There is a difference between the model and the pointer. The option --output-model FILENAME in the train-topics command creates a file (model) that contains the training data along with all the output parameters. --inferencer-filename FILENAME creates a much smaller file (infernizer), which contains only the parameters necessary to display themes for new documents.
The infer-topics command needs a pointer object, not a complete model. You can create a pointer from an existing model file by running train-topics with the --input-model FILENAME and --inferencer-filename FILENAME with --num-iterations 0 .
source share