Change8

ray-2.56.1

📦 rayView on GitHub →
2 features🐛 10 fixes🔧 6 symbols

Summary

This release includes fixes for Ray Data's `to_pandas` regressions, early detection for system-slice memory pressure in Ray Core, and protobuf 7 compatibility with routing fixes for Ray Serve LLM direct streaming.

Migration Steps

  1. To opt-out of Arrow-backed pandas conversion, set the environment variable `RAY_DATA_ENABLE_ARROW_BACKED_PANDAS_CONVERSION=0` or use `DataContext.enable_arrow_backed_pandas_conversion(False)`.

✨ New Features

  • Ray Core: Added early detection for system-slice memory pressure. The memory monitor now snapshots user and system cgroup slices together and logs an error when the system slice exceeds reserved system memory, warning users to raise `--system-reserved-memory` before it causes node deaths.
  • Ray Serve: Added protobuf 7 compatibility.

🐛 Bug Fixes

  • Ray Data: Fixed two Arrow-backed `to_pandas` regressions: added `DataContext.enable_arrow_backed_pandas_conversion` as an opt-out, and reconciled divergent numeric column types before concatenation to avoid int64/`double[pyarrow]` overflow crashes.
  • Ray Data: Fixed a `TensorDtype.__from_arrow__` crash on zero-size tensor elements by using an explicit row count instead of numpy's `-1` dimension inference.
  • Ray Data: Fixed a crash in hash partition caused by read-only hash arrays.
  • Ray Data: Nullified `_input_dependencies` in `_get_args` so exporting operator args no longer triggers an exponential `sanitize_for_struct` call chain over fused operators.
  • Ray Serve: Added protobuf `>=7` compatibility to `_proto_to_dict` by binding to `FieldDescriptor.is_repeated` when the deprecated `label` attribute is absent.
  • Ray LLM: Fixed direct-streaming routing for body-aware routers: the ingress now parses the raw request body into a `SimpleNamespace` over routing-key fields (`messages`, `prompt`) so `choose_replica` receives the message body instead of raw bytes.
  • Ray RLlib: Upgraded the ONNX example from the retired MobileNet v1 to MobileNet v3 via `torchvision` and pinned `onnxscript` in the GPU/ml-build CI dep locks, fixing ONNX export failures.
  • Ray Core: Disabled NCCL `cuMem` host buffer registration in CI pytests to stabilize GPU test runs.
  • Documentation: Repointed the ASHA Tune example links to the renamed `README` document, fixing the `fail_on_warning` ReadTheDocs build on the release line.
  • Documentation: Updated the Python 3.10 CPU `pip freeze` dependency list for the Ray 2.56.0 release.

Affected Symbols