KBlueLeaf / guanaco-7b-leh-v2

The dataset is from alpaca-cleaned and guanaco. With trained embed and head, the model perform better at Chinese and Japanese then original LLaMA, and with instruction based prompt. You can use this model more easily.