socket closed

  #50812 opened Feb 21, 2025

• [serve.llm] remove asyncio_timeout from dependencies

  #50815 opened Feb 22, 2025

• [serve.llm] import boto3 and google lazily but make them part of requirements

  #50820 opened Feb 22, 2025

• [serve.llm] Made json validator a singleton and jsonref packages lazy imported

  #50821 opened Feb 22, 2025

• [core] Subprocess to cleanup resource for parent process

  #50830 opened Feb 22, 2025

• [core][cgraph] Keep cgraph reference alive for teardown

  #50851 opened Feb 24, 2025

• [wip] remove block/unblock calls to see what breaks

  #50852 opened Feb 24, 2025

• [wip] Try removing `NotifyUnblocked`

  #50853 opened Feb 24, 2025

• [data/preprocessors] feat: allow transformer to be executed in append mode

  #50856 opened Feb 24, 2025

• Add perf metrics for 2.43.0

  #50864 opened Feb 24, 2025

• Improvements to PBT example

  #50870 opened Feb 25, 2025

• [llm.serving] Import AutoscalingConfig & DeploymentConfig from Serve

  #50871 opened Feb 25, 2025

• [train] Remove ray storage dependency and deprecate `RAY_STORAGE` env var configuration option

  #50872 opened Feb 25, 2025

• [llm.serving] Made accelerator_type an optional field in LLMConfig

  #50873 opened Feb 25, 2025

• [Train V2] Hide the private functions of train context to avoid abuse

  #50874 opened Feb 25, 2025

• [core] Setup log rotation for runtime env agent

  #50877 opened Feb 25, 2025

• Various improvements to the Getting Started page

  #50878 opened Feb 25, 2025

• [WIP] Rebasing materialized dataset on iterator back-pressure is active upon materialization

  #50880 opened Feb 25, 2025

• [core] Split object manager into small C++ targets

  #50885 opened Feb 25, 2025

• [Core] Logic Added to Cancel Infeasible Tasks in GCS Based on Autoscaler State

  #50886 opened Feb 25, 2025

• [compiled graphs] Remove unused external stream in ExecutableTask

  #50887 opened Feb 25, 2025

• [serve] Wait for proxy to start before running wrk trial

  #50888 opened Feb 25, 2025

• bazel lint all cpp folder

  #50889 opened Feb 25, 2025

• [data] mark pyarrow nightly tests as soft fail

  #50892 opened Feb 25, 2025

• Minor readability improvements to Ray Data Quickstart

  #50893 opened Feb 25, 2025

• Add perf metrics for 2.43.0

  #50894 opened Feb 25, 2025

93 Issues closed by 16 people

• [Serve] Can't autoscale deployment when target ongoing requests is 1

  #24793 closed Feb 25, 2025

• [serve] Prefer stopping pending replicas instead of running ones

  #12929 closed Feb 25, 2025

• [Serve] Handle prefers replicas on the same node

  #13108 closed Feb 25, 2025

• [serve] Extend multi-app unit tests

  #34450 closed Feb 25, 2025

• [Core] Extend Ray Data with IcebergDataSink for Distributed Writes to Iceberg Tables

  #49032 closed Feb 25, 2025

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

  #47295 closed Feb 25, 2025

• [Core] Ray creates a Issue with catboost

  #50843 closed Feb 25, 2025

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

  #46683 closed Feb 25, 2025

• [core] Cover cpplint for ray/src/ray/internal

  #50608 closed Feb 25, 2025

• CI test linux://rllib:examples/connectors/mean_std_filtering_ppo is consistently_failing

  #47435 closed Feb 25, 2025

• CI test linux://python/ray/train/v2:test_v2_api is consistently_failing

  #50879 closed Feb 25, 2025

• CI test windows://python/ray/tests:test_reference_counting_2 is flaky

  #45964 closed Feb 25, 2025

• CI test windows://python/ray/tests:test_actor_retry2 is flaky

  #47415 closed Feb 25, 2025

• Ray cluster launcher on existing GCP VMs, head node unable to SSH to worker nodes

  #34838 closed Feb 25, 2025

• [Ray Data] read_binary_files does not load data from S3 in parallel

  #44215 closed Feb 25, 2025

• [Tune] pbt_transformers.py failing to run - error code

  #37688 closed Feb 25, 2025

• [Train] Ray air benchmark example test break with nightly build

  #43091 closed Feb 25, 2025

• [Ray[Data]] Put failure: FileNotFoundError

  #43282 closed Feb 25, 2025

• RayTaskError in my Trainable class.

  #43137 closed Feb 25, 2025

• [<Ray component: Train>] OOM when loading 70B LLAMA2

  #43318 closed Feb 25, 2025

• tune.with_parameters() throws resource exhaustion exception

  #42590 closed Feb 25, 2025

• [Feature] Add support for GAIL/AIRL

  #19394 closed Feb 25, 2025

• How to enable login and logout button in Ray Dashboard

  #42606 closed Feb 25, 2025

• Ray tune - _TrainingRunMetadata generates incorrect results

  #42628 closed Feb 25, 2025

• [CI] `linux://python/ray/air:test_resource_manager_placement_group` is failing/flaky on master.

  #40505 closed Feb 25, 2025

• [<Ray component: Tune>] TensorflowCheckpoint Temp Folder is hard coded to system variable

  #41205 closed Feb 25, 2025

• [Train] Add user guide for configuring cuda devices

  #41958 closed Feb 25, 2025

• [Train] LightGBM stuck with more workers

  #41790 closed Feb 25, 2025

• Experiment class and search algorithm

  #41295 closed Feb 25, 2025

• Ray 2.6.1 error: pydantic doesn't have attribute `__version__`

  #38343 closed Feb 25, 2025

• [docs][rllib] Broken link to PyTorch Dynamo /rllib/rllib-torch2x.html

  #40868 closed Feb 25, 2025

• [RLlib] "RuntimeError: The learner thread died while training!" after a view training cycles.

