I have a serie of things to predict, but Im doing a single
for, also I see that each predictions shows the loader taking vertical space and it seem that is using a lot of CPU and little GPU (how I know my predictor is running on GPU and if it is, how to use more GPU).
I have two questions regarding this:
- Because it is sequential, should I just spawn some threads? how you do it?
- Instead of a progress bar by prediction, I would love that the loader keep track of all the predictions showing progress over the total batch of predictions.