Yes, you can do that. If your test file has the correct labels (defined values of your target variable) you can use it for evaluation purposes, where your trained model is tested against this test file. If the file does not contain labels, you can use it to create a batch prediction that will generate a label for each instance using your trained model. Just make sure that your test file has the same structure (columns/data fields and data types) as your model’s underlying dataset.