Eric Liang

## Why are these changes needed? Tune's bread and butter use case is parallel execution of trials, but we don't have any such guide for this. Add this as the...

## Why are these changes needed? This PR adds additional breakdowns to track Ray task state when (1) blocked on fetching dependencies in the object manager, and (2) blocked on...

[WIP] Bulk executor initial implementation

1

## Why are these changes needed? Initial implementation of https://github.com/ray-project/enhancements/pull/18 Original prototype: https://github.com/ray-project/ray/pull/30222/files TODO: - [x] Implements block splitting support - [x] Implement legacy plan compatibility - [x] Implement stats...

[core] Object store memory leak in simple while loop

1

### What happened + What you expected to happen Running the following script leads to an infinite object store memory leak and disk spilling. Interestingly, `ray memory` doesn't show any...

[data] Improve str/repr of lazy Datasets

1

### Description Currently, a lazy Dataset's string repr shows something like this: ``` >>> ray.data.range(10).map(lambda x: x).filter(lambda y: y) Dataset(num_blocks=10, num_rows=?, schema=Unknown schema) ``` This doesn't provide a lot of...

[data] Spread map task stages by default for arg size <50MB

4

## Why are these changes needed? 1. Change the default strategy for map tasks to SPREAD for arg size

[WIP] Prototype splitting blocks to ensure requested read parallelism is always met

1

## Why are these changes needed? Today, the number of initial blocks of a dataset is limited to the number of input files of the datasource, regardless of the requested...