lmg-anon/9nine-llama2-7b-experiment-qlora

QLora trained for 1 epoch using all the lines of the Visual Novel "9-nine-".

I'm uploading this mostly to document this experiment, the fine-tune itself isn't that great because it lacks linearity sometimes (probably because of non-linear nature of the dataset), but it did seem to capture the writing style of the VN lines.