  #36801 closed Feb 25, 2025

• [Doc] Help every user understand what a "driver" is

  #41046 closed Feb 25, 2025

• [Doc] Update AIR examples to not use the `preprocessor` arg to Trainer

  #36757 closed Feb 25, 2025

• [Dashboard] Provide GPU stats per actor or per worker process in the node table

  #31998 closed Feb 25, 2025

• [RLlib] timesteps per minute and cpu/gpu load logging

  #25847 closed Feb 25, 2025

• [Dashboard] Confusing `workers` terminology

  #24277 closed Feb 25, 2025

• Callback URL

  #40288 closed Feb 25, 2025

• PPO centralized critic example with more than two agents？

  #40040 closed Feb 25, 2025

• Unable to read '/sys/fs/cgroup/cpuset/cpuset.cpus' through jupyter notebook

  #40351 closed Feb 25, 2025

• [Core] specifying runtime_env using conda env fullpath no work

  #50720 closed Feb 25, 2025

• custom stage name in map_batches

  #40008 closed Feb 25, 2025

• [Core] Crash using C++ API when calling a task and passing an actor handle

  #29879 closed Feb 25, 2025

• Ray

  #39855 closed Feb 25, 2025

• [RLlib] Error in executing demo example with attention net.

  #39769 closed Feb 25, 2025

• ValueError: The two structures don't have the same nested structure. / AttributeError: 'NoneType' object has no attribute 'distributions'

  #38318 closed Feb 25, 2025

• [RLlib] Error importing the library

  #39583 closed Feb 25, 2025

• [Rllib] critic reguliarized regression: wrong action clipping for critic update

  #39683 closed Feb 25, 2025

• ray.get() can't get a large-scale object store

  #39720 closed Feb 25, 2025

• [core] Flaky test `test_release_resources_race` in test_basic when Ray Client is enabled

  #39688 closed Feb 25, 2025

• [RLlib] PPO training on Atari Environment using standard hyperparameters gives poor results

  #39255 closed Feb 25, 2025

• [RLlib] Framework "tf2" raises error in `MLPEncoderConfig`

  #37413 closed Feb 25, 2025

• [core] ObjectManager::PushLocalObject reset wrong size in the ObjectInfo in local_objects_ when handling object size changes

  #39148 closed Feb 25, 2025

• [Tune] `Tuner.restore` does not fully work with WandB Callback

  #38894 closed Feb 25, 2025

• [tune] Can't move file with different disk types

  #38486 closed Feb 25, 2025

• Dreamer error: frame_skip unexpected

  #32508 closed Feb 25, 2025

• [RLlib] PPO speed and performance issues

  #29623 closed Feb 25, 2025

• [External Doc] Update `pytorch_tutorials_hyperparameter_tuning_tutorial.py` in the `pytorch/tutorials` repo

  #38956 closed Feb 25, 2025

• [tune] 'tune.utils.wait_for_gpu' typo

  #38721 closed Feb 25, 2025

• [Doc] Warning about batch size on page /data/examples/huggingface_vit_batch_prediction.html

  #38762 closed Feb 25, 2025

• [Tune] TensorflowCheckpoint cannot save subclassed Keras model because it uses legacy H5 format

  #44804 closed Feb 25, 2025

• [RFC] Ray Serve model multiplexing support

  #33253 closed Feb 25, 2025

• serve run and serve deploy for multiple users

  #42622 closed Feb 25, 2025

• [Serve] Tidy up lifecycle of gRPC context

  #42347 closed Feb 25, 2025

• [Serve] Pull out `disconnected_task` from `ProxyResponseGenerator`

  #40796 closed Feb 25, 2025

• CI test linux://python/ray/dag:tests/experimental/test_accelerated_dag is flaky

  #45922 closed Feb 24, 2025

• [core] Cover cpplint for `ray/tree/master/src/ray/rpc`

  #50647 closed Feb 23, 2025

• [core] Cover cpplint for `src/ray/pubsub`

  #50728 closed Feb 23, 2025

• Please delete

  #50835 closed Feb 22, 2025

• [core] Split giant ray core C++ targets into small ones (GCS)

  #50685 closed Feb 21, 2025

• CI test darwin://python/ray/tests:test_runtime_env_working_dir_3 is flaky

  #44765 closed Feb 21, 2025

• CI test linux://python/ray/train/v2:test_data_parallel_trainer is flaky

  #50698 closed Feb 21, 2025

• [core] Cover cpplint for ray/src/ray/scheduling

  #50679 closed Feb 21, 2025

• Minio as S3 storage

  #50762 closed Feb 21, 2025

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

  #43796 closed Feb 21, 2025

• [core] release test degradation: multi_client_put_gigabytes

  #50769 closed Feb 20, 2025

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

  #43770 closed Feb 20, 2025

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

  #43777 closed Feb 20, 2025

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

  #50724 closed Feb 20, 2025

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

  #46043 closed Feb 20, 2025

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

  #50682 closed Feb 20, 2025

• [core][compiled graphs] Ray 2.42 got RayCgraphCapacityExceeded in vLLM

  #50381 closed Feb 20, 2025

• [Core] Whether the environment variables set in runtime_env can be directly added to ray client?

  #50659 closed Feb 20, 2025

• [Core] Can the path of IGNORE_GITIGNORE be set as configurable?

  #50658 closed Feb 20, 2025

• CI test linux://python/ray/tests:test_network_failure_e2e is consistently_failing

  #49556 closed Feb 19, 2025

• Ray Data checkpoint

  #49438 closed Feb 19, 2025

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

  #44663 closed Feb 19, 2025

• [Core] Learner not respecting Catalog causes warning since 2.39

  #50690 closed Feb 19, 2025

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

  #47332 closed Feb 19, 2025

• Release test serve_microbenchmarks.aws failed

  #50606 closed Feb 19, 2025

• CI test linux://doc:source/serve/doc_code/object_detection is consistently_failing

  #49142 closed Feb 19, 2025

• CI test linux://python/ray/serve/tests/unit:test_pow_2_replica_scheduler is flaky

  #48736 closed Feb 18, 2025

• CI test linux://python/ray/workflow:tests/test_events_with_crash is consistently_failing

  #50193 closed Feb 18, 2025

36 Issues opened by 22 people

• [Data] Ordering of blocks after map and map_batches

  #50890 opened Feb 25, 2025

• [Serve] Ray Serve APIs for users to define when the Ray Serve applications are ready to serve requests

  #50883 opened Feb 25, 2025

• [core] Split object manager into smaller C++ targets

  #50882 opened Feb 25, 2025

• CI test linux://doc/source/llm/examples/batch:vllm-with-lora is consistently_failing

  #50881 opened Feb 25, 2025

• bazel-lint all BUILD files

  #50875 opened Feb 25, 2025

• [Autoscaler][V2] Updating max replicas while Pods are pending causes v2 autoscaler to hang

  #50868 opened Feb 24, 2025

• CI test darwin://python/ray/tests:test_reconstruction_2 is consistently_failing

  #50859 opened Feb 24, 2025

• Installing dependencies from pyproject.toml

  #50858 opened Feb 24, 2025

• [Dashboard] Serve Grafana panels shows metrics from multiple clusters instead of filtering on SessionName or ray_io_cluster

  #50850 opened Feb 24, 2025

• Unable to find Ray Dataset to read files from directories of directories of images from disk to create dataset during run time of training.

