ModelsLab/Astigmatism (Formerly 'Semantic Shift') - Astigmatism 0.4b

This is the more fleshed out version, however there was a single abberant image in the dataset which occasionally can make it obsessed with nuclear apocalypse, and as such I included 0.4(a) as well.Additionally, its my personal philosophy that, if you have multiple loras that go after the same thing, then using as many as possible at low weights is more likely to "average out" overfitting, as opposed to trying to get rid of it by cleverness. Sometimes, two gentle hands gives better results than one strong one.