Qwen lora training isn't going well.
Left no lora right with lora
Lora was trained on 3k steps 0.0001 LR (default ai toolkit settings)

prompt - "This image is of the character Red Hood. This is a digital anime-style illustration featuring a female character with a striking, confident pose. She has long, flowing red hair and is wearing a black and red outfit that includes a tight, black, glossy leather bra accentuating her large breasts, and matching thigh-high boots with red accents. She also has black straps and belts around her waist and thighs. Her outfit is complemented by a red scarf flowing behind her. She is holding a large, futuristic black gun with red details over her right shoulder. Her skin is fair, and she has a playful, slightly mischievous smile on her face. Her eyes are a bright red, matching her hair. The background is plain white, which makes the character stand out prominently. The overall style is dynamic and energetic, with a focus on bold colors and sharp lines. The character's confident pose, with one leg raised and her body slightly twisted, adds a sense of movement and action to the image. The illustration emphasizes a combination of sensuality and strength, typical of many anime and manga character designs."

The prompt was in the training data from auto captioning I did