Use "Bunnygirl" as base keyword.
I trained Carrot on both normal and sulong form, try adding keyword "blonde" for normal form and "white fur" for sulong. Also try clip skip 2 for anime style as fullclip is also tuned towards 3d style images.
Wish I could train better keywords for my LoRAs but my GPU does not have enough VRAM to actually train the text encoding.