All Versions
25
Latest Version
Avg Release Cycle
31 days
Latest Release
1263 days ago
Changelog History
Page 3
Changelog History
Page 3
-
v0.6.5 Changes
March 25, 2019Core
- ๐ท Build system fully converted to Bazel. #4284, #4280, #4281
- Introduce a set data structure in the GCS. #4199
- ๐ Make all arguments to
_remote()
optional. #4305 - ๐ Improve object transfer latency by setting
TCP_NODELAY
on all TCP connections. #4318 - โ Add beginning of experimental serving module. #4095
- โ Remove Jupyter notebook based UI. #4301
- โ Add
ray timeline
command line command for dumping Chrome trace. #4239
Tune
- โ Add custom field for serializations. #4237
- Begin adding Tune CLI. #3983, #4321, #4322
- โ Add optimization to reuse actors. #4218
- โ Add warnings if the Tune event loop gets clogged. #4353
- Switch preferred API from
tune.run_experiments
totune.run
. #4234 - ๐ฒ Make the logging from the function API consistent and predictable. #4011
RLlib
- ๐ฅ Breaking: Flip sign of entropy coefficient in A2C and Impala. #4374
- โ Add option to continue training even if some workers crash. #4376
- โ Add asynchronous remote workers. #4253
- โ Add callback accessor for raw observations. #4212
Java
- ๐ Improve single-process mode. #4245, #4265
- ๐ฆ Package native dependencies into jar. #4367
- ๐ Initial support for calling Python functions from Java. #4166
Autoscaler
- โช Restore error messages for setup errors. #4388
Known Issues
- Object broadcasts on large clusters are inefficient. #2945
-
v0.6.4 Changes
March 06, 2019๐ฅ Breaking
- Removed
redirect_output
andredirect_worker_output
fromray.init
, removed deprecated_submit
method. #4025 - ๐ Move
TensorFlowVariables
toray.experimental.tf_utils
. #4145
Core
- 0๏ธโฃ Stream worker logging statements to driver by default. #3892
- โ Added experimental ray signaling mechanism, see the documentation. #3624
- ๐ท Make Bazel the default build system. #3898
- Preliminary experimental streaming API for Python. #4126
- โ Added web dashboard for monitoring node resource usage. #4066
- ๐ Improved propagation of backend errors to user. #4039
- Many improvements for the Java frontend. #3687, #3978, #4014, #3943, #3839, #4038, #4039, #4063, #4100, #4179, #4178
- ๐ Support for dataclass serialization. #3964
- Implement actor checkpointing. #3839
- First steps toward cross-language invocations. #3675
- ๐ Better defaults for Redis memory usage. #4152
Tune
- ๐ฅ Breaking : Introduce ability to turn off default logging. Deprecates custom_loggers. #4104
- ๐ Support custom resources. #2979
- โ Add initial parameter suggestions for HyperOpt. #3944
- โ Add scipy-optimize to Tune. #3924
- โ Add Nevergrad. #3985
- โ Add number of trials to the trial runner logger. #4068
- ๐ Support RESTful API for the webserver. #4080
- ๐ Local mode support. #4138
- Dynamic resources for trials. #3974
RLlib
- Basic infrastructure for off-policy estimation. #3941
- โ Add simplex action space and Dirichlet action distribution. #4070
- Exploration with parameter space noise. #4048
- Custom supervised loss API. #4083
- โ Add torch policy gradient implementation. #3857
Autoscaler and Cluster Setup
- โ Add docker run option (e.g. to support nvidia-docker). #3921
Modin
- ๐ Upgrade Modin to 0.3.1, see the release notes. #4058
Known Issues
- Removed
-
v0.6.3 Changes
March 06, 2019Core
- ๐ Initial work on porting the build system to Bazel. #3918, #3806, #3867, #3842
- ๐ Allow starting Ray processes inside valgrind, gdb, tmux. #3824, #3847
- ๐ Stability improvements and bug fixes. #3861, #3962, #3958, #3855, #3736, #3822, #3821, #3925
- Convert Python C extensions to Cython. #3541
- ๐ท
ray start
can now be used to start Java workers. #3838, #3852 - ๐ Enable LZ4 compression in
