matthewdeng

### What happened + What you expected to happen When I run an experiment with the GCS Bucket as the `upload_dir`, I can successfully restore the experiment. ```python RunConfig(sync_config=SyncConfig(upload_dir=BUCKET) ```...

Numpy 1.24 Support

1

## Context Numpy [1.24.0](https://github.com/numpy/numpy/releases/tag/v1.24.0) was released on 12/18/22 and contains some breaking changes. This ticket tracks the changes needed to ensure Ray is compatible - full compatibility is targeted for...

DeepSpeed ZeRO3 errors on config initialization

9

### System Info `transformers-cli env`: - `transformers` version: 4.37.2 - Platform: Linux-6.2.0-1017-aws-x86_64-with-glibc2.31 - Python version: 3.9.18 - Huggingface_hub version: 0.19.4 - Safetensors version: 0.4.1 - Accelerate version: 0.26.1 - Accelerate...

sklearn benchmark

1

Create a benchmark comparing the performance of ray-skorch to regular skorch.

## Why are these changes needed? More recent versions of `jax` (e.g. `0.4.28`) will cause the following problem: ``` ╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ /home/ray/default/ray/doc/source/templates/05_dreambooth_finetuning/dreambooth/generate.py:9 in │...

1. Added GPU filtering: - Added `GpuIndex` and `GpuDeviceName` variables to dashboard - Updated GPU panels to use these variables - Variables hidden by default 2. Improved Train metrics: -...

## Changes - Made `TrainRunContext` frozen and added all training inputs (configs, datasets, etc.) - Changed `TrainContext` to contain `TrainRunContext` instead of inheriting from it - Added `after_controller_finish` callback to...

[train] Simplify `test_transformers_trainer_steps::test_e2e_steps`

1

#35270 causes importing `ray.train.huggingface` to import `accelerate`, which takes approximately 4.5 seconds. Since this test is parameterized to run 20x times, this causes the total time to run this test...

## Description Replace GPy with scikit-learn's `GaussianProcessRegressor` in the PB2 (Population Based Bandits) scheduler. ## Additional information - Rewrote `TV_SquaredExp` kernel to implement sklearn's `Kernel` interface instead of GPy's `Kern`...