Retrained with larger dataset and different trigger words.

Changed several trigger words:

1. Changed black gloves to forearm gloves. Since those aren't elbow gloves nor regular short gloves, forearm gloves is the best way to describe them, and it turns out to be quite good with the results.

2. Changed black footware to heelless heels. Now the Lora can recreate her heelless heels, although the result isn't consistent all the time since the dataset is insufficient to make the Lora completely learn the pattern. And this trigger word may produce pics with extra legs or weird gesture.

Less overfitting comparing to V1.

It is such a headache to train this Lora to a useable state with her heels put into consideration. And this version is still not in a state of what I want it to be, so I'll most likely come back again for another retrain when she's officially out in game.