Trained on images from Tesla's recent Optimus gen-2 announcement video. Includes in the dataset some images of gen-1 as well (labeled gen-1, gen-2)
Over fitted towards gen-2, use negative gen-2 to get closer gen-1 outputs.
Trained for {color} strip to control the ambient face light color.
Description
Initial Training, lower network alpha, less predictable but more flexible
FAQ
Comments (2)
What's the difference between v0 (latest release) and v1 (first release)?
Alpha weight, V0 had a lower alpha which makes it less predictable but more flexible. V1 gets more consistent looking bots, but does not vary much from the gen-2.