A question on model training

[joanna910225]

Hi, very nice work!

I used 30 Nobel images and captions as the follow for example:

Nobel_Prizes_style, minimalistic, line art, hand-drawn, bold black outlines, clean strokes, monochromatic outlines, simple design, flat color palette, sketch-like, ink illustration, classic portrait style, contour illustration, plain beige background, minimal background detail, single-subject focus, centered portrait, close-up perspective, symmetrical composition, professional and modern illustration, minimalist backdrop, no shading, no gradients, no textures, flat and simple colors, retro 2D illustration, stylized portrait with neutral tones, bold and clean design, award-style caricature, monochrome with subtle accents, plain light background, no distractions, the subject wears glasses with a thin frame and smiles faintly, Dressed in a casual shirt with a buttoned collar, the figure exudes a calm and intellectual demeanor.

I trained with 20 epoch with sdxl base model, however I still have very poor result with the workflow in your repo.
The result contains large areas of color on clothes.

My question is :

1. does the trigger words used in caption has to be the same as the one used in clip text in the workflow?
2. is there any problem with the caption I used on both style and character description?

Thank you !