I am running a Zero-Shot Text Classification workflow.
Progress is incredibly slow. It has been over an hour, and I am only 68% through the data.
CPU usage tops out at 70% (7950X, 16C/32T). Memory tops out at 25% (of 64GB).
I have since stopped the run and changed the Alteryx Memory limit from ~12GB to ~36GB (and restarting Alteryx), yet memory allocation remains at 24% instead of the expected >50%. (And yes, the setting is 36,864 MBs)
Unfortunately, I do not see any difference in performance. At this pace it will take almost 2 hours to complete the run.
Does AMP have a CPU core and thread limit regardless of system resources?
Being CPU bound (and potentially with an artificial cap for Desktop) makes the Text Classification tools unproductive and forces me to use a Python/CUDA based solution that can harness my GPU to accomplish the same task in a fraction of the time.
Will GPU integration be coming to Desktop anytime soon?
