polish-gpt2-small-v2/README.md

---
license: cc-by-4.0
datasets:
- clarin-knext/msmarco-pl
- clarin-knext/nq-pl
- clarin-knext/hotpotqa-pl
- clarin-knext/scidocs-pl
- clarin-knext/nfcorpus-pl
- clarin-knext/dbpedia-pl
- clarin-knext/trec-covid-pl
- clarin-knext/quora-pl
- clarin-knext/arguana-pl
- clarin-knext/fiqa-pl
- radlab/wikipedia-pl
- radlab/legal-mc4-pl
language:
- pl
library_name: transformers
tags:
- gpt2
- from-scratch
- polish-gpt2
---

## Description
This is the polish gpt2 model in small architecture. 

This model was released on 30.11.2023 and it is the newest version of `radlab/polish-gpt2-small` (https://huggingface.co/radlab/polish-gpt2-small)


## Datasets
Data which are used to train this model:
- clarin-knext/msmarco-pl
- clarin-knext/nq-pl
- clarin-knext/hotpotqa-pl
- clarin-knext/scidocs-pl
- clarin-knext/nfcorpus-pl
- clarin-knext/dbpedia-pl
- clarin-knext/trec-covid-pl
- clarin-knext/quora-pl
- clarin-knext/arguana-pl
- clarin-knext/fiqa-pl
- radlab/wikipedia-pl
- radlab/legal-mc4-pl
- own corpora not published yet

It is about 30,5 GB of data which is 3 times more than the prevoius version.


## Metrics from W&B


![image/png](https://cdn-uploads.huggingface.co/production/uploads/644addfe9279988e0cbc296b/BBhEM3RL4rM1rOwTG-1RC.png)


![image/png](https://cdn-uploads.huggingface.co/production/uploads/644addfe9279988e0cbc296b/Z7BKY2vKPXWkQZEybt2rW.png)

![image/png](https://cdn-uploads.huggingface.co/production/uploads/644addfe9279988e0cbc296b/upsovWy0lVohPGjR1AGmN.png)

## Changelog
- _2023.11.30_ - new dataset