UTA | Onepiece

This is a character model of Lora from the One Piece series in the context of uta.

You can use "utaonepiece" as the trigger word.

I have released two versions of the model here.

The first version has been trained for 30 epochs, while the second one has been trained for 100 epochs.

The second version is highly overfitted and will be very close to the original dataset for some specific prompt combinations, with less variation. However, at the same time, it retains the character's features better in other cases.

I've tried both models, and I think each has its merits. It's up to you to decide which one to choose. For the V1 model, I recommend a weight of 0.6-1, and for the V2 model, I recommend a weight of 0.4-0.8. However, please note that this is not absolute.

In actual applications, you can adjust the weight according to the situation. If the model's pose, screen, and other elements are too fixed, you can try reducing the weight. If the model doesn't retain uta's features well, you can try increasing the weight.

In summary, if you're not satisfied with the drawing, you might try adjusting the weight for better results.

这是一个海贼王里面uta的人物Lora模型。

你可以使用utaonepiece作为触发词。

这里我发布了两个版本的模型。

第一个版本是训练了30个epoch的版本，而第二个是训练了100个epoch的版本。第二个版本会非常过拟合，对于一些指定的prompt组合会非常接近原始数据集，相对来说变化程度较少，但是同时它在其他情况下对人物的特征保留也更好。

我尝试了两种模型，我觉得各有千秋。因此具体选择哪一个由您自己来决定。对于V1的模型，我推荐权重为0.6-1，对于V2的模型，我推荐权重0.4-0.8。但是注意，这并不绝对。

在实际应用中，您可以根据情况来调整，如果模型的姿势画面等元素太固定，您可以尝试降低权重，如果模型对于uta本人的特征保留的不好，您可以尝试加大权重。

总而言之，如果画出的图不是很满意，不妨尝试调整一下权重，也许会有更好的结果。