multilingual-wikihow-qa-16k / source /merge_and_process_datasets.py
0x22almostEvil's picture
Brrrr
b65b873
raw
history blame
940 Bytes
import pandas as pd
n = 2
total_n = 16
# list of all the csv files
csv_files = [
f"en-wikihow-qa-dataset-{n}k.csv",
f"ru-wikihow-qa-dataset-{n}k.csv",
f"pt-wikihow-qa-dataset-{n}k.csv",
f"nl-wikihow-qa-dataset-{n}k.csv",
f"it-wikihow-qa-dataset-{n}k.csv",
f"fr-wikihow-qa-dataset-{n}k.csv",
f"es-wikihow-qa-dataset-{n}k.csv",
f"de-wikihow-qa-dataset-{n}k.csv"
]
# create an empty dataframe to hold the merged data
merged_data = pd.DataFrame()
# loop through all the csv files, read them into dataframes, and merge them
for file in csv_files:
data = pd.read_csv(file)
merged_data = pd.concat([merged_data, data], ignore_index=True)
# write the merged data to new csv and parquet files
merged_data.to_csv(f"multilingial-wikihow-qa-dataset-{total_n}k.csv", index=False)
merged_data.to_parquet(f"multilingial-wikihow-qa-dataset-{total_n}k.parquet", row_group_size=100, engine="pyarrow", index=False)