For the sake of interest, I trained the hypernetwork on the same dataset as LORA. The result was extremely interesting and beautiful.
I used the same promts for clarity.
In general, the hyper network works well with values of 1, but also with values of 0.5 is also not bad.
For the sake of interest, I trained the hypernetwork on the same dataset as LORA. The result was extremely interesting and beautiful.
I used the same promts for clarity.
In general, the hyper network works well with values of 1, but also with values of 0.5 is also not bad.