quantization of occiglot/occiglot-7b-de-en-instruct - which was trained on German and English and code data, with 180M tokens of additional multilingual and code instructions