Can I run all task in parallel during inference?

[wenglihong]

As the example given in readme, we should run each task once in the for loop since they have different mask, I wonder if I can run all the tasks in parallel during inference?