I want to train a model to perform a multioutput regression on an image, the output should be 4 float values (between 1 and 0).
Do you have any suggestions on how to do so?
One possibility I was thinking was to pop up the last layer of a classification transformer model and put some regression on top. does it make any sense? is there any tutorial to do so?
Also, I see the trainer from HF has only one output value, is there any possibility to use it to do multioutput regression?