  #50844 opened Feb 23, 2025

• When I run ray.remote with image_uri as the parameter, it is successful, but when I change it to ray_derve, it cannot run successfully

  #50839 opened Feb 23, 2025

• AttributeError: module 'ray.train._internal.session' has no attribute '_get_session'. Did you mean: 'get_session'?

  #50828 opened Feb 22, 2025

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

  #50827 opened Feb 22, 2025

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

  #50826 opened Feb 22, 2025

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

  #50825 opened Feb 22, 2025

• [Train] Unable to gain long-term access to S3 storage for training state/checkpoints when running on AWS EKS

  #50823 opened Feb 22, 2025

• CI test linux://python/ray/air:test_tensor_extension is flaky

  #50819 opened Feb 22, 2025

• CI test linux://python/ray/air:test_object_extension is flaky

  #50818 opened Feb 22, 2025

• CI test linux://python/ray/air:test_arrow is flaky

  #50817 opened Feb 22, 2025

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

  #50814 opened Feb 21, 2025

• [Core] Handle transient network error for pushing object chunks

  #50803 opened Feb 21, 2025

• [core] Serve microbenchmarks occasionally crash with segfault or invalid memory access

  #50802 opened Feb 21, 2025

• [Data] supper passing `pyarrow.dataset.Expression`s to `Dataset.filter`'s `expr`

  #50799 opened Feb 21, 2025

• Enable Ray debug logging when installing env packages with pip/uv

  #50798 opened Feb 21, 2025

• [Data]: Categorizer fails with non uniform distributions

  #50792 opened Feb 21, 2025

• [Autoscaler] Autoscaler gets into infinite cycle of removing and adding nodes, never satisfies placement group

  #50783 opened Feb 21, 2025

• [Core] Negative available resources

  #50739 opened Feb 19, 2025

• [Core] Stale ray_cluster_<state>_nodes metrics

  #50735 opened Feb 19, 2025

• [core] Cover cpplint for `src/ray/object_manager/plasma`

  #50729 opened Feb 19, 2025

• slow torch.distributed with non-default CUDA_VISIBLE_DEVICES

  #50723 opened Feb 19, 2025

• how to solve this problem

  #50721 opened Feb 19, 2025

• [core] Fix mock dependency

  #50718 opened Feb 19, 2025

• [nsys plugin] How about add an option `name` to nsys dumped file

  #50711 opened Feb 19, 2025

• [Serve] Serve no longer retries deployments after 3 failures

  #50710 opened Feb 19, 2025

• [Serve] group requests by `model_id` in Model Multiplexing

  #50695 opened Feb 18, 2025

• Make sure precommit hook linter and CI matches

  #50694 opened Feb 18, 2025

293 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• (WIP) [core][compiled graphs] Unify code paths for NCCL P2P and collectives scheduling

  #48649 commented on Feb 24, 2025 • 54 new comments

• [core][compiled graphs]Support reduce scatter and all gather collective for GPU communicator in compiled graph

  #50624 commented on Feb 25, 2025 • 23 new comments

• [Data] Adding in per node metrics

  #49705 commented on Feb 25, 2025 • 13 new comments

• [Core][Autoscaler] Refactor v2 Log Formatting

  #49350 commented on Feb 25, 2025 • 10 new comments

• [core] Utils to cleanup cgroup folder

  #49941 commented on Feb 25, 2025 • 8 new comments

• [core][1/N] Make executor to be a long-running Python thread

  #50644 commented on Feb 24, 2025 • 7 new comments

• Various enhancements to the Gradio Ray Serve tutorial

  #50276 commented on Feb 18, 2025 • 6 new comments

• [Fix][Core] Execute user requested actor exit in C++ side

  #49918 commented on Feb 25, 2025 • 6 new comments

• [data/preprocessors] feat: allow hasher to run on append mode

  #50632 commented on Feb 25, 2025 • 4 new comments

• Discover TPU logs in Ray Dashboard

  #47737 commented on Feb 25, 2025 • 4 new comments

• [train][v2] implement state export

  #50622 commented on Feb 25, 2025 • 4 new comments

• [Feat][Dashboard] Remove StateHead's dependencies on DataSource

  #50605 commented on Feb 25, 2025 • 3 new comments

• [core] Implement dup2 wrapper

  #50439 commented on Feb 19, 2025 • 3 new comments

• Add Semi-Random Weighting to AutoScaler Node Scheduler

  #49983 commented on Feb 19, 2025 • 2 new comments

• [core] add RAY_IGNORE_VERSION_MISMATCH when ray start --address

  #50513 commented on Feb 22, 2025 • 2 new comments

• [doc][core] fix a wrong url in ray-dag.rst

  #49980 commented on Feb 25, 2025 • 2 new comments

• [Core] Split stats_metric into smaller targets to improve build performance

  #50595 commented on Feb 24, 2025 • 2 new comments

• [core] Cover cpplint for ray/src/ray/stats

  #50678 commented on Feb 25, 2025 • 2 new comments

• [data] add ClickHouse sink

  #50377 commented on Feb 19, 2025 • 1 new comment

• [Docs] Update Volcano Integration with The New Flag

  #47911 commented on Feb 25, 2025 • 1 new comment

• [bazel] move python rules up

  #47260 commented on Feb 25, 2025 • 0 new comments

• [Doc] RayServe Single-Host TPU v6e Example with vLLM

  #47814 commented on Feb 19, 2025 • 0 new comments

• [core][dashboard] Make updates to DataSource.(node_workers|core_worker_stats) on delta.

  #47186 commented on Feb 25, 2025 • 0 new comments

• [core] GcsPublisher bindings

  #47062 commented on Feb 25, 2025 • 0 new comments

• [core][dashboard] Change the StateDataSourceClient from using gRPC stub -> NewGcsClient.

