WT-solutions/Kratki-Istorii-100k
Kratki-Istorii-100k is a synthetically generated dataset (using INSAIT-Institute/BgGPT-Gemma-2-9B-IT-v1.0) of short stories (3-5) paragraphs, which a young kid should be able to understand. The simplicity of the language used makes it very suitable for training and studying the behaviour of really small Language Models (<500M parameters). The dataset consists of ~100k texts in Bulgarian. You can use the dataset via the HF interface: from datasets import load_dataset ds =… See the full description on the dataset page: https://huggingface.co/datasets/WT-solutions/Kratki-Istorii-100k.
Kratki-Istorii-100k is a synthetically generated dataset (using INSAIT-Institute/BgGPT-Gemma-2-9B-IT-v1.0)) of short stories (3-5) paragraphs, which a young kid should be able to understand. The simplicity of the language used makes it very suitable for training and studying the behaviour of really small Language Models (<500M parameters).
The dataset consists of ~100k texts in Bulgarian. You can use the dataset via the HF interface:
from datasets import load_dataset
ds = load_dataset("WT-solutions/Kratki-Istorii-100k")This dataset aims to democratize experiments and learning on Language Models in Bulgarian by providing a free to use simple dataset for the training of Small Language Models (SLMs). For a dataset suitable for Instruction fine-tuining, please refer to the complimentary dataset: WT-solutions/Kratki-Istorii-Instruct-100k.
Our work is greatly inspired by roneneldan/TinyStories.
license: mit ---