Quoted By:
>trying to train a lora on a chuuba
>be colab-let
>every lora i train is somewhat overcooked and affects my style mixes in some way
>try culling the dataset a bit more
>same issue
>try playing with the unet
>same issue
>try adding/reducing repeats and adding/reducing epochs
>same issue
>now on V9 of the lora
what kind of skill issue is this
