I wanted to start a broader mistakes and lessons thread around GPUs for Local AI.
A useful angle is best workflow, but it also connects with productivity. Different tools and workflows can produce very different results even when the goal is similar.
It would be interesting to compare...