How does validation loss interact with repeat?

_purple is repeat=2, pink is repeat=1_

- I was expecting the validation loss for 1 repeat to be similar to the stretched of 2 repeat, but the turn out that the validation loss of 1 repeat is more like extended of 2 repeat.
- Is this behavior expeted?
- If this behavior is expeted, shouldn't we always use 1 repeat to save time.

Here is validation config
```
validation_seed = 1
validation_split = 0.07
num_repeats = 2
# num_repeats = 1
```

Start from the validation configuration in the issue, comparing runs with num_repeats set to 1 and 2 and reviewing how validation loss is calculated. Determine whether the observed curves are expected and document the answer, including whether using one repeat is valid. No source files or tests are identified in the issue.

Written by the indexing model from the issue text.