metadata
license: pddl
language:
- en
tags:
- synthetic
size_categories:
- 1M<n<10M
This dataset is around 2.4M~ tokens (estimated) and consists of two computer generated prompts and a variety of stories. The quality of these stories may vary, and do display LLM-isms, such as extensive markdown formatting. This dataset is meant for training small transformers and was generated over an hour or so on a RTX 5070 from the Qwen3.5-0.8B model.
Each prompt was generated from a template, and is pretty simple. I will be releasing the synthetic dataset tool I made to create the dataset, eventually. Most prompts include a variety of adjectives for the story and type of story as well as various english cities for the story to be based in.
This dataset is released to the public domain, for any use, for whatever reason.