Ruiyang Wang

Results 78 comments of Ruiyang Wang

[Core] Handle `TypeError` when `RayTaskError.cause` is a `BaseExceptionGroup`

cc @jjyao to take a look and merge. Also we may want to support ExceptionGroups later.

[runtime_env] unify `container` and (new) `image_uri` under plugin mechanism

The name `image_uri`, and its arguments being a plain string is IMO too specific. If later we have some more options, for example some CPU and mem limits, the name...

[runtime-env] automatically infer worker path when starting worker in container

Is this an API break? i.e. `container.worker_path` is no longer respected

[runtime-env] automatically infer worker path when starting worker in container

Can we make it backward compatible, you can choose either: 1. If user set a worker_path, use that. Else, use the inferred one. 2. If user set a worker_path, print...

[data] ray_tqdm does not work with numba

@c21 you are the last one touched that file. Would you mind take a look? Thanks.

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

## Status Report I made it to a point that it can replace *all* of the existing GcsAioClient, and the `internal_kv_*` part of the GcsClient, with some incompatibilities. ## Architecture...

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

`python/ray/serve/tests/test_gcs_failure.py::test_controller_gcs_failure[False]` failed only because interal kv get is expected : ``` except ray.exceptions.RpcError as e: raise KVStoreError(e.rpc_code) ``` so we need to move back all TimeoutError back to RpcError (sad)...

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

TODOs before merging this: - [ ] https://github.com/ray-project/ray/pull/45817 Reconcile PythonGcsClient error handling (e.g. timeout vs rpc error) - [ ] https://github.com/ray-project/ray/pull/46057 Change current get_all_node_info to return `Dict[NodeID, gcs_pb2.NodeInfo]` - [...

[core] Add convenience method `RAY_LOG(INFO).WithField(job_id)`.

auto merge?

[core] Add convenience method `RAY_LOG(INFO).WithField(job_id)`.

can't do this - id.h depends on logging.h, which can't depend on id.h. Tried to forward declare but also failed.