  #47056 commented on Feb 25, 2025 • 0 new comments

• Bump aiohttp from 3.9.5 to 3.10.2 in /release

  #47050 commented on Feb 25, 2025 • 0 new comments

• Bump tensorflow from 2.11.0 to 2.12.1 in /rllib_contrib/simple_q

  #47001 commented on Feb 25, 2025 • 0 new comments

• [ci][core] GCS FT Chaos test

  #46996 commented on Feb 25, 2025 • 0 new comments

• Bump tensorflow from 2.11.0 to 2.12.1 in /rllib_contrib/td3

  #46978 commented on Feb 25, 2025 • 0 new comments

• Bump keras from 2.7.0 to 2.13.1 in /python/requirements/compat

  #46977 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Deprecate algo config (python) dicts; must be `AlgorithmConfig` objects.

  #46896 commented on Feb 25, 2025 • 0 new comments

• [POC] A Reactor style GCS. #1: GcsNodeManager

  #46891 commented on Feb 25, 2025 • 0 new comments

• Add docs link to Serve page of Ray Dashboard

  #46812 commented on Feb 25, 2025 • 0 new comments

• [Docs][hotfix] Correct the desc of nums of blocks

  #47741 commented on Feb 24, 2025 • 0 new comments

• [wip] revive zero copy torch tensor serialization

  #47665 commented on Feb 25, 2025 • 0 new comments

• [ADAG]Enable NPU (hccl) communication for CG

  #47658 commented on Feb 21, 2025 • 0 new comments

• Bump send and express in /python/ray/dashboard/client

  #47643 commented on Feb 25, 2025 • 0 new comments

• Bump serve-static and express in /python/ray/dashboard/client

  #47641 commented on Feb 25, 2025 • 0 new comments

• Bump body-parser and express in /python/ray/dashboard/client

  #47589 commented on Feb 25, 2025 • 0 new comments

• Bump path-to-regexp and express in /python/ray/dashboard/client

  #47588 commented on Feb 25, 2025 • 0 new comments

• [Do not merge] Run release tests for export API

  #47568 commented on Feb 25, 2025 • 0 new comments

• Improvements and Artificial Intelligence-based Improvements for Ray Cross-Language Functionality Testing

  #47499 commented on Feb 25, 2025 • 0 new comments

• [core][dashboard] make a flamegraph on event loop lag.

  #47491 commented on Feb 25, 2025 • 0 new comments

• Bump webpack from 5.76.2 to 5.94.0 in /python/ray/dashboard/client

  #47428 commented on Feb 25, 2025 • 0 new comments

• [todo] Migrate redis kv get sync

  #47348 commented on Feb 25, 2025 • 0 new comments

• [Core][aDAG] Remove busy waiting semaphore acquire in linux

  #47322 commented on Feb 25, 2025 • 0 new comments

• idempotent replies by seq_no for sequential actors.

  #47314 commented on Feb 25, 2025 • 0 new comments

• Bump micromatch from 4.0.5 to 4.0.8 in /python/ray/dashboard/client

  #47310 commented on Feb 25, 2025 • 0 new comments

• [RLlib; docs] New API stack docs: Add `ConnectorV2` documentation

  #47278 commented on Feb 24, 2025 • 0 new comments

• Update py_modules.py AttributeError: module has no attribute '__path__'

  #46302 commented on Feb 25, 2025 • 0 new comments

• [WIP] CI: jemalloc & mimalloc

  #46271 commented on Feb 25, 2025 • 0 new comments

• Pinterest/release 2.9.3+pinterest5

  #46175 commented on Feb 25, 2025 • 0 new comments

• [ADAG] Detect if ADAG is at capacity for execution

  #46158 commented on Feb 25, 2025 • 0 new comments

• Revert "[doc]Make vllm example works with latest vllm version"

  #46094 commented on Feb 25, 2025 • 0 new comments

• [RLlib] - `"Synchronized"` sampling for multi-agent buffers.

  #46083 commented on Feb 25, 2025 • 0 new comments

• [serve] fix lossy serve config

  #45938 commented on Feb 25, 2025 • 0 new comments

• Fix malformed `temp_dir` path when connecting Windows workers to cluster with Linux head

  #45930 commented on Feb 25, 2025 • 0 new comments

• [spark] Fix nvidia-smi hanging issue

  #45896 commented on Feb 25, 2025 • 0 new comments

• Xgui/test subdataset

  #45889 commented on Feb 25, 2025 • 0 new comments

• Adds new working dir upload protocol PLASMA, and use it in job submission.

  #45880 commented on Feb 25, 2025 • 0 new comments

• MADDPG framework should be TensorFlow

  #45863 commented on Feb 25, 2025 • 0 new comments

• Improve code snippet in docs to set up `ray[serve]` gRPC service

  #45862 commented on Feb 25, 2025 • 0 new comments

• [WIP] Benchmark data shuffle

  #45847 commented on Feb 25, 2025 • 0 new comments

• [Core] Add warning when uploading large working dirs

  #45818 commented on Feb 25, 2025 • 0 new comments

• release 2.9.3+pinterest4

  #45807 commented on Feb 25, 2025 • 0 new comments

• remove cleaned metadata

  #45643 commented on Feb 25, 2025 • 0 new comments

• Update rllib-env.rst

  #46750 commented on Feb 25, 2025 • 0 new comments

• Introducing StaleTaskError

  #46705 commented on Feb 25, 2025 • 0 new comments

• [Core] If possible, force flush the trace when the worker ends.

  #46654 commented on Feb 25, 2025 • 0 new comments

• [dashboard] Place the submit job on a separate page

  #46613 commented on Feb 25, 2025 • 0 new comments

• [ADAG] Fix DAG input

  #46604 commented on Feb 25, 2025 • 0 new comments

• [util] remove pygloo support

  #46590 commented on Feb 25, 2025 • 0 new comments

• Fix mlflow artifact logging

  #46570 commented on Feb 25, 2025 • 0 new comments

• [dont review] release 2.10.0+pinterest1

  #46551 commented on Feb 25, 2025 • 0 new comments

• [Core]Fix the issue of actor tasks hanging during resubmission

  #46539 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Optimize rnn_sequencing performance

  #46502 commented on Feb 25, 2025 • 0 new comments

• fix performance bug in arrow to numpy transform

  #46433 commented on Feb 25, 2025 • 0 new comments

• avoid merge errors when blocks contain different type in DelegatingBl…

  #46407 commented on Feb 25, 2025 • 0 new comments

• [Core] Add ray-start option 'session-name'

  #46404 commented on Feb 25, 2025 • 0 new comments

• [test] cpp20

  #46380 commented on Feb 25, 2025 • 0 new comments

• [Docker] Upgrade base deps docker python env to 3.9.7

  #46353 commented on Feb 25, 2025 • 0 new comments

• fixed a typo in ValueError message for contains_tensor

  #46348 commented on Feb 25, 2025 • 0 new comments

• [Doc] Update directory path for installation

  #46318 commented on Feb 25, 2025 • 0 new comments

• Update dyn-req-batch.md with style edits

  #49725 commented on Feb 25, 2025 • 0 new comments

• changes to get ray serve responding on REST API calls when distribute…

  #49730 commented on Feb 25, 2025 • 0 new comments

• [Ray Data Optimization] Remove block merging

  #49743 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Flatten dict-typed observations before comparing them.

