3 ms·
GPU_s_, ideally high-end ones, though.
by johnisgood 2y ago
GPU_s_, ideally high-end ones, though.
- OutOfHere 2y agoHigh-end ones are needed more for large model training and unoptimized tuning, less so for inference alone. Most users would need just inference, at least in the beginning.