新手第一次訓練，感覺大概能用有60分，沒使用正則化，出圖不太穩定，測試了大部分XL大模型只有5個破圖無法兼容，通用性還可以。出圖小特徵有問題局部重繪0.75 幾次就能修正。訓練模型：SDXL1.0官方模型建議權重：1觸發詞：無 ，參數參考例圖。Training model: SDXL1.0 Official ModelRecommended weights: 1Trigger Words:  no Trigger Words推薦 PROMPT:1girl, solo:1.3,smile, two cat_tail,official, head, green eyes,animal ears, ahoge, long straight hair, ponytail, hair ornament,bare shoulders, black crop top, small breasts, navel, detached sleeves, elbow gloves, fingerless gloves, skirt,如果破圖可以使用LoRA Block Weight，去除過擬合。If there's an issue with the image, you can use LoRA Block Weight with parameter references to reduce overfittinglbw:1,0,0,0,0,0,1,1,1,1,1,1,1,1,1如果身體比例變形，可以改變圖片比例或使用ControlNet 。If the body proportions are distorted, you can adjust the image ratio or use ControlNet.貓掌不穩定，有需要可以用 cat paw,paw，局部重繪有機會出來。The cat's paws are unstable, and if necessary, you can use cat paw or paw, with the possibility of partial redrawing.

Files with SHA256: 149b7ef77c51384ce1db069938820531c6598f556716db4950fa654b659d7dac