why not evaluate the metrics on each test set #5
Description
I don't quite understand. Since multiple datasets have already been trained on, why not evaluate the metrics on each test set? Why take out only a few tasks for evaluation?
{{ message }}
I don't quite understand. Since multiple datasets have already been trained on, why not evaluate the metrics on each test set? Why take out only a few tasks for evaluation?