Heidrun-Mistral-7B-base is a generative text model based on Mistral-7B. It has been further pretrained on a subset of the Danish corpus from Wikipedia, Wikibooks and small parts of Hestenettet for 2 epochs.

It is a foundational/completion model with potential for further finetuning.

For inference or chatting please check out Heidrun-Mistral-7B-chat.

Please note that this has been updated since the original release. The old version can be found under branch v0.1.

This mistral model was trained 2x faster with Unsloth and Huggingface's TRL library.