InfinSlice is the capability to go back in time and view your bucket data in any given period, i.e. a start time and an end time. This capability solves the following problems:
- If your deep learning algorithm is capable of continuous learning, you can take InfinSlices of data (perhaps every hour, or every minute) and incrementally train your model
- Use InfinSlice to grab slices of data since your production model was trained, and re-train your model code. Checking the score will tell you if your model is suffering from Data Drift
InfinStor Jupyterlab Sidebar InfinSlice
To use InfinSlice, choose a start time and an end time. Then press the browse button and select a bucket/path. Other operations such as transforms will now use this InfinSlice.
- Hint: Use small 15 second InfinSlices while developing your transforms, then apply the transforms to InfinSnaps consisting of a large number of objects