Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.
COMMITS
March 26, 2026
K
[core] Fix negative RUNNING task metric (#62070)
Kartica Modi committed
S
Support dual compute config schemas in validation (#62013)
Sai Miduthuri committed
L
[Data] Support rapidsmpf 26.2 in gpu shuffle (#62062)
Lee Yang committed
L
[Serve][LLM] SGLangServer: Multi-GPU Deployment (TP/PP Support) fix (#61201)
Limark Dcunha committed
N
[data][llm] Fix trust remote code download (#60344)
Nikhil G committed
A
[Data] Fixing `StreamingSplitDataIterator.schema()` (#62057)
Alexey Kudinkin committed
March 25, 2026
A
[Data] Fixed `ParquetDatasource` handling of `FileSystemFactory.inspect` (#62065)
Alexey Kudinkin committed
A
[CI] Upgrade rayci version from 0.31.0 to 0.36.0 (#61876)
Andrew Pollack-Gray committed
H
B
[Data] Remove old instance-based execution callback API (#62055)
Balaji Veeramani committed
S
Add compute config schema constants and cloud name utilities (#62012)
Sai Miduthuri committed
J
[llm] Promote serve LLM APIs to beta (#62054)
Jeffrey Wang committed
M
[train] Add missing `@pytest.mark.asyncio` to async def tests (#61849)
Mark Towers committed
B
[Data] Remove `get_plan_conversion_fns` (#62050)
Balaji Veeramani committed
B
[Data] Fix planner tests to handle tuple return from planner.plan() (#62053)
Balaji Veeramani committed
E
[raydepsets] adding errors for invalid configs (#62051)
Elliot Barnwell committed
A
[Data] Fixed scheduling metrics computation (#62031)
Alexey Kudinkin committed
L
[Data] Make execution callback dataflow explicit to prevent state leakage (#61405)
Limark Dcunha committed
J
[llm][deps] Upgrade vLLM to 0.18.0 (#61952)
Jeffrey Wang committed
M
[train] Fix StateManagerCallback to accept datasets explicitly (#62042)
matthewdeng committed
A
[serve] Add env var to specify haproxy lb algo (#61988)
akyang-anyscale committed
B
[Data] Remove dead `Benchmark` methods (#61884)
Balaji Veeramani committed
B
[Data] Remove dead release tests (#61883)
Balaji Veeramani committed
I
[data] Return None when no outputs have been produced (#62029)
iamjustinhsu committed
Y
[Data] Fix verbose autoscaler request log output (#61989)
You-Cheng Lin committed
H
[Data] Revert ActorMethod Reference Cycle Workaround (#62036)
HFFuture committed
March 24, 2026
I
[data] pyrefly fix lance table_id (#62034)
iamjustinhsu committed
H
[Data] Add heterogeneous memory batch inference release test (#61899)
Hao Chen committed
S
Add anyscale_sdk_2026 schema flag and Anyscale wrapper properties (#62011)
Sai Miduthuri committed
G
[Data] Add file partitioning for DataSourceV2 [3/n] (#61997)
Goutam committed