You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I already managed to configure everything, and the pre-training process starts without problems.
I noticed that by modifying TRAIN_STEP and eval_period I can control the artificial epochs and steps. However, I found that when I use a smaller or very large data set, the numbers of steps and epochs are the same.
Does it mean that in case the dataset is small, the examples are iterated multiple times to reach the total number of train steps, and if the dataset size exceeds the number of steps, some examples are not used?
The text was updated successfully, but these errors were encountered:
Hello,
I already managed to configure everything, and the pre-training process starts without problems.
I noticed that by modifying
TRAIN_STEP
andeval_period
I can control the artificial epochs and steps. However, I found that when I use a smaller or very large data set, the numbers of steps and epochs are the same.Does it mean that in case the dataset is small, the examples are iterated multiple times to reach the total number of train steps, and if the dataset size exceeds the number of steps, some examples are not used?
The text was updated successfully, but these errors were encountered: