Datasets:

dofbi

/

jolof

like 5

Hello @dofbi , to get a preview of this, I’m sharing the code below where we need to proceed in three steps:

#pip install datasets jsonlines

import jsonlines
from datasets import Dataset

# Load the data from the jsonl file and make two lists for the wolof and french sentences
french = []
wolof = []

with jsonlines.open('dataset.jsonl') as reader:
    data = list(reader)

for line in data:
    wolof.append(str(line['input']))
    if line['output'] is not None:
        if isinstance(line['output'], dict) and 'definition' in line['output']:
            french.append(str(line['output']['definition']))
        else:
            french.append(str(line['output']))
    else:
        french.append(None)  

if len(french) != len(wolof):
    raise ValueError('The number of french and wolof sentences are not equal')

# Build the dataset from the lists then split it into train and test sets
ds= Dataset.from_dict({"wolof": wolof, "french": french})
ds.train_test_split(test_size=0.2)

# We push the dataset to the hub
ds.push_to_hub("dofbi/jolof")

Hello @abdouaziiz
Thank you for this contribution
Dataset Viewer run at Dataset Card

But not running in Viewer