  #49758 commented on Feb 25, 2025 • 0 new comments

• [KubeRay] support suspending worker groups in KubeRay autoscaler

  #49768 commented on Feb 25, 2025 • 0 new comments

• [ci] remove pins in runtime_env usage in train examples

  #49772 commented on Feb 25, 2025 • 0 new comments

• Pass checkpointable args through in tf_learner

  #49861 commented on Feb 24, 2025 • 0 new comments

• [dashboard] Add SubprocessModules to the Dashboard routes, and convert HealthzHead.

  #49864 commented on Feb 25, 2025 • 0 new comments

• Fix complex (dict) observation concatenation in single agent episode

  #49913 commented on Feb 24, 2025 • 0 new comments

• Fix Databricks host URL handling in Ray Data

  #49926 commented on Feb 24, 2025 • 0 new comments

• [RFC][dashboard] Use aiohttp client for inter dependencies.

  #49932 commented on Feb 24, 2025 • 0 new comments

• [core] minor optimization for JoinPaths

  #49946 commented on Feb 24, 2025 • 0 new comments

• [RLlib; Offline] - Add single learner gpu training with preloading in `OfflinePreLearner`.

  #49960 commented on Feb 24, 2025 • 0 new comments

• Explicit comm

  #49979 commented on Feb 24, 2025 • 0 new comments

• [kuberay] fix deserialisation of custom resources in autoscaler config

  #49993 commented on Feb 24, 2025 • 0 new comments

• [dashboard] Remove the dashboard grpc server.

  #50021 commented on Feb 24, 2025 • 0 new comments

• [WIP] Move execution loop to the same thread as the constructor of an actor

  #50032 commented on Feb 24, 2025 • 0 new comments

• [core] Upgrade boost

  #50039 commented on Feb 24, 2025 • 0 new comments

• tsan

  #50105 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Disable callbacks callable check for new api stack

  #50157 commented on Feb 22, 2025 • 0 new comments

• fix: WandbLogger crashing silently on a FileNotFoundError

  #50308 commented on Feb 19, 2025 • 0 new comments

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

  #50414 commented on Feb 18, 2025 • 0 new comments

• [tune] Remove loguniform's base

  #50415 commented on Feb 20, 2025 • 0 new comments

• [agent] Ray metrics use the name parsing as Prometheus

  #50443 commented on Feb 25, 2025 • 0 new comments

• [Core][Dashboard] Convert JobHead to subprocess module

  #50483 commented on Feb 20, 2025 • 0 new comments

• [core] Don't build cpp api on pip install

  #50499 commented on Feb 20, 2025 • 0 new comments

• Various enhancements to Tune Keras example:

  #50581 commented on Feb 18, 2025 • 0 new comments

• [RLlib] Enable spliting and zero padding of Dict observation

  #50589 commented on Feb 21, 2025 • 0 new comments

• Integrate Ray Dataset with Daft Dataframe

  #50630 commented on Feb 20, 2025 • 0 new comments

• [WIP; RLlib] APPO accelerate (vol 17): `LearnerGroup` should not pickle remote functions on each update-call; Refactor `LearnerGroup` and `Learner` APIs.

  #50665 commented on Feb 24, 2025 • 0 new comments

• [WIP / try out] Use UV for Python 3.13 tests

  #50669 commented on Feb 25, 2025 • 0 new comments

• [core] Improve/Fix plasma object store wait logic

  #50680 commented on Feb 24, 2025 • 0 new comments

• Add new serve autoscaling parameter `scaling_function`

  #47837 commented on Feb 20, 2025 • 0 new comments

• (WIP) [ADAG] Support dag.experimental_compile(_custom_nccl_group= nccl_group) in aDAG

  #47987 commented on Feb 25, 2025 • 0 new comments

• [doc] Remove unused/unmaintained `doc/source/templates` folder

  #48295 commented on Feb 25, 2025 • 0 new comments

• [doc] fix: Typo and missing import in doc

  #48311 commented on Feb 20, 2025 • 0 new comments

• Fix invalid type for progress_reporter parameter of RunConfig

  #48439 commented on Feb 25, 2025 • 0 new comments

• remove redundant bazel dependencies

  #48464 commented on Feb 25, 2025 • 0 new comments

• Sjoshi/push manager patch

  #48475 commented on Feb 25, 2025 • 0 new comments

• [data] add opensearch datasource

  #48555 commented on Feb 24, 2025 • 0 new comments

• [Fix][GCS] Implement reconnection for RedisContext

  #48781 commented on Feb 25, 2025 • 0 new comments

• [Build][Deps] Add new `ray[azure]` extra package

  #48847 commented on Feb 25, 2025 • 0 new comments

• [train] Make dataset argument covariant

  #48999 commented on Feb 24, 2025 • 0 new comments

• [Jobs] Add metric to track duration of jobs

  #49035 commented on Feb 25, 2025 • 0 new comments

• [data] fix nodeName When the network in KubeRay is set to hostnetwork

  #49188 commented on Feb 25, 2025 • 0 new comments

• [Dashboard] stop ray submmited job through ui

  #49201 commented on Feb 25, 2025 • 0 new comments

• Fix unpacking zip package treats "../" as the top_level_directory

  #49204 commented on Feb 25, 2025 • 0 new comments

• [wandb] Use wandb Run as a context manager

  #49307 commented on Feb 25, 2025 • 0 new comments

• [Core] fail to download s3 py modules

  #49332 commented on Feb 19, 2025 • 0 new comments

• [Serve] Improve serve deploy ignore behavior

  #49336 commented on Feb 25, 2025 • 0 new comments

• Update tune-search-spaces.rst to correct outdated api use

  #49386 commented on Feb 25, 2025 • 0 new comments

• [core][compiled graphs] Support reduce scatter and all gather collective in compiled graph

