metadata
configs:
- config_name: default
data_files:
- split: train
path:
- text_adventures_deduped.json
SpringDragon text adventure dataset courtesy of Henk717 (thanks Henky!)
Files in this repo:
text_adventures.txt: the original SpringDragon dataset.text_adventures_deduped.json: SpringDragon converted to a json completion set with duplicate samples removed.
The json format is preferred if you're training with axolotl, which parses each line of the text file as separate samples.