The predict subcommand allows you to make bulk JSON-to-JSON predictions using a trained model and its Predictor wrapper.

$ python -m predict --help
usage: run [command] predict [-h] [--output-file OUTPUT_FILE] [--print]
                            archive_file input_file

Run the specified model against a JSON-lines input file.

positional arguments:
archive_file          the archived model to make predictions with
input_file            path to input file

optional arguments:
-h, --help            show this help message and exit
--output-file OUTPUT_FILE
                        path to output file
--print               print results to stdout