Testing a Deployment

After creating a deployment (deployment used as an example: 'Predict My Age'), Deeploy gives the possibility to test the deployment by interacting with it. In order to do so, you navigate to the Deployment Details by clicking View (see figure below).

Next navigate to the Interact tab (see figure below). Below the different elements of the page are explained:

• If you have added example input during the deployment creation, you can use the example input to test the deployment by clicking + Use example input. More information about the Data Payload can be found here. 
• In the textbox an example data payload is displayed with the tensor dimensions as placeholder.
• If the checkbox Save log is selected the test prediction will be saved in the prediction logs. More information about the prediction logs and monitoring a deployment can be found here.
  
  
• When clicking Predict a test prediction is made (see figure below). In this example the output tensor represents a probability distribution over categories, here: age categories.
  
  

To monitor the quality of the model, you can collect the Actual value of a prediction and set up a feedback loop in case the actual values are not available.

Up next: Integrating a Deployment