--- license: apache-2.0 datasets: - cerebras/SlimPajama-627B - bigcode/starcoderdata - mc4 language: - fi - en - da - sv - 'no' - nn - is --- # Viking 7B Viking 7B is a 7B parameter decoder-only transformer pretrained on Finnish, English, Swedish, Danish, Norwegian, Icelandic and code. It has been trained on 2 trillion tokens. Viking 7B is a fully open source model and is made available under the Apache 2.0 License. Please see [the upstream repository](https://huggingface.co/LumiOpen/Viking-7B) for more information. This GGML quantization was done with [akx/ggify](https://github.com/akx/ggify) with llama.cpp b2901 with a small modification to the conversion script to support the Viking tokenizer.