Repeat Label

This is an agent which simply repeats the label for debugging purposes or as a baseline.

Basic Examples

Evaluate on bAbi, returning the answer “I don’t know” roughly 20% of the time.

python examples/eval_model.py -m repeat_label -t babi --cant_answer_percent 0.2

RepeatLabelAgent Options

RepeatLabel Arguments

--return-one-random-answer

Return one answer from the set of labels

Default: True.

--cant-answer-percent

Set value in range[0,1] to set chance of replying with special message

Default: 0.

--cant-answer-message

Message sent when the model cannot answer

Default: I don't know..