Upload training data

You can upload training data to use as a baseline window for drift policies as well as for explainability.

1. Upload your training data to VIANOPS cache. Note: If you want to load a file from a URL, skip to the next step (dataloading) and enter the URL in the srcurl field.
   1. With the /v1/cache/upload endpoint, enter the file with the file body parameter.
   2. Run the endpoint. The endpoint returns a key that you use in the next step to move the file from the cache to the data store.
2. Run dataloading with the cache upload key or with a URL to the file.
   1. With /v1/dataloading/submit, enter or verify the following fields:
      • filetype — Parquet (default), JSON, or CSV.
      • separator — Enter the appropriate separator for you file type. Comma is default.
      • setname — Enter a setname, which you use to reference the data later.
      • tablename — Enter the same string as setname. You also use it to reference the data later.
      • srcurl — If you are loading a file from a URL, enter the address here.
      • upload_key — If you have loaded a file into cache, enter the key here returned from the cache call.
      • connectionname — Must match one of the connections already in the data source connection list.
   2. Adjust any other option as needed.
   3. Submit the endpoint. VIANOPS builds a table in the data store with your data.

3. Add the model tag to the notebook.

   In the cell that defines the model_tags_payload variable, add the following object to the list:

   {"name": "train_set", "value": "TABLE_NAME", "status":"active"},

   where TABLE_NAME is the name of the table (tablename) you specified when dataloading in the previous step.

   For example:

   

4. VIANOPS activates the UI elements for choosing your training data as a baseline for policies or for explainability.
   • Policies
   • Explainability