Herman is an Indonesian language dataset specifically designed
for training LLMs using a single-turn JSON mode. This dataset
is used in Supervised Fine-Tuning (SFT) to improve JSON parsing
capabilities in LLMs. Herman was obtained from Hermes and translated
into Indonesian for the purpose of training Indonesian language models.
Code used for constructing Herman can be found here.
The desired JSON schema can… See the full description on the dataset page:
https://huggingface.co/datasets/SulthanAbiyyu/herman-json-mode.
