loss converge #7
Description
Excellent work!
Recently, I want to adapt Janus-4o to my own dataset for I2I tasks (non-natural images). I'd like to ask, during your training of Janus-4o, to what extent did the loss typically converge (or what was the acc like)?