datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
nepaliflow-romanized-nepali-to-devanagari-dataset
NepaliFlow Romanized Nepali to Devanagari Dataset
This dataset contains instruction-style examples for converting Romanized Nepali words into Nepali Devanagari script.
Task
The task is to convert a Romanized Nepali word into its Devanagari form while returning only the Devanagari output.
Columns
prompt: instruction asking the model to convert a Romanized Nepali word into Devanagari
completion: expected Nepali Devanagari output
Size… See the full description on the dataset page: https://huggingface.co/datasets/dipeshch71/nepaliflow-romanized-nepali-to-devanagari-dataset.cc100-nepali-strictly-cleaned-devanagari-only
CC-100 Nepali — Cleaned(Devanagari Only)
Pipeline
Unicode normalisation (NFC + ftfy)
Rule-based filters (length, Devanagari ratio ≥ 0.5, boilerplate)
Language ID — fastText lid.176.bin, confidence ≥ 0.7
Exact deduplication (MD5)
Near-deduplication (char 13-gram bloom filter)
98/1/1 train/val/test split, seed 42
Usage
from datasets import load_dataset
ds = load_dataset("Basanta55/cc100-nepali-strictly-cleaned-devanagari-only")