# Confused about prediction file size and test set

**URL:** https://forum.crunchdao.com/t/confused-about-prediction-file-size-and-test-set/907
**Category:** ADIA Lab
**Created:** [July 5, 2025, 5:12pm UTC](https://forum.crunchdao.com/t/confused-about-prediction-file-size-and-test-set/907 "2025-07-05T17:12:36Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![technical-sara](https://forum.crunchdao.com/user_avatar/forum.crunchdao.com/technical-sara/32/683_2.png) [@technical-sara](https://forum.crunchdao.com/u/technical-sara)
#### Post date: [July 5, 2025, 5:12pm UTC](https://forum.crunchdao.com/t/confused-about-prediction-file-size-and-test-set/907/1 "2025-07-05T17:12:36Z")

</div>

When I run my local pipeline using X\_test.reduced.parquet and y\_test.reduced.parquet, my prediction.parquet output always contains 101 rows, matching the reduced test set. However, I was expecting to generate predictions for the full test set (10,000+ time series), but I can’t find a full version of X\_test.parquet or y\_test.parquet in the data files I have. Will I get the accurate ROC AUC score from the full y\_test,x\_test only if I upload my code and model ?

---

<div class="post-metadata">

### Author: ![enzo](https://forum.crunchdao.com/user_avatar/forum.crunchdao.com/enzo/32/15_2.png) [@enzo](https://forum.crunchdao.com/u/enzo)
#### Post date: [July 5, 2025, 7:25pm UTC](https://forum.crunchdao.com/t/confused-about-prediction-file-size-and-test-set/907/2 "2025-07-05T19:25:17Z")

</div>

That is correct, locally you are only able to access a (very) small subset of the dataset and you must run in the cloud to make sure you code indeed run.