Please duplicate this space for GPU access

On a CPU only space, inference is slow whereas even on the cheapest GPU hardware available on HF spaces you get quick replies!

(Recommended) Swap the aggregator to any large model of your liking on HF!

Currently, in order to accomodate free tier space capacity both the aggregator and reference models are on the lighter side (yet still very performant!). A more performant aggregator yields better results!