README.md

---
language:
- fi
pipeline_tag: text-generation
base_model: mpasila/gpt3-finnish-8B-gptq-4bit
license: apache-2.0
datasets:
- Finnish-NLP/Capybara-fi-deepl-translated-sft
- mpasila/Capybara-fi-deepl-translated-sft-alpaca
---

# Model Card for Capybara-Finnish-V1-8B

This is a merge of [mpasila/Capybara-Finnish-V1-8B-LoRA](https://huggingface.co/mpasila/Capybara-Finnish-V1-8B-LoRA/).

Base model used: [mpasila/gpt3-finnish-8B-gptq-4bit](https://huggingface.co/mpasila/gpt3-finnish-8B-gptq-4bit/) and the original unquantized model: [TurkuNLP/gpt3-finnish-8B](https://huggingface.co/TurkuNLP/gpt3-finnish-8B/).

Dataset used with the LoRA is [Finnish-NLP/Capybara-fi-deepl-translated-sft](https://huggingface.co/datasets/Finnish-NLP/Capybara-fi-deepl-translated-sft/) with some modifications so it uses Alpaca formatting [modified dataset](https://huggingface.co/datasets/mpasila/Capybara-fi-deepl-translated-sft-alpaca/).

It uses Alpaca format but with a translated instruction at the start:
```
{
    "instruction,output": "Alla on ohje, jossa kuvataan tehtävä. Kirjoita vastaus, joka täyttää pyynnön asianmukaisesti.\n\n### Instruction:\n%instruction%\n\n### Response:\n%output%",
    "instruction,input,output": "Alla on ohje, jossa kuvataan tehtävä ja joka on yhdistetty kontekstia lisäävään syötteeseen. Kirjoita vastaus, joka täyttää pyynnön asianmukaisesti.\n\n### Instruction:\n%instruction%\n\n### Input:\n%input%\n\n### Response:\n%output%"
}

```

Merged using this [Colab notebook](https://colab.research.google.com/drive/1a76Y21GfPtmVs71Uztlgk2xzPA4_vVjs?usp=sharing). It might not be the best way to merge a quantized LoRA on to a float16 model but I just wanted to quickly do something. You can try merging it better if you want.

### Framework versions

- PEFT 0.8.2