eyousefifar/

Parto-Persian-v1

Parto is a large-scale, synthesized dataset of Persian instruction-answer pairs, designed for training conversational and instruction-following models.

Language: Persian
Rows: 4.49M
License: mit Splits: 80% train, 20% test
Total Tokens: ~1.78B
Average per row: ~397 tokens

from datasets import load_dataset ds = load_dataset(eyousefifar/Parto-Persian-v1, split="train")

@misc{{Parto_Persian_v1, author =… See the full description on the dataset page: https://huggingface.co/datasets/eyousefifar/Parto-Persian-v1.