InfinSnap fine grained snapshots enable you to train your model against a frozen read only view of the data. This helps with the following problems:
- Even if the bucket contents change while training, InfinSnap ensures that the training software reads a frozen view of the data
- Model Provenance is facilitated since InfinSnap gives you the ability to go back and view the bucket data as it was when used for training
InfinSnap is useful for Data Provenance. If you want to recreat model artifacts, choose the same InfinSnap and run the same set of transforms to generate the exact same artifacts.