pyarrow
build. #3931 - โก๏ธ Update Redis to version 5.0.3. #3886
- ๐ Use one memory-mapped file for Plasma store. #3871,
Tune
- ๐ Support for BayesOpt. #3864
- ๐ Support for SigOpt. #3844
- ๐ Support executing infinite recovery retries for a trial. #3901
- ๐ Support
export_formats
option to export policy graphs. #3868 - ๐ฒ Cluster and logging improvements. #3906
RLlib
- ๐ Support for Asynchronous Proximal Policy Optimization (APPO). #3779
- ๐ Support for MARWIL. #3635
- ๐ Support for evaluation option in DQN. #3835
- ๐ Bug fixes. #3865, #3810, #3938
- Annotations for API stability. #3808
Autoscaler and Cluster Setup
- โก๏ธ Faster cluster launch and update. #3720
- ๐ Bug fixes. #3916, #3860, #3937, #3782, #3969
- ๐ง Kubernetes configuration improvements. #3875, #3909
Modin
- โก๏ธ Update Modin to 0.3.0. #3936
Known Issues
- Object broadcasts on large clusters are inefficient. #2945
-
v0.6.2 Changes
January 17, 2019๐ฅ Breaking Changes
- โฑ The
timeout
argument ofray.wait
now uses seconds instead of milliseconds. #3706
Core
- 0๏ธโฃ Limit default redis max memory to 10GB. #3630
- Define a
Node
class to manage Ray processes. #3733 - Garbage collection of actor dummy objects. #3593
- Split profile table among many keys in the GCS. #3676
- ๐ณ Automatically try to figure out the memory limit in a docker container. #3605
- ๐ Improve multi-threading support. #3672
- โ Push a warning to all users when large number of workers have been started. #3645
- ๐จ Refactor code
ray.ObjectID
code. #3674
Tune
- ๐ Change log handling for Tune. #3661
- ๐ Tune now supports resuming from cluster failure. #3309, #3725, #3657, #3681
- ๐ Support Configuration Merging for Suggestion Algorithms. #3584
- ๐ Support nested PBT mutations. #3455
RLlib
- โ Add starcraft multiagent env as example. #3542
- ๐ Allow development without needing to compile Ray. #3623
- ๐ Documentation for I/O API and multi-agent improvements. #3650
- Export policy model checkpoint. #3637
- ๐จ Refactor PyTorch custom model support. #3634
Autoscaler
- โ Add an initial_workers option. #3530
- โ Add kill and get IP commands to CLI for testing. #3731
- ๐ง GCP allow manual network configuration. #3748
Known Issues:
- Object broadcasts on large clusters are inefficient. #2945
- โฑ The
-
v0.6.1 Changes
December 24, 2018Core
- โ Added experimental option to limit Redis memory usage. #3499
- โ Added option for restarting failed actors. #3332
- ๐ Fixed Plasma TensorFlow operator memory leak. #3448
- ๐ Fixed compatibility issue with TensorFlow and PyTorch. #3574
- ๐จ Miscellaneous code refactoring and cleanup. #3563 #3564 #3461 #3511
- ๐ Documentation. #3427 #3535 #3138
- Several stability improvements. #3592 #3597
RLlib
- ๐ Multi-GPU support for Multi-agent PPO. #3479
- Unclipped actions are sent to learner. #3496
rllib rollout
now also preprocesses observations. #3512- Basic Offline Data API added. #3473
- ๐ Improvements to metrics reporting in DQN. #3491
- AsyncSampler no longer auto-concats. #3556
- QMIX Implementation (Experimental). #3548
- ๐ IMPALA performance improvements. #3402
- ๐ Better error messages. #3444
- ๐ PPO performance improvements. #3552
Autoscaler
Ray Tune
- Lambdas now require
tune.function
wrapper. #3457 - ๐ Custom loggers, sync functions, and trial names are now supported. #3465
- ๐ Improvements to fault tolerance. #3414
- ๐ Variant Generator docs clarification. #3583
trial_resources
now renamed toresources_per_trial
. #3580
Modin
- Modin 0.2.5 is now bundled with Ray
import modin
afterimport ray
- Modin 0.2.5 release notes
- ๐ Greater than memory support for object store. #3450
Known Issues
- Object broadcasts on large clusters are inefficient. #2945