data_split.py

import json
import random 
random.seed(42)
def split_data():
    f = open("./PsyQA_full.json")
    lines = json.load(f)
    random.shuffle(lines)
    train_data = lines[:int(len(lines)*0.8)]
    valid_data = lines[int(len(lines)*0.8):int(len(lines)*0.9)]
    test_data = lines[int(len(lines)*0.9):]
    print(len(train_data), len(valid_data), len(test_data))
    with open("train.json", "w") as f:
        json.dump(train_data, f)
    with open("valid.json", "w") as f:
        json.dump(valid_data, f)
    with open("test.json", "w") as f:
        json.dump(test_data, f)  
split_data()