# Using Ray Tune with Distributed Training Causes Object Spilling

**URL:** <https://discuss.ray.io/t/using-ray-tune-with-distributed-training-causes-object-spilling/21923>\
**Category:** Ray Tune\
**Created:** [February 26, 2025, 9:03pm UTC](https://discuss.ray.io/t/using-ray-tune-with-distributed-training-causes-object-spilling/21923 "2025-02-26T21:03:40Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![ceruleanwarbler](https://avatars.discourse-cdn.com/v4/letter/c/4da419/32.png) [@ceruleanwarbler](https://discuss.ray.io/u/ceruleanwarbler)\
**Post date:** [February 26, 2025, 9:03pm UTC](https://discuss.ray.io/t/using-ray-tune-with-distributed-training-causes-object-spilling/21923/1 "2025-02-26T21:03:40Z")

</div>

Hey team, we have a `TensorflowTrainer` for distributed training that we plug into `tune.Tuner` and the TensorflowTrainer uses `datasets={"train": train_dataset}` where train\_dataset is a Ray Dataset. When I run concurrent trials with Tune, this results in object spilling. We’ve noticed less object spilling when the model is smaller, which makes it seem like perhaps TF objects are being spilled