  #49404 commented on Feb 25, 2025 • 0 new comments

• [Core] Persist the Driver Console Log When Job Execution Not Through Job API

  #49452 commented on Feb 25, 2025 • 0 new comments

• [data] add dataloader for lance datasource

  #49459 commented on Feb 25, 2025 • 0 new comments

• [core][cgraph] Use threadpool and one io_context for mutable object provider

  #49500 commented on Feb 25, 2025 • 0 new comments

• [ci] Remove redundant ML doctests from running in unit test pipelines

  #49516 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Add NPU and HPU support to RLlib

  #49535 commented on Feb 25, 2025 • 0 new comments

• [core] Minor improvements to core worker get

  #49567 commented on Feb 25, 2025 • 0 new comments

• [core] Move observable store client logic into in memory store

  #49570 commented on Feb 25, 2025 • 0 new comments

• [core] Don't get dashboard address after each dashboard connection failure

  #49584 commented on Feb 25, 2025 • 0 new comments

• [Core] Streaming generator supports num_returns

  #49586 commented on Feb 20, 2025 • 0 new comments

• [Dashboard] Support multiple accelerator monitoring and flexible display

  #49610 commented on Feb 25, 2025 • 0 new comments

• [Core] Add virtual cluster

  #49717 commented on Feb 24, 2025 • 0 new comments

• [RLlib] Fix broken stats accumulation for 'MeanStdFilter' connector.

  #49718 commented on Feb 25, 2025 • 0 new comments

• [Distributed Debugger] Newly added breakpoint not works: Breakpoint in file that does not exist

  #48778 commented on Feb 23, 2025 • 0 new comments

• [vLLM] The ray serve using vLLM example on the website does not work.

  #50275 commented on Feb 22, 2025 • 0 new comments

• Test issue (please ignore) - more text then even more text

  #49867 commented on Feb 22, 2025 • 0 new comments

• [Ray Serve] Expose public interface for user to customize the router

  #50465 commented on Feb 21, 2025 • 0 new comments

• [tune] Can't reload a past experiment (pickling error in pyarrow?)

  #46740 commented on Feb 21, 2025 • 0 new comments

• [Jobs] Add metric to track duration of jobs

  #48962 commented on Feb 21, 2025 • 0 new comments

• [RLlib] Adding transformer models

  #50648 commented on Feb 21, 2025 • 0 new comments

• [RLlib] [Solution Found] Action masking example not working when LSTM enabled.

  #50526 commented on Feb 21, 2025 • 0 new comments

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

  #50340 commented on Feb 21, 2025 • 0 new comments

• Cannot Install ray[rllib] on Python 3.13

  #50226 commented on Feb 21, 2025 • 0 new comments

• [RLlib] MetricsLogger API problem with series-based data

  #50294 commented on Feb 21, 2025 • 0 new comments

• Release test rllib_learning_tests_pong_appo_torch.aws failed

  #50217 commented on Feb 21, 2025 • 0 new comments

• [RLlib] PPO algorithm can't be trained from checkpoint

  #50136 commented on Feb 21, 2025 • 0 new comments

• [RLlib] Callbacks class input check is invalid for new api stack

  #50135 commented on Feb 21, 2025 • 0 new comments

• [RLlib] "TypeError: 'int' object is not iterable when using from_jsonable with nested Discrete in Dict"

  #50131 commented on Feb 21, 2025 • 0 new comments

• [RLlib] TFLearner does not correctly implement `restore_from_path`

  #49860 commented on Feb 21, 2025 • 0 new comments

• [RFC] LLM APIs for Ray Data and Ray Serve

  #50639 commented on Feb 20, 2025 • 0 new comments

• [Core] Python 3.13 wheel

  #49738 commented on Feb 20, 2025 • 0 new comments

• [AIR: TransformersTrainer] Local rank detection from Deepspeed conflicts with Ray multi-node provision

  #37212 commented on Feb 24, 2025 • 0 new comments

• `gcp_cluster_launcher_full` release test failure

  #37276 commented on Feb 24, 2025 • 0 new comments

• [core] Dedup session name/cluster ID

  #37546 commented on Feb 24, 2025 • 0 new comments

• Object error when running /ray-air/examples/torch_image_example.html on cluster

  #37623 commented on Feb 24, 2025 • 0 new comments

• [Tune] Unset signal catching event after tune.run() finished.

  #37737 commented on Feb 24, 2025 • 0 new comments

• [AIR] Multiple API Changes in 2.7

  #37868 commented on Feb 24, 2025 • 0 new comments

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

  #37910 commented on Feb 24, 2025 • 0 new comments

• page /serve/tutorials/streaming.html adding Token Streaming LLM example for serve

  #38094 commented on Feb 24, 2025 • 0 new comments

• [core] Recent windows test flakiness

  #38413 commented on Feb 24, 2025 • 0 new comments

• Registering & Using OpenAI Gym environments from the RoboHive API in Ray RLlib & Tune

  #37231 commented on Feb 24, 2025 • 0 new comments

• [Core|Data] read_csv(): Exception from as task of operator "ReadCSV->SplitBlocks(100)"

  #47839 commented on Feb 24, 2025 • 0 new comments

• Unexpected node deaths cannot be recovered from checkpoints

  #46814 commented on Feb 24, 2025 • 0 new comments

• [aDAG] Support torch profiling with configurable parameters

  #47745 commented on Feb 24, 2025 • 0 new comments

• [environments] Support PEX executables

  #15518 commented on Feb 24, 2025 • 0 new comments

• Make infeasible tasks error much more obvious

  #45909 commented on Feb 24, 2025 • 0 new comments

• [Ray Complied Graph] NCCL Internal Error

  #49827 commented on Feb 24, 2025 • 0 new comments

• [RLlib] Unable to replicate original PPO performance

  #45655 commented on Feb 24, 2025 • 0 new comments

• [Core] Plugable storage backend besides Redis

  #50656 commented on Feb 24, 2025 • 0 new comments

• Fatal Python error: Segmentation fault

  #49998 commented on Feb 20, 2025 • 0 new comments

• [Core] Spot preemption related retries do not count towards the max retries

  #50640 commented on Feb 19, 2025 • 0 new comments

• RAY_IGNORE_VERSION_MISMATCH=True has no effect on ray start --addr.

  #50511 commented on Feb 19, 2025 • 0 new comments

• Ray's execution of the command "start --head" does not perform deduplication.

  #50510 commented on Feb 19, 2025 • 0 new comments

• Exception: The current node timed out during startup. This could happen because some of the Ray processes failed to startup.

  #50474 commented on Feb 19, 2025 • 0 new comments

• Compiled Graphs torch.Tensor serialization device

  #50452 commented on Feb 19, 2025 • 0 new comments

• Fatal Python error: Floating point exception when running on H20

  #50418 commented on Feb 19, 2025 • 0 new comments

• [StateAPI] StateAPI request truncates recent elements

  #50378 commented on Feb 19, 2025 • 0 new comments

• 【bug】Ray.data.write_parquet will write twice when use fsspec local filesystem

  #49741 commented on Feb 19, 2025 • 0 new comments

• [data] sort large dataset by ray.data.Dataset always fail

  #49679 commented on Feb 19, 2025 • 0 new comments

• [Data, Train] ray::SplitCoordinator is very slow at every epoch + takes up too much memory

  #49190 commented on Feb 19, 2025 • 0 new comments

• [Data]Fuse operator

  #49587 commented on Feb 19, 2025 • 0 new comments

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

  #47465 commented on Feb 19, 2025 • 0 new comments

• [Agent] Make sure Ray metrics perform same validation as Prometheus to weed out invalid names early on

  #40586 commented on Feb 19, 2025 • 0 new comments

• [Air] WandB logger / _WandbLoggingActor crashes silently when logging a video with relative path or if it cannot find the given file.

  #50307 commented on Feb 19, 2025 • 0 new comments

• [core][logging] Customizable Python standard log attributes.

  #49502 commented on Feb 19, 2025 • 0 new comments

• [core] Add a util function to initialize NCCL communicator

  #50681 commented on Feb 19, 2025 • 0 new comments

• trial_name_creator does not change name in get_dataframe()

  #50635 commented on Feb 19, 2025 • 0 new comments

• [Data] ray.data.from_torch fails on datasets with variable shaped images

  #50229 commented on Feb 18, 2025 • 0 new comments

• [RLlib] 'MeanStdFilter' connector is broken.

  #49716 commented on Feb 20, 2025 • 0 new comments

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

  #49609 commented on Feb 20, 2025 • 0 new comments

• The mistake of using multi-head networks

  #49582 commented on Feb 20, 2025 • 0 new comments

• [RLlib][Windows] Windows Invalid Directory Name Error in Ray RLlib

  #49477 commented on Feb 20, 2025 • 0 new comments

• [RLlib|Custom Policy]Custom Policy Implementation in Reinforcement Learning

  #49334 commented on Feb 20, 2025 • 0 new comments

• [usability][Feature] Throw error message if resolved ip address doesn't match the localhost

  #19052 commented on Feb 20, 2025 • 0 new comments

• [Core] Getting node id for usage in NodeAffinitySchedulingStrategy

  #28195 commented on Feb 20, 2025 • 0 new comments

• [Core] Remove pg resource notation from user-facing APIs.

  #31064 commented on Feb 20, 2025 • 0 new comments

• [Core] classmethod support for actors

  #36986 commented on Feb 20, 2025 • 0 new comments

• [core] Check if a ray task has errored without calling `ray.get` on it

  #45229 commented on Feb 20, 2025 • 0 new comments

• [Serve] Use concurrency group

  #20054 commented on Feb 20, 2025 • 0 new comments

• [Clusters][Azure] Custom ARM template for Azure Clusters

  #50684 commented on Feb 20, 2025 • 0 new comments

• [Core] Too many threads in ray worker

  #36936 commented on Feb 20, 2025 • 0 new comments

• [Ray debugger] Unable to use debugger on Ray Cluster on k8s

  #45541 commented on Feb 20, 2025 • 0 new comments

• [Core] Please provide better message where 'RuntimeError: Failed to unpickle serialized exception'

  #49885 commented on Feb 20, 2025 • 0 new comments

• [core] Cover cpplint for ray/src/ray/raylet

  #50687 commented on Feb 20, 2025 • 0 new comments

• [Ray Core] Task/resource not properly reclaimed after ray job exception or stopped

  #47531 commented on Feb 20, 2025 • 0 new comments

• [Core] Why does the statistical information of node report that the message is too large?

  #50661 commented on Feb 20, 2025 • 0 new comments

• [WIP] poc / hack relpath

  #45003 commented on Feb 25, 2025 • 0 new comments

• [testing] cumulative pinterest upgrades for 2.10

  #45000 commented on Feb 25, 2025 • 0 new comments

• add more execution and iteration metrics to prometheus

  #44971 commented on Feb 25, 2025 • 0 new comments

• [Core] Profile Ray start

  #44818 commented on Feb 25, 2025 • 0 new comments

• Update bert.ipynb

  #44455 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Cleanup `examples` folder 02: Add shared value function example script for MultiAgentRLModule.

  #44421 commented on Feb 25, 2025 • 0 new comments

• change naming to intel gaudi habana for ray train example

  #44412 commented on Feb 25, 2025 • 0 new comments

• [misc] Reformat RLLib BUILD files

  #44153 commented on Feb 25, 2025 • 0 new comments

• [misc] Reformat train/tune BUILD files

  #44151 commented on Feb 25, 2025 • 0 new comments

• verify windows wheels.

  #43442 commented on Feb 25, 2025 • 0 new comments

• [docs][clusters] Improve instructions for GPU autodetection and manual cluster launching

  #43219 commented on Feb 25, 2025 • 0 new comments

• [RLlib] ConnectorV2 API: Add heuristic action logits mixin example script.

  #43107 commented on Feb 25, 2025 • 0 new comments

• [docs][Serve] add text about pip-pack installation

  #41088 commented on Feb 25, 2025 • 0 new comments

• [docs] Documentation fixes (logging and profiling)

  #40915 commented on Feb 25, 2025 • 0 new comments

• [docs] [train] copy edits to E2E ingest example

  #40176 commented on Feb 25, 2025 • 0 new comments

• [Logging] Fix Deduplication URL

  #39830 commented on Feb 25, 2025 • 0 new comments

• Upgrade default AWS DLAMI

  #39721 commented on Feb 25, 2025 • 0 new comments

• [ci] remove is_automated_build in setup.py

  #36547 commented on Feb 25, 2025 • 0 new comments

• Unpin setproctitile

  #45640 commented on Feb 25, 2025 • 0 new comments

• [core][2/2] Kill worker on root detached actor died.

  #45638 commented on Feb 25, 2025 • 0 new comments

• [core][1/2] Add SubscribeAllActors to GcsClient.

  #45637 commented on Feb 25, 2025 • 0 new comments

• [core] Eagerly kill idle workers on job finish.

  #45633 commented on Feb 25, 2025 • 0 new comments

• Create a singleton io context and thread, and standalone gcs client on it.

  #45524 commented on Feb 25, 2025 • 0 new comments

• [train] Update Torch default timeout_s to use Torch's default timeout

  #45501 commented on Feb 25, 2025 • 0 new comments

• blind try on ubuntu upgrade ..

  #45427 commented on Feb 25, 2025 • 0 new comments

• [Data] Allow configuration of MAX_IMAGE_PIXELS in ImageDatasource

  #45415 commented on Feb 25, 2025 • 0 new comments

• add ray debugger references to ray docs

  #45414 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Enhance callbacks test case for EnvRunners; Add (optional) explicit `enable_multi_agent` setting to AlgorithmConfig.

  #45385 commented on Feb 25, 2025 • 0 new comments

• [POC][core] GcsClient async binding, aka remove PythonGcsClient.

  #45289 commented on Feb 25, 2025 • 0 new comments

• [core] Change all object_size to uint64_t and use 0 for unknown. Also adds a method `ray.experimental.get_local_object_locations`

  #45247 commented on Feb 25, 2025 • 0 new comments

• [RFC] Splitted Dashboard Heads.

  #45175 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Fix async (multiprocessing) gymnasium vector envs in `SingleAgentEnvRunner`.

  #45144 commented on Feb 25, 2025 • 0 new comments

• [RLlib; Tune] Fix default behavior of default tune `CLIReporter` (based on `Algorithm._progress_metrics`).

  #45122 commented on Feb 25, 2025 • 0 new comments

• [wip][train][tune] handle s3fs permissions

  #45100 commented on Feb 25, 2025 • 0 new comments

• Add roundtrip (ping-pong) microbenchmarks for accelerated DAG channels

  #45064 commented on Feb 25, 2025 • 0 new comments

• [WIP] add env var to enable debug

  #45009 commented on Feb 25, 2025 • 0 new comments

• lonnie's workspace

  #36406 commented on Feb 25, 2025 • 0 new comments

• [Feature] Remote call timeout required.

  #18916 commented on Feb 25, 2025 • 0 new comments

• [Dashboard] Explain Disk usage for KubeRay

  #36362 commented on Feb 25, 2025 • 0 new comments

• [Doc] Broken links due to the o11y refactoring

  #36330 commented on Feb 25, 2025 • 0 new comments

• [Train] Provide a list of models for people to choose from in the HF transformer example

  #36837 commented on Feb 25, 2025 • 0 new comments

• [Serve| Observability] Show the duration of each request

  #36633 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Deprecate methods/configs for RLlib 2.0

  #36592 commented on Feb 24, 2025 • 0 new comments

• Tests are not being skipped when they fail more than 5 times in a row.

  #36499 commented on Feb 24, 2025 • 0 new comments

• [data] Add migration guide for DataConfig

  #36668 commented on Feb 24, 2025 • 0 new comments

• [Workflow] Logger can't handle PyTorch Lightning progress bar

  #33106 commented on Feb 24, 2025 • 0 new comments

• Rllib - is_training never true

  #35833 commented on Feb 24, 2025 • 0 new comments

• [Serve| Observability] The "Logs" tab is very overwhelming

  #36632 commented on Feb 24, 2025 • 0 new comments

• [runtime env] make the `working_dir` field more flexible

  #31588 commented on Feb 24, 2025 • 0 new comments

• GradioIngress do not support Blocks.queue()

  #36977 commented on Feb 24, 2025 • 0 new comments

• random.sample

  #35245 commented on Feb 24, 2025 • 0 new comments

• Issue on page /ray-core/objects/object-spilling.html

  #37002 commented on Feb 24, 2025 • 0 new comments

• Issue on page /cluster/vms/user-guides/community/slurm.html

  #35711 commented on Feb 24, 2025 • 0 new comments

• [Core] [RLlib] single machine inside kubernetes/docker hangs at ray init

  #37144 commented on Feb 24, 2025 • 0 new comments

• CheckpointConfig does not work on Windows

  #37226 commented on Feb 24, 2025 • 0 new comments

• [serve] Make latency buckets configurable

  #38223 commented on Feb 25, 2025 • 0 new comments

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

  #48859 commented on Feb 25, 2025 • 0 new comments

• [<Ray component: Serve>] Support strawberry-graphql in serve with FastAPI

  #50677 commented on Feb 25, 2025 • 0 new comments

• [Ray Core] After the ray job is finished, it will stably trigger resource leakage

  #49999 commented on Feb 25, 2025 • 0 new comments

• [Core] ray.actor.exit_actor() does not seem to work from within an async background thread

  #49451 commented on Feb 25, 2025 • 0 new comments

• [Epic][CI] Migrate linter to Ruff

  #47991 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Attribute error when trying to compute action after training Multi Agent PPO with New API Stack

  #44475 commented on Feb 25, 2025 • 0 new comments

• [core] Cover cpplint for all C++ folders

  #50583 commented on Feb 25, 2025 • 0 new comments

• [Dashboard] Update speedscope from 1.5.3 to the latest 1.13.0

  #23118 commented on Feb 25, 2025 • 0 new comments

• [KubeRay] Documentation of using custom docker images with KubeRay needs improvements

  #31641 commented on Feb 25, 2025 • 0 new comments

• [Core] Collecting and monitoring metrics documentation should reflect different possible grafana.ini file paths

  #42430 commented on Feb 25, 2025 • 0 new comments

• [core] Split giant ray core C++ targets into small ones

  #50586 commented on Feb 25, 2025 • 0 new comments

• [core][runtime env] Remove code_search_path for job config and unify this functionality in Runtime Env

  #26784 commented on Feb 25, 2025 • 0 new comments

• [serve] Status should switch from UPDATING to UPSCALING when in [min, max_replica] range

  #41038 commented on Feb 25, 2025 • 0 new comments

• [RLlib] Moving all Envs to have the c'tor signature

  #21611 commented on Feb 25, 2025 • 0 new comments

• [RFC] More programmable API that has same output as `ray memory` command

  #13792 commented on Feb 25, 2025 • 0 new comments

• [Rllib] InvalidArgumentError: cannot compute ConcatV2 as input #1(zero-based) was expected to be a double tensor but is a float tensor

  #36364 commented on Feb 25, 2025 • 0 new comments

• How to get started with the ray docker image?

  #36533 commented on Feb 25, 2025 • 0 new comments