| import random
|
| import pandas as pd
|
| from tqdm import tqdm
|
| from sklearn.model_selection import train_test_split
|
|
|
| df = pd.read_parquet('gemago2_dataset.parquet')
|
|
|
| texts = []
|
| for _, row in tqdm(df.iterrows(), desc="row", leave=False, total=len(df)):
|
| texts.append(rf"<kor>{row['korean']}</kor>\n\n<{row['language']}>{row['target']}</{row['language']}>")
|
| texts.append(rf"<{row['language']}>{row['target']}</{row['language']}>\n\n<kor>{row['korean']}</kor>")
|
| del df
|
|
|
| train_texts, test_texts = train_test_split(texts, test_size=0.2, random_state=42)
|
| del texts
|
|
|
| random.shuffle(test_texts)
|
| with open("test.txt", "w", encoding="UTF-8") as f:
|
| f.write("\n".join(test_texts))
|
| del test_texts
|
|
|
| random.shuffle(train_texts)
|
| with open("train.txt", "w", encoding="UTF-8") as f:
|
| f.write("\n".join(train_texts))
|
| del train_texts
|
|
|
| |
