Datasets
Description: Explains the concept of dataset objects in Valohai, and how they're used in experiments and pipelines.
Datasets are immutable collections of files.
You can create new versions of datasets, (optionally) based on previous ones, but you cannot change the existing ones. This offers full reproducibility, in the terms of the data used for training (or executions in general) - results produced by the execution in which datasets are used as a method of selecting the input files, will always yield the same results.
In this section you can learn how to:
Last updated
Was this helpful?
