TinyStoriesV2-GPT4-train.txt - Is a new version of the dataset that is based on generations by GPT-4 only (the original dataset also has generations by GPT-3.5 which are of lesser quality). It contains all the examples in TinyStories.txt which were GPT-4 generated as a subset (but is significantly larger).
This dataset was used to train
https://github.com/noanabeshima/tiny_model/.
The data was preprocessed with:
from… See the full description on the dataset page:
https://huggingface.co/datasets/noanabeshima/TinyStoriesV2.
