[Train] Shard all Input Datasets in Ray Train by default #37668
Labels
data
Ray Data-related issues
enhancement
Request for new feature and/or capability
train
Ray Train Related Issue
Description
Currently, Ray Train only shards the "train" datasets by default, leaving all other datasets unsharded. Users can configure the
DataConfig
to shard these other datasets.This is not satisfactory because:
We should consider enabling dataset shading by default.
Use case
No response
The text was updated successfully, but these errors were encountered: