Recommended weight： 1V1.5 Recommended weight： 0.5~0.8Trigger word：<pre><code>dusk \(arknights\), bare shoulders</code></pre>The "bare shoulders" prompt is mainly to keep the outfit closer to the original design. If you want to change the outfit, you can skip this option.I tried training a Flux LoRA with my RTX 3060 and encountered quite a few challenges, but the results turned out pretty well, although some details were missing. I believe this is due to insufficient preparation of the training dataset.Additionally, if you're looking to try local training and run into some issues, and you happen to have the same GPU as mine, the RTX 3060, feel free to leave a comment below. Maybe my experience can help you out.Feel free to leave feedback in the comments below, or share images generated using this model!_________________________________________________________________________________________________________建議權重：1V1.5 建議權重：0.5~0.8觸發提詞：<pre><code>dusk \(arknights\), bare shoulders</code></pre>bare shoulders主要是讓服裝更符合原本。如果要更換服裝可以不加。嘗試用我的RTX 3060訓練Flux LoRA，遇到不少麻煩，但結果還不錯，只是缺失一些細節，這部分我認為是訓練集準備的不夠好。目前還是建議本地訓練如果vram不夠的話還是線上訓練，例如civitai或使用雲算力就好，不然訓練時間相當可觀，整體時間成本不大符合。當然，如果你想嘗試本地訓練但遇到一些麻煩，顯卡也剛好跟我一樣是RTX 3060，歡迎在下方留言，或許我的經驗能幫你解決。歡迎底下留言反饋，或是用這模型產出圖片返圖!

Files with SHA256: 48311f68c554455b7732e722e562e601af32612b7964ab67fe45807644372239